Slot Gacor
Features Archives ✔️ News For Finance
Home Archive by category Features

Returnal is a sci-fi video game about a person constantly reliving the past in order to find a new future. In some ways, it feels like a pointed metaphor for the game’s creators.

PlayStation fans are likely familiar with Finnish game studio Housemarque, whose best modern games have masterfully combined classic arcade chops with modern flourishes. Yet even its biggest PS3 and PS4 games (Super Stardust HD, Resogun, Nex Machina) have mostly felt like translations from classic cabinets, thanks to fixed perspectives and allegiant action. Blow stuff up, aim for the high score, game over, and repeat.

This week, Returnal sees the studio aim its pedigree at a much higher scope: a game that combines the pure action of ’80s arcade games with the plot, production value, and world exploration of a full-blown “adventure” game. It’s as if someone at Housemarque looked at 1981’s Galaga running next to 2018’s God of War and said, “Can we somehow combine these two?”

The result feels like a statement game for Housemarque, arguably in the same way that 2019’s Control solidified Remedy Studios’ own reputation—though this effort isn’t quite as successful. At its best, Returnal delivers the studio’s finest-yet action and tension within a phenomenal 3D-shooting system. I’ve gone to sleep thinking about the game’s best blasting moments, eager to wake up the next day and return (returnal?) for “one more run.” Yet at its worst, Returnal‘s roguelite trappings sometimes threaten to bring the whole package down—especially if you’re not very good at high-speed shooter games.

Returnal will be some people’s favorite game of 2021. But even those players should prepare to strap in for a bumpy, weird start.

A Selene for every (deadly) occasion

Returnal product image

Returnal [PS5]

(Ars Technica may earn compensation for sales from links on this post through affiliate programs.)

Returnal stars a modern-day astronaut named Selene, whom players take control of the moment she crash-lands onto a mysterious planet named Atropos. You crawl out of your wrecked ship, get your bearings, and run (as seen from a third-person, over-the-shoulder perspective) to find a useful weapon… next to a dead astronaut with the same outfit and callsign as yours.

More dead Selenes appear, usually clutching personal audio recording devices that spoonfeed more of the game’s mysterious plot. You’ll add to that pile of corpses before long, since the opening tutorial segment includes a brutally difficult monster that traps you in a pit and kills in two hits. Immediately after your death, the screen flashes black, and the opening crash-on-a-planet sequence plays again with different camera angles; Groundhog Day stuff, but instead of “I’ve Got You, Babe,” your mornings always open with screams and smoke.

Your version of Selene remembers dying and coming back to life this time, but the world you’ve landed on looks different. The opening door reveals a new zone to run through. Different dead Selenes lie in different places (sometimes with new audio logs). Different rooms, lairs, and caverns appear, now full of new arrangements of enemies, items, and secrets.

Hence, we’re in roguelite territory, and the object is to die-and-retry while unraveling Atropos’ mysteries and finding a mix of temporary and permanent upgrades within every randomly generated sequence. That’s different than a roguelike, where each death starts you from scratch; roguelites bring you back to life with some upgrades remaining persistent after every death, while other stuff vanishes if you don’t use or spend it before you die. (The latter is much more common in the modern gaming era, with popular examples like Hades and Dead Cells; comparatively, Spelunky is the best example of a modern roguelike, if not the titular PC classic itself, Rogue.)

You’ll get stronger, you’ll keep going, you’ll keep dying

By the time you find that stronger monster again, you’ll have recovered a permanent “alt-fire” mode for your gun that shoots a charged, concentrated blast, along with a likely assortment of temporary upgrades. You’ll get stronger, in both permanent and temporary ways, and you’ll keep going.

But you’ll also keep dying.

Selene’s spacesuit has some nice batteries built in.

Returnal would rather you learn its systems within the course of the game’s die-and-retry loop, instead of breaking its brutality out to a tutorial. To some extent, I get it. This serves a plot that hinges on Selene’s confused dedication to breaking a time loop, and plot morsels emerge at a steady rate—which, I assure you, I haven’t spoiled here in the slightest. Still, you’ll have a better time if you understand the type of difficulty you’re getting yourself into.

Like Demon’s Souls before it, Returnal establishes a unique, tough-as-nails ruleset for combat, only this game’s take feels so much more like a Housemarque game. In a major departure from the Souls-like genre, Selene’s default movement speed is “damned fast,” and this is helped by unlimited “run” stamina (for even faster movement) and an instant dash-dodge button, which needs a second to recharge. In fact, there’s no “stamina” limitation anywhere in Returnal. Selene’s spacesuit clearly has some nice batteries built in.

This speed is imperative because players see Selene from a tight third-person angle as she runs through a mix of open fields and fallen-apart architecture. To live, you’ll have to keep moving, lest you get caught by waves of enemy bullets (usually large, slow-moving laser orbs) coming from all sides, along with foes that hunt for Selene’s body and pounce with melee attacks. To organically encourage your “git gud” mentality, Housemarque offers built-in help in the form of “adrenaline,” a meter that fills up every time Selene kills a foe without taking damage. Get your adrenaline high enough, and you’ll get perks. Some are combat boosts like increased melee damage (for the sword you eventually find), but others contribute directly to combat visibility. One perk makes ghosts of enemies appear if they’re behind cover. Another creates a warning radius around Selene’s body, lighting up to indicate whether you’re being targeted by bullets (white), turrets (purple), or pouncing foes (red) in any direction.

The only “wave” I will do

Hence, your path to survival is visibility, landscape mastery, constant movement, and strategic retreat—though like other modern action games, you’ll want to be close to your enemies’ corpses shortly after they die, since they spew a coin currency that you’ll need to purchase upgrades, health refills, and more during a given run. Kill enemies from afar and their coins will vanish by the time you reach them—which isn’t a novel concept in these kinds of games.

And getting into the weeds of how Returnal‘s individual parts resemble other roguelites could miss the larger point. Instead of describing game aspects in isolation—each gun, each power-up, each difficult min-max decision—I’d like to set the stage by describing one of my more successful gameplay runs.

I open the game with Selene once again gasping in the despair of dying and returning to the same stupid crash site, on the rainy, forested surface of Atropos. I have a few permanent upgrades at this point, including a melee attack and certain unlocked perks for the four guns I’ve found thus far: a pistol, a shotgun, a machine gun, and a weird rifle (the “Hollowseeker”) that shoots slower, weaker, heat-seeking shots than other guns.

Newer weapons may list perks with a percentage next to them. These indicate that you need to rack up a certain number of kills to unlock that perk, and doing this will make future guns you discover more powerful.
Enlarge / Newer weapons may list perks with a percentage next to them. These indicate that you need to rack up a certain number of kills to unlock that perk, and doing this will make future guns you discover more powerful.
Sony Interactive Entertainment / Housemarque

Every run starts with a base, perkless version of the pistol, and before long, I find other weapon options in randomly generated treasure chests. Some have perks I’ve already unlocked; others have grayed-out descriptions of perks I can unlock by killing enough foes. My first chest contains a Hollowseeker with a new listed perk called “wave.” I’m not a fan of this rifle yet, but I’d like to unlock and understand this perk. After getting through early rooms full of weaker foes, I learn that “wave” means “shoot big, slow, purple saw blades in addition to my standard firepower.” Wave is rad. Really, most of the perks are as rad, and they routinely put you in the head-scratching position of picking between guns. The shotgun, for example, is a tricky weapon when enemies float from afar, but once you perk it up with things like additional heat-seeking missiles and “stagger” abilities, it becomes a game-changer in some of the toughest battles.

In fact, in this run, I survive some of my toughest fights against hunt-and-pounce foes by staggering them, which drops them to their knees and exposes a weak point long enough for me to focus fire, kill them up close, grab their useful coins, and resume strafing-and-dodging other nearby pests.

When random isn’t random enough

This run might be my 12th time in the opening “Overgrown Ruins” biome, however, and things are starting to look familiar.

Roguelites revolve around premade building blocks for their stairwells, their treasure zones, their pools of lava, and so on. When a roguelite game is in 2D, those jigsaw puzzle pieces aren’t just simpler to jumble up; they’re also smaller. You might see tiny nooks reappear from one Spelunky run to the next, but generally, their block arrangements have a lot more leeway to work with—and benefit from a crapton of variables. You can mash together hundreds of 2D sprites without worrying about whether 3D geometry is going to intersect awkwardly or break the game.

I need that sweet, sweet adrenaline.

I wondered how Returnal would grapple with this as a fully 3D roguelite—a rarity in the genre—and the answer, as it turns out, is a bare-minimum effort. Atropos is largely made up of premade “rooms,” which I put in scare quotes because they’re marked on your handy mini-map with entrances and exits but can vary from constrained caverns to wide-open outdoor spaces. If you walk into a room and recognize its stairwell, then you’ll recognize the rest of it—every platform, every gap, every bottomless pit. This even applies to the very mild puzzles found in various rooms; the Ruins biome in particular is rich with a “shoot a hidden target to unlock a treasure chest” gimmick, yet it doesn’t use the roguelite system as an opportunity to randomize where the target is. It’s always in the same spot. Meh.

Instead, the game’s tension largely comes from wondering exactly what’s behind the next door. Will a room be full of enemies? Which ones and where? And what items are scattered around a room’s outskirts? This all works in service of Returnal‘s very satisfying combat—which, again, benefits from letting players enjoy freedom from stamina-related restrictions—and that means whenever you get into a fight, you can expect battles to play out in Housemarque’s finely honed mixes of cramped spaces, hidey-holes, open gaps, and temporary, destructible cover. In this particular run, I’m getting the added benefit of some zone mastery during these Ruins biome fights. “Over there’s the bottomless pit,” I remark as I dash around, knowing that an errant fall will cost me a bit of health and my entire accrued adrenaline meter. I need that sweet, sweet adrenaline to keep tabs on foes, after all. I’m doing better this time.

But I’m also a bit bored by the noncombat rooms’ familiarity. Was there really no way to move things around in the seen-this-before zones, Housemarque?

Building a machine in the Ruins

The beautiful thing about the run in question, however, is that I’d finally gotten a strong handle on the opening Ruins biome, along with some familiarity with the second “Crimson Wastes” biome. The latter opens the rooms up considerably and adds new and remixed foes—including a few fantastic air-controlling pests that take great advantage of newly opened skies.

Once you accrue enough permanent gear, Returnal offers players a few shortcuts when dying-and-retrying to skip ahead to later biomes, and part of me was eager to jump ahead to newer stuff. But for this run, I decided to explore every corner of the opening Ruins biome, since I had no idea which rooms, fights, and treasures it would hide. I did know I’d be more likely to accumulate useful temporary upgrades, along with a few more gun-specific perk unlocks, if I ran through the familiar biome one more time. (I also spent more time practicing Returnal‘s active-reload system, which largely resembles the one found in Gears of War and must be mastered to withstand the game’s worst fights.)

The result was exactly the build that roguelite fantatics drool over: a perfect “machine” of complementary bonuses, which came thanks to a mix of smart purchases and dumb luck.

Part of this was my selection of “mutations,” a risk-reward system that adds one benefit and one drawback. You’ll find mutations as alien blobs on the ground in some treasure rooms, and their pros and cons are clearly described via your HUD before you choose whether or not to equip them. I had also found an “artifact,” the game’s name for temporary, one-run-only bonuses, that gave me additional “repair speed” for every mutation I’d equipped, and another artifact that made each mutation’s positive aspect stronger. The negatives became noticeably less negative.

That repair-speed bonus coupled nicely with an artifact that automatically healed me whenever my health got low—which normally takes way too long to recharge. As described, this build’s mutations overclocked the auto-heal process, so whenever I took too much damage, my health would almost instantly refill to a small-but-doable amount. I had other perks that helped with things like shields and damage output, along with consumable items that (thanks to an artifact) would sometimes remain in my inventory after use. Suffice it to say, I’d finally stumbled upon a build that stared back at Returnal‘s high difficulty and growled, “Bring it.”

I began feeling confident. I leaned into each weapon’s special abilities. I got extra bandwidth to understand new enemies’ patterns and how multiple creatures’ AI might crescendo to rain lightning-orb terror on Selene if I didn’t react appropriately—and that bandwidth came as much from the run’s bonuses as from the die-and-retry learning I’d done up to that point.

Story and visuals

The run, sadly, ended with a wacky game crash (see above), but not before I’d reached an entirely new threshold of game progress. Returnal establishes solid stopping points within its campaign, and the biggest ones automatically wipe your temporary inventory, anyway, so there’s less inherent heartbreak no matter how far you get or how brutal your last death was.

Though the action is randomized, the plot is not, and you’ll follow Selene’s journey of interstellar discovery and existential dread via some trippy, haunting sequences. Occasionally, Housemarque’s Finnish take on an English-speaking hero tips its hat with awkward phrases and pronunciations, but it’s not enough to devalue one of the cooler repeating-loop stories I’ve seen in a long time—and certainly the best and most polished I’ve seen in a game. (If you’re wondering, this year’s Loop Hero comes in at second place.)

I can’t ascribe the same praise to Returnal‘s visual makeup, and the more I have played it, the more I’ve wondered whether Housemarque had originally planned this game for PlayStation 4 instead. Enemies look cheap and plasticky up close, and tight zooms on Selene’s in-game face, typically behind an astronaut helmet, are remarkably gaudy. Landscapes don’t teem with wind-swept plant life or other tiny, dense details.

And while the game’s official site mentions “ray tracing,” I have yet to see any particular in-game proof that either its reflections, shadows, or other lighting pipelines actually benefit from a next-gen light-bounce system. Instead, I’ve seen particularly unattractive reflections on some waterfalls and lakes, and absolutely no dynamic light bounces on others.

Perhaps most telling of all, PS5’s PCI-e 4.0 storage isn’t leveraged to render massive, unbroken worlds. Instead, most biomes are broken apart with doors or barriers, as if to help an older game platform (like, cough, PS4) contain each battling zone in a safe amount of system RAM.

The game’s PS5 benefits are mostly found in nearly instant loading times when warping between distant rooms and dense particle and alpha effects, evidenced by flashy enemy laser-orb attacks and rainy, windy weather systems. Perhaps the game will add “locked 60 fps refresh” to its visual bullet-point list at some point, but until I see the day-one patch, I’m left contending with a few inconsistent, sub-60 fps stutters during some fights in larger biomes. The slowdown I’ve seen thus far hasn’t been bad or frequent enough to criticize the game over—everything runs fast and fluid enough, and I’m not sure whether PS4 would’ve run the same game anywhere near 60 fps—but Returnal doesn’t look next-gen enough to make me shrug off those stumbles, either. Since Returnal prioritizes the moments when attacks fly fast and Selene dashes madly, I’m ultimately in the thumbs-up column about how it looks and runs.

This zigs where Hades zags—and that’s fine

I must admit, I haven’t beaten Returnal yet. The game is freaking hard, for one, and even upon completing the campaign, I imagine Returnal will offer other ways to continue engaging over time. As one early example, a “daily challenge” run gives players a static perk loadout, then asks them to play through a specific “seed” of combat-filled rooms and compete for a spot on the day’s leaderboards. It’s a nice combat-forward option to help new players experiment with higher-powered loadouts and battle through smaller stakes, and I’m glad this unlocks quite early in the campaign.

But my 24 hours of play, in spite of being unceremoniously interrupted by crashes, have convinced me that this is a fine addition to the roguelite pantheon, even if it’s not quite up to the same caliber as Hades, last year’s Ars Game of the Year.

Hades‘s Greek-god trappings, humor-filled plot, and gorgeous art helped a lot, but its systems truly made it a tantalizing over-and-over roguelite. You could aspire to tailor your runs to certain play styles but still be left with compelling min-max decisions all along the way as other options and power paths shouldered in. These were combined with six fantastic weapon options—and each had its own variety of sub-styles, along with a solid rewards system encouraging players to try each as they moved into endgame mastery.

Hades‘ results tapped more successfully into the genre’s endorphin-producing potential than anything I’ve played before or since. Returnal is clever enough with its randomized min-max decision trees, thanks to systems like the aforementioned mutations and a mix of temporary and permanent currencies that drive a number of “Ugh, should I spend my cash?” decisions along the way. But on a systems level, it’s not surpassing what Hades got so right.

Returnal launch trailer

Instead, Returnal stands out thanks to how its brilliant 3D combat is bolstered by the unpredictability of its randomized worlds—and how Housemarque somehow managed to marry that content with a cinema-caliber story. Sadly, the studio left too much randomization off of Returnal‘s table, arguably in service of the plot. Additionally, the game often beats its players silly before they become “good enough” to see its most compelling plot twists—and how those twists affect Returnal‘s worlds and mechanics.

I really like Returnal, but if you check the Ars Slack logs, you’ll find that I complained quite a bit along the way. I needed a full 10 hours for its combat and universe to click in a crucial, “I want to beat this game” way, and I’m still left wondering how many good ideas and systems were left out of this game just to get its sky-high aspirations out the door. Maybe some of my positive bias comes from dreams of a sequel, which might build upon Housemarque’s first stab at the genre. But I won’t blame anyone for having less patience with Returnal‘s uneven ambition (or its $70 price point, which, from what I’ve seen, does not favorably compare to last year’s $60 Last of Us Part 2 or Ghost of Tsushima, also published by Sony).

But this is the stuff that keeps Sony fanboys drooling: ambitious new IP that succeeds more than it fails while turning the familiar into something fresh. Returnal clearly heralds a new era for Housemarque, in terms of turning the focused arcade-blasting likes of Super Stardust HD into quest-worthy 3D action. Keep it coming, Sony and Housemarque.

The Good:

  • Housemarque’s arcade-action pedigree scales tremendously to 3D combat against waves of enemies and their lasers, in ways that feel dramatically different than other modern third-person shooters.
  • Enemy variety lays it on thick in terms of tricky attack patterns and AI—and these organically and fairly ramp up, no matter how difficult the results may be.
  • Die-and-retry system, as bolstered by a dense soup of temporary and permanent buffs, ultimately serves the game’s battling mechanics and tension, even if it’s not your cup of tea.
  • 3D positional audio and a sensible “adrenaline” meter keep players keenly aware of nearby threats—which they can whimsically dodge and dance around thanks to the lack of limiting “stamina” meters.
  • Masterful weave of compelling time-loop plot into an otherwise randomly generated game.
  • Daily challenge system adds powerful, quick-burst play as a palate-cleansing alternative to this game kicking your butt.
  • I don’t want to spoil them, but the boss fights are redonkulously good. This is the first time I’ve ever typed “redonkulous” at Ars Technica.

The Bad:

  • “Roguelite” suggests a lot more variation in repeating environments than what Housemarque has built.
  • In terms of endgame replayability, other roguelites do it better.
  • If you like a gentle ramping-up of difficulty, Returnal is not for you.
  • The story and atmosphere are pretty desolate. It’s a proper match for the Alien and Dead Space mold, but be ready for zero sense of humor.

The Ugly:

  • Too many crashes, ahead of the game’s day-one patch. We don’t know if those will be fixed.
  • A lack of a crystallized save-game option. Housemarque may need to be petitioned to change that.

Verdict: “Buy” is tough to flatly state at a $70 price point, but Returnal certainly isn’t the worst purchase a content-starved PS5 owner could make right now.

Listing image by Sony Interactive Entertainment / Housemarque

When you get down to it, “shooting” in a video game is really just a way of projecting a directed line of intent from your character to another visible point on the map. This basic fact is a large part of why shooting a gun has become such a natural means of interacting with games from a first-person perspective. If your character is looking at something, shooting a gun lets you instantly and easily engage with whatever you’re looking at.

There’s one other major real-life action where this simple point-and-shoot mechanic applies: photography. Nintendo was among the first game-makers to realize this over 20 years ago, creating Pokémon Snap for the Nintendo 64 as a new type of first-person “shooter” (it doesn’t hurt that cameras fit much better than guns with Pokémon’s family-friendly branding). In the years since, though, only a handful of games have taken Nintendo’s lead and replaced “shoot a gun” with “shoot a photo” as the main verb.

So it’s down to Nintendo to revive and expand its own good idea with the awkwardly titled New Pokémon Snap for the Switch. Though the update can get a bit repetitive and tedious at times, this secret-packed photo safari is a great mix of chill moments and competitive personal striving for the best shots.

Ride the rails

New Pokémon Snap product image

New Pokémon Snap [Switch]

(Ars Technica may earn compensation for sales from links on this post through affiliate programs.)

If we’re comparing New Pokémon Snap to gun-based shooters, it bears less resemblance to a Call of Duty and more resemblance to an arcade light-gun shooter like House of the Dead or Time Crisis. That’s because the game, like its N64 predecessor, limits your movement to a preset path, automatically pushing you forward as you tilt and zoom your camera to try to get the best shots (the descriptor “rail shooter” is popular enough to have colonized Wikipedia).

To be fair, you do unlock a few subtle ways to change things up as New Pokémon Snap progresses. You can eventually revisit each course during the night, for instance, completely changing which pokémon are present and how they might act in front of you. You can also earn your way to higher “research levels” for each course, introducing further slight variations on the available fauna. Some courses also have branching paths, letting you see new parts of the level if you trigger a preset marker.

Still, the game’s strict on-rails design can be frustrating, especially if you’re accustomed to open-world shooters that let you explore every last nook and cranny. Pokémon will sometimes remain stubbornly far away or keep their backs turned to you, creating awkward shots that would be easy to fix if you could just take a few steps to the side. Being on rails is also frustrating if you miss a carefully planned shot, forcing you to restart the entire course from the beginning and wait to get back to that same position before trying again.

The slow-paced way the game unlocks new tools and areas seems designed to encourage these kinds of repetitive runs through its largely linear environments. This feeling is exacerbated by the limited methods for filling up your in-game Photodex of Pokémon (a photographic version of the usual Pokédex).

Getting a full Photodex requires finding at least four different photos of each monster that ranks at each level from one to four stars. This ranking process can be inscrutable, but photos of rarer poses and actions by the pokémon generally earn better star rankings. The star rankings are made even more confusing because each photo also gets a bronze-to-gold color rating as well as an individual score rating based on factors like size, centering, and the presence of other pokémon in the photo.

In any case, at the end of each run you can only choose one photo per monster to submit to the professor for final evaluation and addition to your Photodex. Even if you snap dozens of photos at every star ranking in a single run, in the end you’ll have to throw most of them out and give the course another go to capture another pose for the Photodex (you can save unused pictures in a more permanent photo album, though, just for posterity and in-game online sharing).

Getting into the groove

At first, this structure seems like a deliberate obstruction meant to slow New Pokémon Snap down and pad it with fake replayability. After a while, though, I began to give in to the chill monotony of playing through each course again and again. Instead of rushing to get as many pictures as possible in each run, I could focus on the careful observation of each pokémon I spotted on each path through a stage, and in each run, I could figure out how to coax it into a new pose or photogenic positioning.

The game slowly unlocks three main ways to interact with the world besides just snapping photos. There’s fluffruit, a light apple-like snack you can throw to briefly distract monsters. There are Illumina balls, which can light up special lantern-like plants or the monsters themselves. And there’s a flute that plays a short ditty that can sometimes get monsters to dance in place briefly.

Figuring out how to use these tools to get the best shots introduces an effective puzzle element to what would otherwise be dull and repetitive trips. Throwing a fluffruit might lure a fish out of its hiding place, for instance, which then encourages a flying pokémon to swoop down and grab fluffruit out of the water, creating an amazing shot (if your shutter finger is quick enough to grab it). Another fluffruit could lure a ground Pokémon into a swirling sand tornado for an amazing midair shot. Or playing a flute at just the right time can get a Pokémon to pause just long enough for your automated car to swing around and get the perfect angle on a happy, dancing monster.

New Pokémon Snap is full of these and other kinds of secrets that take multiple trips and educated guesses to unlock. This is where the genius of the game’s on-rails design comes in. Because NPS generally knows where you are and where you’re going, it can time out semi-scripted events and behaviors for the Pokémon in each area. Going through those planned scenes multiple times encourages you to experiment and pay attention to small changes on each new run, in order to figure out how to change things just enough to lure out a reluctant Pokémon or get a great new shot. The game also embeds a subtle hint system via an in-vehicle scanner and specific photo requests made by your fellow researchers.

Come on and safari with me

How appealing all this is probably depends on how appealing you find the Pokémon themselves. Watching the cute little beasts cavort and react to your actions in lovingly detailed HD is a large part of the bill of goods being sold here (though some monsters develop very jagged edges during zoomed-in photos). If your heart doesn’t melt just a little seeing a Pichu frolicking through a field with a Grookéy, you’re probably not going to get much out of this game.

But NPS can still be enjoyable for players who can’t tell a Pichu from a Grookéy. It’s the kind of game that provides a great excuse to grab your favorite legal intoxicant, throw on a podcast, and just chill out, snapping pics while cruising through some fun, lightly interactive environments (that prospect might sound even more appealing after a year trapped inside amid a global pandemic).

If you want to get competitive and compile all the best shots, NPS has multiple ranking systems (and the ability to upload shots online for digital kudos from strangers). After a while, though, going through an entire new run just to get an incremental score improvement on a previous photo begins to reach the point of diminishing returns.

Still, after 15 hours snapping thousands of photos of hundreds of distinct Pokémon, I feel like there are plenty of secrets left to be uncovered in New Pokémon Snap‘s varied environments. What’s more, I’m eager to uncover them in quick, five or 10-minute safaris whenever I happen to have a Switch handy.

The good

  • The perfect game for chilling and taking photos
  • Plenty of secrets and hidden interactions to encourage frequent replays
  • Tons of cute Pokémon and adorable environments to cruise through
  • Fun, easy-to-use online sharing options

The bad

  • On-rails paths limit your interaction with the environment
  • Repeating an entire course for a single shot can get annoying
  • Pacing for unlocks and “new” Photodex entries can feel slow and padded

The ugly

  • Seeing distinct jagged lines around Pikachu’s ears on a zoomed-in shot

Verdict: Buy it if the idea of taking photos of cute animals has any appeal to you at all.

In the future, what might lunar exploration look like if NASA can send multiple Starships there each year? This SpaceX rendering offers a vision of one such future.
Enlarge / In the future, what might lunar exploration look like if NASA can send multiple Starships there each year? This SpaceX rendering offers a vision of one such future.

When NASA astronauts return to the Moon in a few years, they will do so inside a lander that dwarfs that of the Apollo era. SpaceX’s Starship vehicle measures 50 meters from its nose cone to landing legs. By contrast, the cramped Lunar Module that carried Neil Armstrong and Buzz Aldrin down to the Moon in 1969 stood just 7 meters tall.

This is but one of many genuinely shocking aspects of NASA’s decision a week ago to award SpaceX—and only SpaceX—a contract to develop, test, and fly two missions to the lunar surface. The second flight, which will carry astronauts to the Moon, could launch as early as 2024.

NASA awarded SpaceX $2.89 billion for these two missions. But this contract would balloon in amount should NASA select SpaceX to fly recurring lunar missions later in the 2020s. And it has value to SpaceX and NASA in myriad other ways. Perhaps most significantly, with this contract NASA has bet on a bold future of exploration. Until now, the plans NASA had contemplated for human exploration in deep space all had echoes of the Apollo program. NASA talked about “sustainable” missions and plans in terms of cost, but they were sustainable in name only.

By betting on Starship, which entails a host of development risks, NASA is taking a chance on what would be a much brighter future. One in which not a handful of astronauts go to the Moon or Mars, but dozens and then hundreds. In this sense, Starship represents a radical departure for NASA and human exploration.

“If Starship meets the goals Elon Musk has set for it, Starship getting this contract is like the US government supporting the railroads in the old west here on Earth,” said Rick Tumlinson, a proponent of human settlement of the Solar System. “It is transformational to degrees no one today can understand.”

We will nonetheless try to understand some of the ways in which Starship could prove transformational.

1. Starship ahead of schedule

Ahead of NASA’s announcement on April 16, I did not expect SpaceX to receive the only, or even the largest, award from NASA this early on in the lander-development process.

About a year ago, NASA selected three different bids for a Human Landing System. Over the course of 10 months, each of the three contractors fine-tuned its design and worked with NASA engineers to explain how its lander could meet the space agency’s needs. A team led by Blue Origin submitted the most conventional design, tailored to NASA’s request for a three-stage lander. Dynetics proposed an innovative lander, with a nod toward reusability, but it was also sized to bring just a few astronauts to the lunar surface.

SpaceX, by contrast, submitted a version of its Mars vehicle as a lunar lander. For the last five years, SpaceX has largely self-funded development of Starship as the reusable upper stage of a massive rocket, Super Heavy. The vehicle is intended to take dozens of people to Mars at a time in a six-month voyage. Thus, Starship is massively oversized to take two or four astronauts down to the surface of the Moon. But of the three landers, it is the only one with a direct path toward full reuse.

Starship is also the most technically demanding of the three vehicles because of its size and aspirations. Among the biggest hurdles are learning to land Starship, both on the Moon and back on Earth. And to conduct missions to the Moon and beyond, SpaceX must develop the technology to refuel Starship with methane and liquid-oxygen fuel in low Earth orbit.

“One of the hardest engineering problems known to man is making a reusable orbital rocket,” SpaceX founder Elon Musk told me about a year ago. “It’s stupidly difficult to have a fully reusable orbital system.”

Because there are so many technological miracles needed to validate the Starship design, I felt that NASA would not fully commit to the SpaceX vehicle as a potential lander until it had flown. Perhaps launching Starship into orbit would be enough of a technology demonstration for NASA. Or maybe SpaceX would have to land one on the Moon. This perceived need to demonstrate the viability of Starship is one reason why Musk and SpaceX have built and launched Starships at such a frenetic pace in South Texas during the last year. Only by doing, the thinking went, would NASA believe in Starship.

Instead, NASA has committed to the ambitious program even before Starship has safely landed after a high-altitude flight test. In this sense, NASA’s support for Starship has come ahead of schedule.

2. SpaceX needs NASA for Mars

After seeing SpaceX launch more than 100 rockets over the last decade, what has become abundantly clear is that its engineers are now the best in the world at designing, building, and flying new and innovative rockets. The execution of the Falcon 9 program, proving out first-stage reuse, and development of the Falcon Heavy rocket attest to this.

But building great rockets is one thing. It is another thing to develop all of the other capabilities needed to ensure that humans can travel to Mars, land on the red planet, and survive there.

When it comes to in-space activities, SpaceX has leaned on NASA’s expertise for Crew Dragon as part of the commercial crew program. And with respect to the kinds of technologies needed for long-duration travel to Mars, through deep space, SpaceX has limited experience—there is very little recycling of air, water, and other consumables on a Crew Dragon spacecraft. NASA, on the other hand, has been working on these problems for more than a decade with astronauts on the International Space Station.

The space agency has also been conducting studies of Moon and Mars missions for decades, said Abhi Tripathi, who worked as a systems engineer at NASA from 2000 to 2010 performing these kinds of analyses. Tripathi left NASA to work at SpaceX on the cargo and crew versions of the Dragon spacecraft until 2020, when he moved to the University of California, Berkeley.

“NASA will undoubtedly bring to bear a wealth of invaluable information, technology, and subject matter experts to help SpaceX achieve their shared goal of putting humans on Mars,” Tripathi told Ars.

NASA and SpaceX collaborating this early on Starship also helps with a host of other issues not related to transportation. A government agency will be needed to facilitate the development of nuclear-based power for the surface of Mars, for example. And any human missions to Mars will raise planetary protection questions and other international concerns. Having NASA alongside SpaceX means the US government will help address all of these issues.

Suddenly, human landings on Mars about a decade from now seems a lot more realistic.

3. NASA bets on game-changing technology

The world has never seen a vehicle like Starship before. If successful, the massive spacecraft would open up new possibilities to NASA not before available. This is because Starship could realize the long-desired goal of rapid, low-cost reuse of a launch system.

Consider the status quo. The large Space Launch System rocket under development by NASA will be able to launch 95 metric tons into low Earth orbit. NASA and its contractors, led by Boeing, will be able to build one a year. The expendable vehicle will launch one payload, at a cost about $2 billion per mission, and then drop into the ocean.

In terms of lift capacity, the vehicles are similar. Starship and Super Heavy should be able to put about 100 tons into low Earth orbit. However, SpaceX is already capable of building one Starship a month, and the plan is to reuse each booster and spacecraft dozens of times. Imagine the kind of space program NASA could have with the capacity to launch 100 tons into orbit every two weeks—instead of a single annual mission—for $2 billion a year. Seriously, pause a moment and really think about that.

In their decision to select SpaceX, NASA officials appeared to recognize this potential. “We were looking to see what industry partners could bring in terms of innovation and solutions,” said Lisa Watson-Morgan, the Human Landing System program manager. The emphasis here is on innovation and new solutions to old problems.

“In picking the Starship architecture, NASA is helping enable a path toward a super heavy launch vehicle, in-space propellant storage, in-space refueling, and large up and down mass to planetary surfaces,” said Tripathi, who has examined these problems from both NASA and SpaceX’s perspective.

Put another way: if Starship is successful, NASA no longer needs to pick just one or two big things to do in space. The agency will be able to do many different things at the same time.

4. NASA funds an SLS competitor

So why is NASA funding a launch system that will directly compete with its SLS booster? That, to be clear, was not the space agency’s intent. In explaining the award during a news conference, agency officials were careful to say that SLS and the Orion spacecraft remain an essential part of the Artemis architecture. But in reality, NASA may well be putting its SLS rocket out of business.

With the Human Landing System award, NASA has put its stamp of approval on Starship and Super Heavy. The launch system will eventually go into the catalog maintained by NASA’s Launch Services Provider program, allowing other agency programs to procure the vehicle for missions. This could be a real boon for large, space-based telescopes that would find the large volume of Starship’s payload fairing useful.

“If I were an official in one of NASA’s other directorates, I would personally be dreaming up all kinds of ideas for what I can someday do with all these substantive new capabilities,” Tripathi said.

In the big picture, $2.89 billion is not a lot compared to what NASA has already invested in the SLS rocket. The space agency spends that much every year in development costs for the rocket and its associated ground systems. Because the SLS rocket is funded through cost-plus contracts to major space contractors like Boeing, there is less incentive to control costs or deliver a timely product. Predictably, the SLS vehicle is significantly over budget and now five years behind its original launch date of late 2016.

All of this has led to criticisms that SLS is a jobs program. Indeed, it provides jobs in all 50 states and supports hundreds of small businesses. And perhaps this explains why Congress has steadfastly supported SLS despite its costs and delays.

By contrast, Starship is not a jobs program. Rather, it’s a jobs-killer program from the perspective of Congress.

5. Is SpaceX too dominant?

SpaceX has enjoyed a remarkable string of NASA contract wins. Over the last decade, it has landed NASA awards to deliver cargo and crew to the International Space Station, launch the Lunar Gateway, supply this Gateway with cargo, and now deliver humans to the surface of the Moon.

The Artemis Program could also plausibly morph into the SpaceX Lunar Program. How? Under the current plan, a Super Heavy rocket would launch Starship to lunar orbit. Days later, an SLS rocket would launch crew inside an Orion spacecraft, which would dock with Starship in lunar orbit. The crew would transfer to Starship and go down to the Moon. After coming back to lunar orbit on Starship, the astronauts would board Orion and fly back to Earth.

But if Starship is safe for humans to land on the Moon, why would it not be safe for humans simply to launch from Earth on board the vehicle? This would save NASA the cost of an SLS plus Orion launch—about $3 billion per mission, combined—and a tricky rendezvous and docking in lunar orbit. This is probably the future of a truly sustainable lunar exploration program.

That’s good for NASA and for SpaceX, but what about the other spaceflight companies? Under the (much) more expensive plan using SLS and Orion, NASA is also funding a who’s who of aerospace companies: Boeing, Lockheed Martin, Northrop Grumman, Aerojet Rocketdyne, United Launch Alliance, and many, many other smaller players across the United States. Starship directly supports SpaceX, its limited number of suppliers, and… whatever company ends up building spacesuits for lunar forays.

It is therefore difficult to see a SpaceX-only exploration program winning broad congressional support for Artemis. History suggests that all of the losing contractors would urge the politicians they bolster with contributions to actively oppose the program.

And what of international partners and the geopolitical implications of this? During a confirmation hearing this week before the US Senate Commerce committee, incoming NASA Administrator Bill Nelson said expanding the coalition of nations participating in the Artemis Program was one of his big goals. Increasing the space agency’s reliance on SpaceX likely would work against this.

With the low-cost, reusable Falcon 9 rocket, SpaceX has already badly damaged the commercial launch industries in Europe, Russia, and Japan. For the Artemis Program, Europe is contributing the Service Module for the Orion spacecraft. How would these officials react if NASA now says, “non merci” to that contribution because of SpaceX?

“The nation’s activities in deep space remain very tied up in international policy, alliances, adversaries, and security, as well as space exploration and science,” an industry source told Ars. “There are a number of foreign policy interdependencies and offsets that are managed through or impact space, generally below the surface. What does this choice signal to all of those players?”

In summary

With Starship, SpaceX has offered what appears to be the best technical solution to NASA’s stated goal of a sustainable lunar exploration program. Starship would be able to take far more people and cargo to the Moon than any other solution for NASA—and it could do the job for far less money and far more often.

Furthermore, in awarding the Human Landing System contract to SpaceX, NASA has embraced a risky yet highly rewarding technology.

But whereas NASA is a space agency, its feet remain very much grounded in the political orbit of Washington, DC’s beltway. Technically, Starship may be the best solution to NASA’s needs. But politically, would it be? Probably not. If NASA wants to go to the Moon and beyond, it must work with a multitude of contractors and countries, at least for now.

Ultimately, physics will win out. If SpaceX can make Starship work, eventually NASA’s other options for human exploration of the Solar System may come to look ridiculous by comparison. By placing an early bet on Starship last week, NASA has increased the ultimate odds of Starship’s success.

For the space agency, this is an audacious and surprising play. But the potential payoff is huge. One day it may allow us to boldly go not just back to the Moon, but far, far beyond.

This tiny PC's Ryzen 5 Pro 2500U is a couple generations out of date—but it's inexpensive, and it still packs a serious punch.
Enlarge / This tiny PC’s Ryzen 5 Pro 2500U is a couple generations out of date—but it’s inexpensive, and it still packs a serious punch.
Jim Salter

Minisforum’s UM250 is a very small form factor PC with the power and the ports to take on a lot of tasks. And due to its choice of an older CPU, it’s pretty cheap, too.

A couple of months ago, we reviewed Minisforum’s Comet Lake i5-powered U850. The UM250 we’re looking at today is cut largely from the same cloth—it’s got 16GiB RAM, flagship Intel Wi-Fi 6, a 256GB SSD, two wired Ethernet ports, and an attractive VESA-mountable case that’s easy to work on (and in).

The biggest real-world difference between the two models is price: $430 for the fully loaded, AMD-powered UM250 versus $700 for the Intel-powered U850.


(Ars Technica may earn compensation for sales from links on this post through affiliate programs.)

Like most of Minisforum’s models, the UM250 is an unassuming little silver-and-black brick stuffed with ports—including four USB type-A ports and enough video out to drive three displays via USB-C, DisplayPort, and full-size HDMI.

The UM250 we tested is “fully loaded” with 16GiB of socketed dual-channel RAM, a 256GB Kingston M.2 SSD, and a copy of Windows 10 Professional. If you’re looking to supply your own RAM, SSD, and OS there’s also a bare-bones version on Minisforum’s store at $320.

The reason the UM250 is so relatively inexpensive (not much more than half the cost of the Intel-powered U850) is the Ryzen 5 Pro 2500U powering it. The UM250’s 2500U is almost two years older than the Comet Lake i5 in the U850, but it goes neck-and-neck with the newer, more expensive Intel part in most benchmarks. Heck, the Ryzen even wins in some areas.

Minisforum also shaved off some cost by only providing a single SATA port versus the U850’s two, and by using a slower M.2 SATA model of the Kingston SSD. The UM250 also offers dual RTL8111 Gigabit Ethernet versus the U850’s RTL8111 Gigabit + Intel 2.5Gbps Ethernet. We suspect most of the folks in the market for this sort of mini-PC won’t mind those sacrifices, especially when considering they come at nearly $300 off the retail cost.

Moving past raw specs, the UM250 is pleasant to share an office with. Even in Time Spy and Cinebench R20 multi-threaded testing, its cooling fan stays reasonably quiet. If you’re close to it in a dead silent environment, you’ll be able to hear it—but even then, it’s a steady clean whoosh without any bearing whine. This mini-PC is slow to change RPMs rather than rapidly spinning up and down repeatedly.

Inside the UM250

Specs at a glance: UM250
CPU Ryzen 5 Pro 2500U
OS Windows 10 Pro (pre-installed) / Linux supported
GPU Vega 6 (integrated)
Wi-Fi M.2 Intel AX200 Wi-Fi 6, dual-band + BlueTooth 5.1
SSD Kingston M.2 256GB SATA SSD
  • one SATA port
  • one full-size HDMI 2.0
  • one full-size DisplayPort
  • one USB-C (full featured)
  • DC barrel jack
  • four USB3.1 Type-A
  • two 1Gbps Ethernet (Realtek 8111H)
  • one 3.5 mm audio
  • integrated mic
Price as tested $430 at Amazon / $470 at Minisforum

Much like the U850, the UM250 is extremely easy to get into and work on/in. The top plate can be removed by gently pushing two corners and letting it pop out (similar to some kitchen cabinet doors). Once inside the UM250, you’re presented with a socketed NVMe SSD on the left, an unpopulated SATA power+data connector in the center, and two socketed DDR DIMMs on the right.

Unlike the more expensive U850, the UM250 only offers a single SATA connector—and no sunken drive bays in the chassis itself. Instead, you can bolt a 2.5″ SATA HDD or SSD to the underside of the top plate. This is functional but a little irritating, since it means your SATA cable is attached to the plate you must remove to get into the box.

But again, considering the massive price disparity between the U850 and UM250, we’re not complaining. We’re just happy there’s a SATA connector and mounting bracket at all, given that the primary drive is NVMe.

Multi-threaded CPU Performance

Those of you who made sour faces at the Comet Lake powered U850 and said “I’ll wait for the AMD version” should feel pretty vindicated right now. Despite UM250’s Ryzen 2500U being almost two years older, it runs neck-and-neck with U850’s much newer (and more expensive) Comet Lake i5 CPU.

The two processors turn in essentially identical Cinebench R20 runs. Meanwhile, Passmark slightly favors the Ryzen 2500U, and the Intel-loving Geekbench 5 gives both the i5-10210U and i5-1035G1 moderate wins. If the only criterion you have for a mini-PC is performance, the U850 is the slightly better bet… but it costs nearly twice as much and has a somewhat more obnoxious fan noise to boot.

Single-threaded CPU performance

Unsurprisingly, the Ryzen 5 Pro 2500U doesn’t do as well in single-threaded testing, losing by margins of 20%-30% to newer, more expensive processors. Just as unsurprisingly, we warn readers not to get unduly excited about this—truly single-threaded workloads are rare.

Even workloads which bottleneck on a single CPU thread generally do so with significant activity on at least one other CPU thread. And “the one thread I care about the most” inside a multi-threaded workload is generally not the same thing as the truly single-threaded results you see above.

Storage performance

Storage performance is the most painful compromise the UM250 makes, with a SATA-only M.2 drive.
Enlarge / Storage performance is the most painful compromise the UM250 makes, with a SATA-only M.2 drive.
Jim Salter

The biggest real compromise the UM250 makes to keep the cost down is dropping down to a SATA-only M.2 drive. The 256GB Kingston SNS-8180-S3 in this little PC isn’t slow enough to make a Windows desktop feel pokey, but its shortcomings are pretty obvious in comparison to the NVMe drives in the more expensive U850 and in the Acer Swift 3 laptop.

Gaming performance

Nobody ought to think of any of the machines in these charts as “gaming PCs” by any stretch. But within its own bracket, the UM250 turns in some outstanding results—its somewhat-elderly Ryzen 5 Pro 2500U turns in Time Spy scores almost double what either Ice Lake i5-1035G1 or Comet Lake 15-10210U can manage. The UM250 even comes within yelling distance of the Ryzen 7 4700U-powered Acer Swift laptop.

When we drop down to Night Raid—the gaming benchmark targeted to non-gaming PCs—the lead shrinks a little, but it by no means disappears. UM250 still manages a solid 35% lead on its Intel-powered competitors. This is a bigger lead than the newer processors had in single-threaded CPU tests.

Linux support

Linux support on the UM250 is flawless. We installed Ubuntu 20.04 to test, and the only unclaimed devices are an IOMMU and a crypto accelerator.

Both Ethernet interfaces, the Intel Wi-Fi 6, and all onboard sound and video worked flawlessly. The UM250 is simply an excellent choice for small form factor Linux computing, for either desktop or medium-power “server” workloads.


UM250 is another solid effort from Minisforum, providing an extremely cost-effective mini-PC suitable for either desktop or project use. With dual gigabit Ethernet plus Intel AX200 Wi-Fi 6, this little PC would make an excellent router, too.

In practice, this is a very easy machine to share an office with—although actively cooled, its fan is quiet and unobtrusive, with no bearing whine or rattle in what noise it does make. The UM250 also tends to stay at the same RPM for long periods, rather than spinning rapidly up and down as CPU load changes.

The biggest performance compromise the UM250 makes is in its storage, which is a Kingston M.2 SATA SSD. The Kingston isn’t horrible, by any means—the Windows desktop felt snappy, and the machine boots pretty quickly. But because of this choice, the UM250 won’t really hang with high-end SATA SSDs, let alone M.2 NVMe drives like you’ll find in the more expensive U850.

The good

  • Intel Wi-Fi 6
  • Windows 10 Pro on a <$500 PC
  • Toolless chassis
  • Socketed everything
  • Dual Ethernet
  • Excellent Linux support
  • Great performance-per-dollar
  • Quiet cooling
  • VESA mountable
  • Supports up to 3 4K displays (USB-C + DisplayPort + HDMI)

The bad

  • M.2 SATA SSD
  • No free puppy

The ugly

  • Ryzen 5 Pro 2500U is three years old
black German car drives through a modern tunnel

The self-driving technology industry is in a strange state right now. A number of companies have been pouring millions of dollars into self-driving technology for years, and many of them have prototype self-driving vehicles that seem to work.

Yet I know of only one company—Waymo—that has launched a fully driverless commercial taxi service. And I only know of one company—Nuro—that’s running a driverless commercial delivery service on public roads. You’d expect these companies to be capitalizing on their early leads by expanding rapidly, but neither seems to be doing that.

Meanwhile, several other players, including Cruise and Mobileye, say they’re planning to launch large-scale commercial services by 2023. But plenty of self-driving companies have blown past self-imposed launch deadlines in the past, so it’s not clear if that will actually happen.

In short, predicting what the next couple of years will bring is a challenge. So rather than offering a single prediction, here are eight: I’ve broken down the future into eight possible scenarios, each with a rough probability. Hopefully, breaking things down this way offers a good overview of the many different strategies being pursued by self-driving companies today. A decade from now, we’ll be able to look back and say which companies or approaches were on the right track. For now, we can only guess.

1. Waymo wins (20 percent)


Waymo has been viewed as the self-driving industry’s technology leader ever since it started as the Google self-driving car program more than a decade ago.

In the optimistic scenario, Waymo will maintain and expand its current lead. It will grow its current taxi service from one corner of the Phoenix metro area to all of Greater Phoenix, then steadily expand to other metro areas. Running the largest driverless taxi service could give Waymo access to more real-world driving data and operational experience than any other company has, which could allow it to further improve its software and maintain its lead.

So why do I only give Waymo a 20 percent chance? While Waymo still seems to be the technology leader, it hasn’t capitalized on its lead as well as many people—apparently including Waymo’s own leadership—expected a few years ago.

In 2018, Waymo announced deals to purchase “up to” 82,000 vehicles for use in its taxi fleet, suggesting the company thought it was on the verge of large-scale commercial launch. Yet today its fleet still numbers in the hundreds of vehicles.

I don’t know why Waymo is moving slowly. Maybe its software has become excessively optimized for suburban Phoenix. Maybe its hardware or back-end support costs are too high to operate profitably. Maybe there are lingering safety or reliability concerns that Waymo wants to squash before expanding in a big way.

But whatever the issue, it may not go away any time soon. Which could leave an opening for other companies.

2. Another robotaxi company wins (25 percent)


Plenty of other companies are pursuing the same basic strategy as Waymo—building and operating a robotaxi fleet. These include:

  • Cruise (owned by GM, Honda, and others)
  • Argo (owned by Ford and Volkswagen)
  • Motional (owned by Hyundai and auto-parts supplier Aptiv).
  • Zoox (a startup that was recently acquired by Amazon)
  • Aurora (a startup that recently acquired Uber’s self-driving program)

If Waymo falters, I think it’s most likely to be on business execution: Waymo continues to have industry-leading technology but fails to expand rapidly enough to take full advantage of it. Running a taxi service with a few hundred vehicles in one metro area (as Waymo is doing now) is a very different proposition from running a taxi service with hundreds of thousands of vehicles in dozens of cities.

Automaker-backed companies like Cruise, Argo, and Motional might have a greater ability to rapidly scale up production of self-driving vehicles. Amazon obviously has a lot of experience with large-scale logistical problems. And Aurora has a close relationship with Uber, which might provide Aurora with preferential access to its ride-hailing network.

3. Tesla (and wins (5 percent)

This will make Tesla fans mad, but I think it’s true: Tesla is a long shot.

The bullish case for Tesla is that it has access to a vast trove of real-world driving data harvested from customers’ vehicles. If you think limited training data is a major bottleneck for improving self-driving algorithms, then this might be a significant advantage. Tesla CEO Elon Musk also has a bigger appetite for risk than most of the other companies working on self-driving technology. Musk’s willingness to put unproven technology on public roads may accelerate Tesla’s progress even as it creates a greater risk of fatal accidents.

On the other hand, Tesla has significant disadvantages. The company’s business model—selling cars to end users—puts lidar sensors and high-density maps financially out of reach. Elon Musk has tried to spin this as a positive, calling lidar a “crutch.” But the fact remains that almost every other company is using lidar and HD maps because it believes they are helpful.

More fundamentally, it’s hard to watch videos of Tesla’s software in action and conclude that Tesla is in a leading position—or even that it is catching up to the leaders. Tesla’s unfortunately named “full self-driving beta” software routinely flubs scenarios that Waymo’s cars have been able to handle for years.

One of those Tesla self-driving videos referenced above

If I’m wrong and Tesla’s strategy does succeed, that would be very good news for, a self-driving startup founded by legendary hacker George Hotz. Comma is building an open source self-driving system designed to run on a smartphone. Comma’s strategy is to enable early adopters to modify their own cars to take steering inputs from Comma’s smartphone-based software—and then use the data harvested from those early customers to further improve the software in much the same way as Tesla. Like Tesla, Comma has eschewed lidar, arguing that it can achieve adequate performance with smartphone-grade cameras.

Hotz’s ultimate goal is for Comma to be the Android to Tesla’s Apple. That is, if Tesla emerges as a clear leader in self-driving technology, other automakers will need to license their own self-driving technology to compete with Tesla. Hotz hopes that Comma’s software will become an industry standard among automakers, much as Android is an industry standard for smartphones not made by Apple.

4. Mobileye (and its partners) wins (10 percent)

A self-driving vehicle from Mobileye's autonomous test fleet navigates the streets of Detroit.
Enlarge / A self-driving vehicle from Mobileye’s autonomous test fleet navigates the streets of Detroit.

Mobileye, an Israeli company acquired by Intel in 2017, has some big advantages. As the leading provider of advanced driver-assistance systems (ADAS), Mobileye has relationships with a bunch of automakers. The company has leveraged those relationships to get data from numerous customer cars, giving Mobileye access to a king’s ransom in real-world road data. Mobileye’s dominant position in the ADAS market also gives it a deep bench of engineering talent.

Like Tesla (and unlike Waymo and some other companies in the space), Mobileye believes that full autonomy can be achieved through incremental improvements to ADAS. Mobileye is hoping to gradually deliver better and better ADAS until, at some point, customers can take their hands off the wheel and take a nap in the driver’s seat.

At the same time, Mobileye isn’t as hemmed in by financial constraints as Tesla. With Intel as its parent company, Mobileye has the resources to explore multiple technological approaches simultaneously. Like Waymo, Mobileye is experimenting with lidar and testing fully self-driving technology on public streets with safety drivers behind the wheel.

Mobileye’s plan is to license its chips, sensors, and software to a range of customers. So if Mobileye’s technology works, it could allow smaller companies to ride Mobileye’s coattails. The delivery startup Udelv, for example, recently announced plans to use Mobileye’s technology in at least 35,000 delivery robots between 2023 and 2028. Mobileye is hoping to have dozens of customers like Udelv in a few years.

An industry favorite?

A Mobileye victory would also be pretty good news for other automakers, because it would mean they’re unlikely to get disrupted by Waymo, Tesla, or other upstarts. Mobileye’s main business is licensing technology to automakers, so if Mobileye develops a leading self-driving stack, carmakers should be able to easily integrate it into their existing product lines.

Mobileye has released some impressive videos of its technology in action. However, it’s hard to judge self-driving technology based on a few videos. Lots of companies have released impressive videos, but most of them haven’t been confident enough in their technology to launch driverless commercial services.

There’s also a risk that Mobileye’s business model—licensing its technology to a wide variety of customers, from automakers to companies running delivery fleets—could hamper the company’s progress. Developing self-driving technology for one specific application is hard enough. Mobileye is hoping to develop a single technology stack that will perform well in a wide range of applications. That might not be realistic.

5. China wins (15 percent)

Autonomous vehicles from in Beijing in April 2021.
Enlarge / Autonomous vehicles from in Beijing in April 2021.

There’s a lot of international collaboration on self-driving projects, with overseas companies like Honda and Volkswagen investing in American self-driving projects and Intel buying the Israeli Mobileye. But China is mostly in a world of its own.

China has its own stable of self-driving companies that have largely developed independently from Western nations. Leading self-driving companies in China include AutoX,, WeRide, search giant Baidu, and ride-hailing firm Didi Chuxing.

Last December AutoX became the first company to test fully driverless vehicles on public roads in China—though this was merely a test of its technology, not the launch of a commercial driverless service.

I haven’t done very much reporting on these companies, so I can’t talk at length about their prospects. But what seems clear is that, if American self-driving companies face serious overseas competition, it’s most likely to come from China. (The most formidable company without American or Chinese ties is probably Russia’s Yandex.) China’s authoritarian but relatively competent government could theoretically find ways to speed the development of Chinese self-driving technology, whether that’s by building supportive infrastructure or shielding companies from liability concerns.

But so far, Chinese companies don’t seem to be dramatically ahead of their American counterparts, so I don’t rate this an extremely likely possibility.

6. Self-driving trucks win (5 percent)

Kodiak Robotics

Self-driving cars get the most attention, but a substantial minority of self-driving companies believe that trucks, not cars, will be the first application for self-driving technology.

Companies like Kodiak, Embark, and TuSimple focus on automating long-haul trucking routes. A couple of other companies—Waymo and Aurora—are hedging their bets, simultaneously working on trucking and taxi projects. Nuro, a delivery startup I’ll discuss more below, is also working on long-haul trucking via its <a href=””>acquisition</a> of Ike last year.

Some of these companies are designing their systems to work autonomously only on freeways. Under this model, a human driver will drive the truck to a transfer point near the freeway. The trailer will then be transferred to an autonomous truck that will haul it on the freeway to another city. Then the trailer will be switched to another human-driven truck to navigate the last few miles of tricky urban driving to get the truck to its final destination.

This model has a number of advantages. Freeways are controlled environments that are generally free of pedestrians, cyclists, and other obstacles. They are generally well-marked and lack complex intersections. These factors could make the task of automating freeway driving relatively simple.

And in theory, a company could parlay the data and experience garnered from early success with self-driving trucks to other markets, growing to eventually be a major player in a range of self-driving applications.

However, I remain a skeptic of self-driving trucks for a simple reason: while freeways are easier to navigate in some ways, the potential costs of failure are extremely high. If a fully loaded semi truck loses control at 70 miles per hour, it could cause real damage. And there’s no obvious way to test this technology incrementally. At some point, each of these companies is going to have to push a button and send a truck driving on its own at freeway speeds. I’m skeptical anyone is going to feel confident enough to take that risk that any time soon—or that regulators in many jurisdictions will feel comfortable enough to allow it.

7. Delivery robots win (10 percent)

Nuro recently announced a pilot project to deliver Domino's pizzas in Houston.
Enlarge / Nuro recently announced a pilot project to deliver Domino’s pizzas in Houston.

Nuro is another company that’s planning to haul stuff instead of people. But instead of automating large, high-speed trucks on the highway, Nuro is building small, low-speed delivery robots optimized for streets in residential neighborhoods (Nuro’s vehicles are bigger and faster than sidewalk robots). This is an attractive initial application for self-driving vehicles because a smaller robot traveling at a top speed of 25 miles per hour (40 km/h) is less likely to kill someone.

Two years ago, I made the argument that startups like Nuro that focused on low-speed applications could disrupt Waymo. The other startup I pointed to in that piece was Voyage, which was building a low-speed taxi service at a retirement community in Florida. I argued that focusing on low-speed operations in a controlled environment would greatly simplify the technology challenge, allowing these companies to enter the market much sooner than companies trying to build full-speed taxi services serving entire metro areas.

I then argued that once these companies had mastered these simpler applications of self-driving technology, they might be able to use the data and operational experience they glean to move upmarket and eventually beat companies (like Waymo) focused on more challenging driving tasks.

This prediction hasn’t held up well. It’s true that Waymo has struggled to bring its own technology to market at scale. But it’s not clear that startups like Nuro and Voyage are doing better. Voyage was acquired by Cruise earlier this year.

Nuro is still an independent company and appears to be going strong. However, Nuro seems to be in a similar holding pattern as Waymo. Earlier this week, for example, Nuro announced a pilot project delivering pizza for Domino’s—a partnership Nuro first announced way back in 2019. Like Waymo, Nuro has seemed to be on the verge of large-scale commercial operations for two or three years.

This doesn’t necessarily mean Nuro is doing anything wrong—it might just inherently take years of practice to get an autonomous delivery vehicle running. But this pace does suggest that focusing on low-speed delivery robots isn’t much of a shortcut to commercialization.

Still, the fact remains that Nuro is one of the few companies actually running a commercial self-driving service. And it may be gaining experience that will allow it to expand rapidly in the next few years. If Nuro does succeed, that could be good news for Zoox. While Zoox is primarily focused on building a taxi service, it could easily pivot to building driverless delivery vehicles for its parent company, Amazon.  Also, if Nuro succeeds we can expect a lot of other companies building self-driving stacks to court Udelv as a potential partner or acquisition target.

8. Somebody else wins (10 percent)

Technology markets have a way of surprising us. It’s possible that some startup I haven’t noticed yet is destined to embarrass the well-funded companies mentioned above.

Another possibility is an “everyone wins” scenario. Maybe self-driving isn’t one problem but a bunch of individual problems, each requiring its own carefully designed technology stack. Maybe driverless taxis, driverless vehicles, and driverless trucks are three distinctive markets that will each be dominated by its own group of companies. And maybe early versions of the technology will require so much human supervision that we’ll have a slow and gradual transition from human-driven vehicles to driverless ones. I don’t think this is a likely outcome, but it’s possible.

Another possibility is that nobody wins: maybe self-driving is an even harder problem than people appreciate, even after years of setbacks, and it will take decades, rather than years, to get working. In that case, we could have yet another “AI winter” where a lot of companies scale back their research in this area. Again, I don’t think this is likely, but it’s a possibility.

Recently, we managed to get our hands on an Asus ROG Zephyrus G15 laptop. We had to do it just like anybody else does, by finding one and buying it retail. That’s notable because this laptop combines AMD’s Ryzen 9 5900HS processor with an Nvidia RTX 3070 Mobile GPU, a combination that means this device sells out extremely quickly.

Make no mistake, the RTX 3070 in this year’s AMD-powered Zephyrus is a distinct step up. Last year, if you wanted an RTX 3000 series GPU in a laptop, you had to settle for an Intel CPU to go with it.


Specs at a glance: Asus ROG Zephyrus G15 GA503, as tested
OS Windows 10 Home
CPU 3.0GHz 8-core AMD Ryzen 9 5900HS (4.5GHz boost)
RAM 16GiB DDR4-3200
GPU AMD Radeon 8 core / Nvidia GeForce RTX 3070 MaxQ
SSD SK Hynix M.2 NVMe PCIe3.0 1TB
Battery ASUStek 90Wh
Display 1440p WQHD, non-glare, 165Hz, adaptive sync
  • two USB-A ports
  • two USB-C ports
  • 3.5mm phone/mic combo jack
  • DC power jack
  • full-size HDMI out
  • RJ-45 wired Ethernet
  • micro SD reader
  • Kensington lock slot
  • no camera
Price as tested $1,800 at Best Buy

This year’s Ryzen-powered Asus ROG Zephyrus G15 is almost everything we’d want in a gaming laptop. Beastly CPU? Check. Beastly GPU? Check. Full-size wired LAN port? Check. Loud speakers? Absolutely. 1440p display with high refresh? Yes. Even the storage on this laptop—a model of SK Hynix NVMe SSD we’d never seen before—is blistering fast.

Unfortunately, this year’s Zephyrus G15 shares the same Achilles’ heel as last year’s Ryzen-powered G14: there’s no webcam.

This puzzling omission feels almost like somebody at Asus lost a bet. Not only do game streamers frequently want a cam view, we’re a year and change into a global pandemic with near-universal need for daily teleconferencing. So, why Asus… why?

Aside from that, this is a great general-purpose laptop as well as a gaming-specific beast. Even its fans were much better behaved than last year’s G14, and this setup remained near-inaudible throughout everything but Time Spy testing.

Internals and upgrades

Getting into the Zephyrus G15 is relatively easy. There are no individual compartment panels; the entire back plate lifts off after removing 10 visible screws and three hidden ones. (The screws in the center of the panel are the “hidden” ones, concealed under sticky rubber caps for some reason.)

Note that the lower right screw is a captive design—although you can force it to, it’s not supposed to actually come out of the back plate entirely. The idea is that after removing the first 12 screws entirely, you can use the unscrewed-but-still-inserted thirteenth as leverage to pop the panel loose in that corner. We’d still recommend a spudger and some patience if you want to remove the panel without damage, whether or not you use the captive screw to begin.

Once you’re into the G15, everything’s easy to find and work with—there’s none of this silly “remove the keyboard to access the Wi-Fi” business. You’ll find the single DIMM slot (initially populated with an 8GiB Samsung DIMM, on our system) just left of center. The C: drive, an SK Hynix 1TB M.2 NVMe drive, is just above and to the left of the DIMM slot. A second, unpopulated M.2 NVMe slot is off to the right of the first M.2, and the Intel AX201 Wi-Fi is socketed just below that.

The only somewhat sour note here is that single DIMM slot—8GiB of the laptop’s base 16GiB is soldered to the board. This limits how far you can upgrade the RAM. The manual says it only goes to 24GiB, but we got it to 40GiB without problems using a 32GiB HyperX DIMM. Still, you’re effectively stuck in single-channel mode for 24GiB of that, since it’s not possible to upgrade both banks evenly.

Performance—multi-threaded CPU

The Ryzen 9 5900HS in the Zephyrus G15—to absolutely no one’s surprise—wipes the floor with every competitor in multi-threaded testing, including the unusually Intel-friendly Geekbench 5.

In Cinebench R20—generally our favorite one-size-fits-all CPU test—the 5900HS only improves on last year’s 4900HS (found in the 2020 ROG Zephyrus G14) slightly, but they both deliver roughly double the performance of their closest competitors.

Passmark shows an enormous boost to the 5900HS which didn’t show up in Cinebench R20. Finally, the 5900HS gets a modest but noticeable victory over Tiger Like i7-1185G7, our previous Geekbench 5 champion in the laptop space.

Performance—single threaded CPU

Single threaded performance testing is more of a mixed bag. The fight here is entirely between the Ryzen 9 5900HS in the ROG G15 and the Tiger Lake i7-1185G7 running flat-out at 28W TDP.

In Cinebench R20, the Tiger Lake i7 and the Ryzen 9 5900HS are neck and neck, with a very narrow victory going to Tiger Lake. Passmark shows us the same situation, but with Ryzen 9 the just-barely-victor. Geekbench 5 grants the Tiger Lake i7 a more noticeable victory.

As usual, we caution readers not to get too excited about single-threaded numbers—the margins here are much narrower than in multi-threaded testing, and the tests themselves model a pretty uncommon workflow.

It’s much more common to bottleneck on a single thread in a multi-threaded workload than to have a truly single-threaded workload running standalone. Also, these numbers do not demonstrate “per thread performance in a multi-threaded workload;” they test true single threaded performance with all other cores idle or very near-idle.


The SK Hynix SSD in the G15 flexes serious storage muscle in CrystalDiskMark testing, nearly doubling competing drives' performance in large writes.
Enlarge / The SK Hynix SSD in the G15 flexes serious storage muscle in CrystalDiskMark testing, nearly doubling competing drives’ performance in large writes.
Jim Salter

We weren’t sure what to expect from the SK Hynix NVMe SSD in the G15—that’s not a name you see often in the consumer SSD space, and the specific model seems to be new to the Zephyrus G15 as well.

CrystalDiskMark testing definitely put our minds at ease; the SK Hynix drive outperformed not only the Kingston Design-in drive from the Minisforum U850, but also the WD Black 2TB and WD SN730 512GB (not pictured) from this spring’s gaming-focused Ars System Guide.

This is a seriously fast SSD, which contributes strongly to the overall impression the ROG G15 laptop gives a new user. It’s not just a narrowly-focused FPS monster, it feels extremely fast in general use cases as well.


When we first tested the ROG G15, we did so with completely out-of-the-box drivers—and it didn’t do nearly as well as you see in the charts above. In fact, it barely outpaced last year’s G14, which only has an RTX 2060 Mobile in it.

After downloading and installing GeForce Experience drivers directly from Nvidia, we pulled much more commanding results from the 3DMark gaming benchmark suite, which you see above. With the correct, up-to-date drivers installed, this system actually hangs neck-and-neck with the Lenovo Legion 5i gaming PC from this spring’s System Guide—that’s one heck of an accomplishment for a laptop.

On the other hand, the G15—despite having a GPU that says “RTX 3070” on it—does not come within shouting distance of the HP Omen 30L, which has a desktop RTX 3080 on board. If you’re intimately familiar with wide ranges of CPU and GPU parts, this doesn’t come as a surprise. But there are always new folks who don’t realize that “desktop 3070” and “laptop 3070” perform wildly differently, as do, for example, “desktop i7” and “laptop i7.”

That said, an RTX 2070 Super equipped desktop gaming rig is no slouch—nor is it cheap. At the $1,800 we paid for the ROG G15, we’re getting almost the exact gaming performance we would from a similarly-priced desktop rig. That’s pretty exciting.

Battery life

Nobody buys a gaming laptop thinking they’ll get the best battery performance around—and that mindset should be adopted for the G15. Even though it doesn’t hang at the very top of our battery-life charts, this laptop scores solidly in the middle, alongside a pack of very good other laptops.

PCMark 10’s Modern Office battery test simulates “a typical day in the office”—there are documents to be edited, videoconferences to be had and presentations to be made, with significant periods of desktop idle in between. The G15 scores just under 11 hours on this test—less than the HP Dragonfly Elite or Dell XPS 13, but better than Acer’s Swift 3, let alone last year’s ROG G14. This is mostly thanks to the whopping 90Wh battery in the G15.

Gaming battery life is a different story. The PCMark 10 Gaming battery test runs a moderately demanding 3D demo loop, swooping around the drone-haunted, lava-filled canyon from Fire Strike. The G15 only manages an hour and 18 minutes here—and the last eighteen minutes limped along at only 14 frames per second.

Gaming on battery is a joke. We don’t recommend that anyone plan on doing much of that—on this or any other laptop.

Linux? Linux!

To our surprise and joy, the Zephyrus G15 worked pretty well under Linux, with only a small amount of finagling. We did need to do a full install, since the live boot environment from an installer stays in safe video mode.

Given the bleeding-edge nature of the laptop—and the dubious Ubuntu results we had with last year’s similarly-designed Zephyrus G14—we decided to go with the Ubuntu 21.04 final beta, rather than the 20.04 LTS. We also didn’t want to screw up the Windows installation on the C: drive, since this was neither my own personal laptop nor a review unit.

This Orico portable SSD from the low-end USB storage shootout made a great testbed to try a full Ubuntu installation on the laptop without screwing up its C: drive.
Enlarge / This Orico portable SSD from the low-end USB storage shootout made a great testbed to try a full Ubuntu installation on the laptop without screwing up its C: drive.
Jim Salter

Booting into the Ubuntu 21.04 installer went off without a hitch, as did the actual installation onto the Orico portable SSD we plugged into another of the laptop’s USB ports. Unfortunately, the new installation wouldn’t boot—as usual, the default Nouveau open-source driver for Nvidia GPUs crapped in its own mess kit. In order to work around this, we rebooted the system and pressed “escape” to get the GRUB menu, then selected safe graphics mode boot.

In safe graphics mode, we had no difficulty getting to the desktop—and from there, we opened up the “Additional Drivers” applet and selected the current proprietary Nvidia driver for our RTX 3070. One reboot later, everything was running fine—we had sound, 3D-accelerated graphics, Wi-Fi, and everything else. The only unclaimed devices were the fingerprint readers, a Renoir IOMMU, and a Renoir audio processor (which did not prevent overall sound from working fine).

From here, we ran the glmark2 benchmark to confirm that our Nvidia RTX 3070 really was running, and running properly; then dove into the battery test. For battery testing on Linux, we used the BBC’s “Open Ocean: 10 Hours of Relaxing Oceanscapes” video on Youtube. A script echoes the battery percentage, date, and time to a logfile once per minute, and we let it run down from there.

The ROG G15 went from 98% charge to 3% charge in two hours, 48 minutes before shutting down. This certainly isn’t phenomenal, but it’s not unusable, either. (For what it’s worth, Ubuntu estimated about 10 hours total runtime on battery while we were faffing about on the desktop.)


Asus’ ROG Zephyrus G15 is a heck of a laptop. Despite being a flat-out gaming machine, it’s not very bulky, it doesn’t feel crazy heavy (4.2lbs / 1.9kg), and its fan profile is pretty modest. While doing “normal laptop stuff,” including tons of software installation with its attendant gigabyte after gigabyte of file decompression, the fans never even spun up to audibility in our relatively-quiet office. (They only became noticeable during Time Spy and Time Spy Extreme testing.)

There wasn’t a single piece of hardware in this laptop we disliked—the SK Hynix SSD is blistering fast, the 165Hz display is oily smooth with great color range and decent (270 nits) brightness, and the CPU and GPU are both UFO-level fast as well. The Wi-Fi chipset is top-of-the-line Intel Wi-Fi 6 2×2, and there’s both a full-size HDMI out and a full-size, non-collapsing RJ-45 Ethernet port!

The one piece of hardware we’re mad about is the one which wasn’t present in the G15 at all—a webcam. Just like last year’s Zephyrus G14, the G15 is cameraless. There are enough USB ports available that it’s probably not a big deal for most people to attach one when they need it, but we maintain that you shouldn’t have to. We’re a year-plus into our current reality, and the pandemic has elevated the importance of videoconferencing beyond anything the world’s ever seen before. New devices should have webcams by default, period.

At least the battery life in the G15 is great—for a gaming laptop, it’s phenomenal; for a high-end non-gaming laptop, it would still be solid. This machine (again, apart from the curious lack of webcam) makes a great daily driver, not just a laser-focused FPS-monster. Yet it is an FPS-monster, with gaming performance roughly equivalent to a high-end desktop PC with an RTX 2070 Super.

The good

  • Blistering-fast CPU, GPU, and SSD
  • Enormous 90Wh battery
  • Restrained aesthetic without too much “gamer bling”
  • Reasonably svelte and lightweight
  • Top-quality Wi-Fi
  • Full-size HDMI and non-collapsing Ethernet ports
  • Butter-smooth 165Hz 1440p display
  • Equivalent gaming performance to a 2070 Super equipped desktop gaming PC

The bad

  • Keyboard backlights flash “U” in Morse Code when the laptop sleeps (can be disabled in Aura controls)
  • Weighs more than a European swallow

The ugly

  • No webcam
  • Still no webcam
  • Once more for the people in the back: why isn’t there a webcam?!

Listing image by Jim Salter

The most inescapable thing about driving a Rolls-Royce Ghost is never being able to forget how much it costs. So let’s get that out of the way: with a sticker price of $460,350, it’s a car that costs more than a house in many parts of the country. This knowledge never escapes you, a constant reminder that you’re visiting a world where just the cost of the options fitted to the car seen here would buy a pretty good pre-owned Mercedes S-Class.

You remember it when you park or whenever another car gets a little too close in traffic, given the pricey consequences of a scratch or scrape. But the price tag also looms when you feel the weight of the metal switchgear or your feet sink into the deep shag of the lambswool carpets. You quickly realize the amount of time and materials that went into putting everything together.

The award for most improved goes to…

Despite the above, it was hard not to like the 2021 Ghost. When we tried the previous model a couple of years ago, it was similarly solid and sumptuously trimmed, but it thoroughly underwhelmed, particularly from behind the wheel. Part of the problem, perhaps, was that car sharing a platform with BMW’s 7 Series.

But that’s a problem no more, for the bones of the new Ghost come from Rolls-Royce’s new aluminum spaceframe platform (called Architecture of Luxury), shared with the Cullinan SUV and the even more expensive Phantom. The result is a car that’s as enjoyable to drive as the previous one wasn’t.

In the past, writers have likened the experience of driving a Rolls-Royce to piloting the drawing-room of a British stately home. I don’t think that’s an accurate analogy for the 21st-century Ghost, though. The chassis—made from extruded aluminum box sections—is too stiff and the ride and handling too composed to draw unfavorable comparisons with masonry and furniture.

Ironically, that very stiff chassis is one of the reasons the Ghost really does offer a magic carpet ride. It has a rather clever “planar suspension system” that does things like use GPS data and the Ghost’s forward-looking optical sensors to anticipate how firm the air springs should be at any given moment. It does a remarkably good job of soaking up big bumps, and this “road preview” function is now beginning to show up on other vehicles, too. But no other car also features an extra damper for each of the front upper wishbones. These soak up the high-frequency shocks and jolts that aren’t filtered out by the air suspension.

As a result, the Ghost seems to float, but never wallow, down the road, almost devoid of mass; in fact, it tips the scales at 5,445 lbs (2,470 kg). There are no drive modes, nor paddle shifters for the eight-speed automatic transmission, just a column stalk to select drive, neutral, reverse, or park. Instead, like the suspension, the gearbox—yet again ZF’s 8HP for those keeping score—uses GPS data to select the appropriate gear at the appropriate time. It all goes on behind the scenes and just works.

In the city, the Ghost is not a car you feel like driving quickly. A lot of that is its size: at more than 218 inches (5,545 mm) long, it is one of the bigger cars on the road, and as mentioned already it’s hard to be anything but hyper-aware of the inevitably massive bill that would result from even a little scrape. That slight hesitation before the power arrives also contributes—this just isn’t the sort of vehicle that you throw into the first available gap in traffic.

The Ghost is a joy to drive out in the country as it glides along the lanes.
Enlarge / The Ghost is a joy to drive out in the country as it glides along the lanes.
Jonathan Gitlin

That sort of behavior might upset whoever is fortunate enough to be riding in the back. And in this particular Ghost, rear passengers would be very fortunate, with not one but two drinks cabinets: one refrigerated with a pair of champagne flutes, the other with a decanter and glasses. The rear seats recline this way and that, and you’re ensconced behind a thick C pillar, hidden from potential paparazzi. Large screens are concealed behind the seat-back tables, giving either rear seat occupant full control over the Ghost’s infotainment system, and my wife spoke highly of the passenger experience.

On a winding road you tend to give the throttle just a little touch here and there, coasting along using the car’s momentum to make progress. The steering is fingertip-light, but it’s accurate and precise at the same time. You can purr along with the engine power gauge—nothing so common as a tachometer here—reading just 10 percent, almost completely isolated from the outside world. Honestly, the more I drove it, the more I understood why people get the urge to turn them into rally cars.

If you do decide to explore the higher reaches of the 563 hp (420 kW) engine’s output, the 6.75 L twin-turbocharged V12’s characteristic growl is present but quiet; this is no Ferrari or Aston Martin, after all. There’s a little too much of a delay between the initial input of the throttle pedal and the power arriving, noticeably from a standstill. It’s as if the engine is taking a deep breath in those first couple of hundred milliseconds—whether it’s the turbos spooling up or simply a pedal programmed to German tastes (which prefer a 0.25-second pause between input and reaction, according to some engineers in the industry), I’m not sure.

The downside to all of this is a fearsome thirst, even when driven gently. Although rated at a combined 14 mpg (16.8 l/100km), the onboard data revealed this particular Ghost had averaged 16.5 mpg (14/3 l/100km) over the course of nearly 700 miles (1,100 km). You can’t help but think that an electric Rolls-Royce is the answer. The more immediate torque of an electric motor would overcome the car’s substantial mass from a standstill, for one thing. Adding batteries would obviously involve adding even more weight, but there might be a quite significant weight saving without the need to insulate the occupants from engine noise. And the designers could get away with much shorter overhangs, which would result in a Rolls-Royce better optimized for life in town, where most of them spend most of their lives.

Wouldn’t an electric Rolls-Royce be even better?

The good news is that Rolls-Royce agrees with that assessment. “We’ve gone on the record saying that we will have an electrified Rolls-Royce within this decade, and quite possibly sooner,” Ghost engineering lead Jonathan Simms told Ars. The company has yet to confirm if it’ll be a standalone model or an EV based on something currently in the company’s line. “So the nice thing for us, at Rolls-Royce, is we fully agree with your comments.”

Most Rolls-Royces spend most of their lives at lower speeds, and in the city. Which means they're ideal for electrification—something Rolls-Royce agrees with. Watch this space...
Enlarge / Most Rolls-Royces spend most of their lives at lower speeds, and in the city. Which means they’re ideal for electrification—something Rolls-Royce agrees with. Watch this space…
Jonathan Gitlin

Simms said perhaps the biggest challenge to fully realizing a Rolls-Royce EV is just the limitations of electric vehicle ownership in general—not every community is currently set up with reliable charger access or other niceties that may normalize the experience. “You’ll never get past a Rolls-Royce press release without reading the word effortless,” he said. “And as you’re surely aware, at the moment, in many places in the world, owning an electric car is not described as effortless. There’s very pleasant elements to driving and owning them, for sure, but the entire experience end-to-end is not one that’s marked by that word.”

Still, an electric Rolls-Royce is coming, sometime. And without hesitation, Simms said the company has the technology and they are working to bring it to market.

“There’s a number of brands at the moment you see are really struggling to get their head around what electrification means with their brand values—everything from creating standalone model lines through trying to integrate some kind of hybrid into a hypercar, and everything in between,” he said. “Actually for us, it’s very simple, because electrification is perfect for our brand values. As you say, it’s silent, it’s instant torque, fairly often our cars are used at lower speeds—it fits perfectly.”

Listing image by Jonathan Gitlin

Farah Alibay is living her dream to unlock the mysteries of Mars. But the 28-year-old Canadian engineer is also helping to chart a new path in aviation history — and doing it on another world.

Alibay is part of a team of engineers that designed and tested a space helicopter — nicknamed “Ingenuity” — that is set to take off over the red planet in the next few days.

It is a major “first” in space exploration — namely, the first time an autonomous aircraft has ever taken flight on another planet.

Read more:
NASA rover Perseverance takes first spin on surface of Mars

“If we can demonstrate that we can fly on Mars, that opens all sorts of opportunities for future missions,” Alibay told Global News’ The New Reality.

“Right now, we’ve been limited to driving on Mars, which is still super cool, but it’s fairly slow,” she said.

Ingenuity was affixed to the Perseverance rover that landed on the red planet on Feb. 18. Earlier this month, the chopper “emerged” from the belly of the rover, and is now operating on its own.

“I mean, we’re flying on another planet. Come on!” Alibay said.

“It’s kind of crazy because we’ve only been flying on Earth for about 100 years. And now we’re saying we’re going to go to another planet where gravity is different.”

Detached from the rover, the helicopter, which weighs less than two kilograms, has had to survive temperatures that dip to -90 Celsius, drawing upon enough of its own stored energy to remain functional. Mars receives only about half of the solar energy that reaches Earth during the daytime. Building a machine that can stay warm and power itself up enough to fly under those conditions — a mere 272 million kilometres from Earth — is not easy. That, and the fact that Mars’ atmosphere is just one per cent as dense as the Earth’s, makes it extremely challenging to build a flyable helicopter.


Click to play video: 'NASA successfully lands first helicopter on Mars'

NASA successfully lands first helicopter on Mars

NASA successfully lands first helicopter on Mars

Overcoming hurdles

If landing a rover, and flying a helicopter, on Mars came with plenty of twists and turns, so too did Alibay’s path to working on the mission.

Her job at NASA began with an internship — and an unlikely one at that. She submitted many applications before she landed the job. “I probably had like 50 of them rejected. And one day, someone … at a dinner, at a conference, took interest in my research and offered me an internship.”

The Montreal native moved to England at the age of 13, with what she describes as wobbly English. She says she “went to public school; I didn’t have access to privileged education,” and yet she was accepted to prestigious universities, first at the University of Cambridge, and then at the Massachusetts Institute of Technology, where she completed her PhD. Her parents were role models, but when it came to the field of space exploration, there were not many people she could look up to and identify with.

“I didn’t have someone who looked like me in this position,” she says. “I grew up in the ’90s so there was a lot of interest in space, but a lot of people in these positions were white men, and it took a while for me to even allow myself to dream that I could be part of these people.”

But she persevered, and, in her role as a systems engineer at Nasa’s Jet Propulsion Laboratory, emphasizes that “it’s really important, now that I’m here, to show girls, to show minorities, that, hey, it doesn’t matter what you look like,” she says. “There’s a place for you here.”

Read more:
Take a look around Mars with Perseverance rover’s HD photo panorama

Bridging the arts & sciences

That story resonates deeply with Chimira Andres, a planetary geologist from Barrie, Ont., who, like Farah Alibay, grew up feeling dismayed at the lack of representation in the field by women, and especially women of colour. That’s finally changing now, Andres said.

“It’s just great to see that diversity in mission control that wasn’t evident before,” she said. “There are a lot of inspirational people out there, but if you can’t really see them, you can’t see yourself in that position as well.”

Her path to space was also rather unusual. She began her career as a dancer — and has since come to discover the many parallels between dance and the world of space exploration.

“I wanted to pursue science and space, and dance, at the same time, and just didn’t know how,” she said. Dance, she says, has taught her “everything from discipline to resilience” — skills that are critical for success in space.

After nearly two years at the Canadian Space Agency, Andres has spread her wings to Europe.

Last September, she landed a job in the Netherlands at the European Space Agency, where she is a graduate trainee in the agency’s science and technology education program.

“I really want to spread the love for STEM, the love for space,” she says. “Space is contagious, I think. So once you’ve talked about space to someone … you’ve got them hooked.”

Read more:
Wobbly muon particles hint at a new, secret force of nature

Homegrown talent

Though Alibay and Andres are both working abroad, there is no shortage of brainpower in Canada when it comes to aerospace engineering and space exploration. Where Canada excels is in education, and post-secondary programs here attract the best and brightest from all over the world.

That includes aerospace graduate Eitan Bulka, who came to Montreal from Boston a decade ago, and recently defended his PhD at McGill.

“There are a lot of opportunities here in Montreal, so I’m actually only looking to stay in Montreal,” he told Global News.

His work looks at increasing the autonomous capacities of unmanned aerial vehicles. These include everything from drones that deliver goods purchased online to helicopters and other aircraft that can fly autonomously on other planets — including Mars.


“It’s challenging enough to do this on Earth, and then there are extra challenges on Mars,” he says. “But, a lot of the tools that we develop on Earth are also applicable on Mars.”

He says that the idea that Canadians “go abroad” to pursue dreams in space — or that there is somehow a brain drain —no longer really holds.

“In the past, you kind of had to move to California. But now a lot of these (aerospace) companies are … in Canada and in Montreal, and they’re more open to having remote employees.”

And yet, the aerospace field in Canada is still “somewhat scattered,” says Inna Sharf, the director of McGill’s aerospace mechatronics lab. There are dozens of aerospace companies in Canada, from startups to large employers, and several university research groups working on everything from space engineering to self-driving cars, but “we don’t have NASA. … We don’t have anything uniting all this work, pulling it together.”

Read more:
How cigarettes, coffee and Canadian engineers helped put men on the moon (July 22, 2019)

Challenging terrain

There are definitely no shortcuts in getting to Mars, and the Perseverance mission is proof of that. NASA has landed the rover — and the Ingenuity helicopter with it — in one of the most inhospitable parts of the planet.

The area is known as Jezero Crater, and NASA chose the site because there is evidence of an ancient river delta that may have been able to support life more than 3.5 billion years ago. The crater has been shown to contain clay deposits, which can only form when there is water. Similar clay formations are found in the Mississippi River delta.

And making those connections with Earth is where Andres, the planetary geologist from Barrie, excels. She began her space career by learning about our planet. “I majored in Earth and environmental sciences,” she said, “because I heard that to be a good planetary scientist, you had to be an expert on Earth first.”

Of course, Mars is a “whole other world.” Yet, the parallels to science on Earth, whether it’s about engineering an autonomous aircraft or mapping ancient glaciers and streams, are remarkable.

“Space,” Andres says, “is not just rocket science. It’s not just biology. It’s all of them. It’s physics, math, combined with Earth sciences, and everything that you need, to help scientists bring a mission to Mars.”

See this and other original stories about our world on The New Reality airing Saturday nights on Global TV, and online at

Illustration of a house, with frayed fiber wires that don't extend far enough to connect the home to Internet service.
Aurich Lawson | Getty Images

When Edward Koll and his girlfriend, Jo Narkon, bought and moved into a new house in Draper, Virginia in late September 2020, they had every reason to think that Comcast Internet would be readily available.

They had done their due diligence before buying the house, plugging the address into the Comcast website to make sure it had cable Internet and contacting Comcast directly to make extra sure. Both the Comcast website’s availability checker and a Comcast sales rep confirmed to them that the house had service.

But Comcast had given the couple false information because of an error in its coverage map, and it didn’t tell the couple about this mistake until after they bought the house. That was over six months ago, and Koll and Narkon still don’t have home Internet service.

“It wasn’t until after we closed on the house that Comcast finally notified us that we were in a ‘dead zone,'” Koll told Ars when he first contacted us on February 12, 2021. “The ends of our street are hooked up, but the middle where we live does not have the service pedestal.”

This wasn’t the first time that Comcast, the nation’s largest cable company, has falsely told customers that service would be available.

Even though it was Comcast’s fault, the company demanded $5,000 up front to extend its network to Koll and Narkon’s house. The couple eventually paid up in late November after determining it was the only viable option. But while Comcast initially promised service within 90 days of receiving the $5,000 payment, it failed to complete the project in that time frame.

Comcast now says the project will finally be completed by April 15, over six weeks after the original 90-day estimate passed on March 2. The couple has been using mobile hotspots for Internet access the entire time they’ve lived in the house because of Comcast providing false availability information and then failing to complete the installation on time.

Comcast: “[We] did not live up to our standards”

Comcast seemed to take the problem more seriously after Ars contacted the company’s public relations department, eventually promising a $5,000 credit.

“Mr. Koll’s experience did not live up to our standards and we are very sorry for the inconvenience this has caused, therefore, we will be applying a credit to his account equal to $5,000,” Comcast told Ars on March 29, after a few weeks of back-and-forth communication.

“The shocking part of it was Comcast’s refusal to even own this,” Koll told Ars in a phone interview last week, after Comcast agreed to the $5,000 credit. “They just kind of brushed it off and we really did not get any traction on this until we reached out to you [at Ars].”

A Comcast rep called the couple yesterday to confirm that service will be installed on Thursday, April 15, sometime between 8am and 10am, Koll told us. Even after that call, Koll said he is “not convinced” the project is far enough along to be finished by then. But they’ve ordered Internet, TV, and phone service, and the couple is switching their cell phones over to Comcast’s mobile service. The total price is estimated at $249 a month, Koll said. At that price, the $5,000 credit will last about 20 months. Virginia is one of the states where Comcast reluctantly agreed to delay data-cap enforcement until 2022.

No option but to pay $5,000

Between emails, phone calls, and in-person visits, Koll said he had dozens of contacts with Comcast over the past six months to try to sort the mess out. “Comcast said, ‘yeah, our agents told you that service was available and they promised you that it was right there at your doorstep, but they were wrong,'” Koll said.

One installer arrived at the house and “said, ‘I need to go find out where your service line is,’ and then he disappeared. He never came back,” Koll told Ars. “And then we had two other installers and there was finally the third guy who said, ‘You don’t have a service pedestal close enough and you’re looking at a hefty bill just to have it installed.'”

After hearing Comcast’s demand for $5,000, Koll complained to the FCC but the agency “followed up with Comcast and accepted their explanation without comment,” Koll told Ars. Koll said he also contacted the offices of US Sens. Tim Kaine (D-Va.) and Mark Warner (D-Va.) without any results.

Edward Koll at his home.
Edward Koll at his home.

Gigabit Internet “is available for your address”

Koll provided Ars a copy of the $5,000 invoice dated November 30 along with chat and email transcripts from conversations he had with Comcast.

In an online chat before they bought the house, Koll asked a Comcast rep to verify that gigabit Internet service and cable TV would be available. The rep told him that “the Gigabit Speeds Internet plan is available for your address” and that a 2-year contract would cost “$169.99/mo plus equipment, taxes, and fees.”

The rep told Koll to order service five to eight business days in advance in case cable needed to be run to the house, making it sound like it would be an easy process and not a six-month ordeal. But the Comcast rep was wrong because Comcast’s system was providing false information. Koll placed an order for Comcast service over the phone on August 25, and the couple closed on the house on September 24, moving in the same day. Even when accepting the order, Comcast didn’t realize that it hadn’t extended its network close enough to the house to provide service.

Comcast boasts about its gigabit network in a slide from a press kit.
Comcast boasts about its gigabit network in a slide from a press kit.

On November 1, 2020, Koll replied to an email from a Comcast rep in the executive customer relations department, and pointed out that Comcast’s service map showed his address as having service. Koll’s email also pointed out that he had placed an order for service on August 25 “and at the same time put a deposit into the account so that service would be installed on or before closing on this house.”

“Given the multiple assurances from Comcast for service and the lengthy delays we have already endured, I anticipate a quick resolution to this matter,” Koll wrote in that email.

But there was no quick or cheap resolution. “We reluctantly agreed to pay $5,000 in November to have the service pedestal installed despite Comcast’s earlier promises regarding existing service availability,” Koll told Ars, months later.

“We really need help”

Koll first contacted us after reading one of our articles about a 90-year-old AT&T customer, Aaron Epstein, who bought an ad in the Wall Street Journal to complain about AT&T’s slow Internet service. At the time, Koll wasn’t sure that Comcast would install service at all.

“We really need help, similar to Mr. Epstein,” Koll told Ars in his first email to us. “We need to get the service we were promised installed. We would like to have the $5,000 back, but at this point are willing to accept that as a credit towards future bills. Comcast has offered two free months, which does not come close to the amount we have paid.”

Although Comcast later agreed to a credit equal to the full amount they paid, the company did not tell us why it won’t refund the $5,000 instead of providing it as a bill credit. That essentially means they had to pay $5,000 in advance for about 20 months of service, an initial expense that new Comcast customers normally would not have to pay.

Comcast fixes map, months too late

In late March, Comcast finally updated its coverage map and online availability checker to correctly show Koll’s address as unserved. So why did Comcast’s website and a Comcast sales rep falsely tell Koll that service was already available at the home?

“This is not a common occurrence and we have to look into these instances on a case-by-case basis,” Comcast said. “For this specific address, it was included in a database from a company we acquired many years ago. Our local teams are planning a further review of that regional database.”

Cable companies have failed for years to develop more accurate service maps. The cable industry’s biggest lobby group fought against requirements to submit address-level data to the Federal Communications Commission, telling the FCC in 2017 that it’s too difficult and expensive to figure out exactly where they can and can’t offer service.

The FCC finally ordered ISPs to submit geospatial maps of where they provide service in 2019, but then-FCC Chairman Ajit Pai did not finish setting up the new system before leaving office in January 2021. Upon taking over from Pai, Acting FCC Chairwoman Jessica Rosenworcel learned “that we had a lot of work to do and hadn’t yet begun many of the steps required to actually build a collection system,” she wrote.

Comcast: “It’s on us and we are very sorry”

Comcast has to work with the local power company to extend service to the house, but Comcast admitted that it didn’t start that process as quickly as it should have. Comcast told Ars that installers who went to the home and found that they could not install cable service did not communicate that fact to the construction team. The lack of communication “caused initial delays. It’s on us and for that we are very sorry,” Comcast said.

Comcast told us on March 29:

Once the right teams were engaged, we began the process to extend service to Mr. Koll’s home. Before we can start construction, we need to obtain a permit to attach our lines to the utility pole near his home. We estimated 90 days for the process, but sometimes, like in this case, it may take the pole owners more time to get the poles ready before issuing the permit. We have been in touch with the pole owner [American Electric Power] and they expect to complete their work this week. Once that is complete, we can begin our construction, which we expect to be complete on or before April 15.

Surviving on mobile hotspots

To get by without cable Internet, the couple bought three extra cell phones to use solely as hotspots. When the data plan on one runs out, they turn on the next.

“We had to purchase three additional cell phones and go with Verizon’s largest data plan, which is 30GB a month [on each hotspot], and we’re using up all of that data,” Koll said. “It’s called very judicious use. We’re not streaming anything, we’re not even playing music. It’s been a difficult six months.”

The hotspots have 4G speeds. “Where we live, the connection is not that great,” Koll said. “We get two bars on a good day. It’s fairly reliable at two bars, but once you’ve hit your data cap on a given phone it drops it to 600kbps.”

Picture of a Comcast router/modem gateway from the company's website.
Picture of a Comcast router/modem gateway from the company’s website.

Verizon offers landline telephone service in the area but it’s too far from the central office for DSL Internet, Koll said. Verizon offers fiber-to-the-home service in some other parts of Virginia.

“We went round and round with Verizon, looking for the most economical way to be able to operate for the last six months, and this was it,” Koll said.

Koll, 52, is a consultant and project manager. He works at home and needs a VPN for his job. Narkon was working in an office until the pandemic began and hasn’t gotten a work-at-home job in part because of the couple’s extremely limited Internet access, Koll said. “Two of us trying to work full-time off of a hotspot, it just wouldn’t work,” he said.

They had tried HughesNet satellite where they previously lived but the connection wasn’t good enough to maintain the VPN connection, and satellite “data caps were prohibitive,” Koll said. They didn’t have wired Internet access at their previous home, and they were looking forward to finally getting fast broadband at the new house.

“It wasn’t the driving factor, but a consideration in purchasing this home was dedicated Internet service that already existed or could easily be turned on,” Koll said.

Frustrating delays

Shortly after receiving the $5,000 payment, Comcast told Koll that the 90-day timeline could be extended by factors such as weather, permitting delays, and the pandemic, he said. Koll was surprised. “We’ve been dealing with COVID for a year,” he told Ars. “Most places have figured out the impact of COVID on their line of business, how to work around it.”

Koll exchanged emails over several months with a Comcast construction coordinator and a member of Comcast’s executive customer relations department. In early December, Comcast said the project would probably be done by mid-March, only a couple weeks past the 90-day time frame. But it later became clear that wouldn’t happen and Comcast was slow in providing a new estimated date of completion.

Koll had to repeatedly follow up because of delays in getting responses, including once when the construction coordinator misspelled Koll’s email address and another time when the executive department rep said an email “got stuck in my outbox.”

On January 14, Comcast told Koll that it was “waiting on permit approvals” and couldn’t yet provide a “better estimated completed time.” Koll asked Comcast for updates on the project status on February 4 and February 12 but got no response. Finally, on March 1—five days after Ars contacted Comcast’s public relations department about the case—he was told it would be completed by April 15.

“My apologies it looks like my response got stuck in my outbox the ETA is 4/15,” the March 1 email from the Comcast executive customer relations representative said.

“No concern for us as a customer”

Koll described his frustration with Comcast in our phone interview last week. “Even after we signed the contract with them and gave them the money to do the construction, there was really no concern for us as a customer,” he said.

Koll was disappointed that Comcast didn’t seem to take their plight seriously until Ars contacted the company.

“You reached out to Comcast and now all of a sudden, we’ve got traction from their PR department, from their construction department,” he said. “I actually got an in-person visit from [the Comcast construction manager] last week, giving me the rundown and the breakdown on everything that was going to happen and the order it was going to happen in.”

“I’m a project manager,” he also said. “The number one rule, especially when communicating with anybody, is making sure I’m setting the right expectations. Up until last week when the [Comcast] PR department contacted me, the expectations weren’t even being set, let alone met.”

Though it looks like everything is finally on track, Koll said that he and his girlfriend weren’t fully convinced they’ll get service by April 15. “I think we will both be surprised if it actually happens,” he said.