I wrote that no one should buy MSI and Intel’s original handheld gaming PC. I literally called it an embarrassment, and the company blacklisted me after that. MSI stopped pitching me news, and stopped answering my emails, even after the company began to write off its dud of a handheld.

So you can imagine my surprise to find: MSI and Intel have gone from worst to nearly first. In many games, it’s twice as fast as the original Claw. And with new drivers that bump its performance up to 30 percent higher since launch — I tested — the newer $1,000 MSI Claw 8 AI Plus might just be the best Windows handheld you can buy.

The Claw 8 has become a fixture in my life as I carve and puzzle through the hauntingly beautiful painted worlds of Clair Obscur: Expedition 33 and Blue Prince, respectively. I have a lot of other handheld review units floating around, but I rarely want to reach for a different one.

Let’s get this out of the way: it’s called the “AI Plus” because this “AI Engine” is supposed to automatically configure performance using Intel’s NPU. But it’s very dumb in practice and not a reason to buy.

Yes, it runs Windows, and if you’ve read any of my handheld reviews you’ll know how I feel about that. Windows 11 has become a bloated mess filled with annoying upsells and unwanted AI cruft, is annoying to navigate by controller alone, and often wakes up poorly from sleep. I can’t count the number of times I’ve sworn at Windows handhelds for popping awake in the middle of the night, or spontaneously deciding it was time to turn their remaining battery life into hot exhaust in the middle of my sealed backpack.

But the MSI Claw 8, with Intel’s Lunar Lake, isn’t as bad as the Windows norm. When you combine that with the newly improved performance, the longest battery life of any handheld in all but the most lightweight games, and an excellent 8-inch screen, it’s enough to knock the Asus ROG Ally X off its high perch — at least until the Xbox version arrives this October.

The MSI Claw 8 AI Plus at a distance: it’s a gaming handheld PC in grey with black trim and LED lit joysticks.

$999

The Good

  • Best battery life in a handheld
  • Excellent performance after updates
  • Great variable refresh rate screen
  • Fewer Windows annoyances than usual

The Bad

  • $1,000
  • Windows is bloated and can’t be trusted to sleep
  • Iffy rumble for games
  • AI tuning feature doesn’t work well

The MSI Claw 8 AI Plus is the biggest mainstream handheld, at nearly a foot long, 5 inches tall, and an inch thick, and it’s the only current-gen handheld with an Intel chip. Last year, both size and Intel pedigree were liabilities — but thankfully every handheld maker is paying more attention to ergonomics this year, and Intel’s Lunar Lake is a big improvement.

Scalloped grips and balanced weight distribution make the Claw comfortable for me to hold, it no longer looks like a ROG Ally knockoff, and while it isn’t light at 1.75 pounds (795g), that’s only a quarter-pound heavier than the Ally X. That’s despite the Claw having an extra inch of diagonal screen real estate and the same 80 watt-hour battery capacity as Asus.

The MSI Claw is wider, taller, and roughly as grippy as the Asus ROG Ally X…

And it’s bigger and grippier than the Lenovo Legion Go S, which also has an 8-inch screen.

I honestly found it tough to go back to playing Expedition 33 on the Asus ROG Ally X after using the Claw 8, partially because its screen is more colorful and more spacious (with a far smaller bezel), and partially because the game ran smoother. (I beat the game, including the entire Endless Tower and four of the game’s toughest boss fights, on the Claw 8.)

But before all that, I had to tangle with Windows 11 — and was surprised how little detangling was needed.

The first time I fired up the MSI Claw 8 AI Plus, I couldn’t believe how fast Windows setup had become. Instead of the typical 45 minutes of annoying upsells and mandatory updates, I was able to get to the Windows desktop just seven minutes after I pressed the power button.

MSI’s “Quick Settings” are now part of the Xbox Game Bar in Windows.

That’s still slower than setting up a SteamOS handheld, but fast enough I thought there must be some mistake! (Did this really get past Microsoft certification?) Another curiosity: When I hit the button that pulls up the Claw’s quick settings menu to adjust brightness, volume, and my processor’s TDP (giving it more wattage/gas), it launched a new widget in the Xbox Game Bar with those handy controls instead of a dedicated MSI process! Did somebody accidentally slip me a preview of Microsoft’s “best of Xbox and Windows together” that’ll ship on the Xbox Ally later this year?

But I won’t lie and say the Windows experience was flawless after that. MSI’s Game Bar widget turned out to be incredibly sluggish and unreliable out of the box. It got better after I changed the Windows power mode from “Balanced” to “Best Performance,” but it still isn’t nearly as fast as Asus’ Armory Crate, which has gotten extremely responsive since the Ally first launched, or as reliable as SteamOS, which doesn’t require a separate utility at all.

Here’s how much MSI has improved:

Game and power mode

MSI Claw 8 AI Plus (Aug ’25)

MSI Claw 8 AI Plus (Jun ’25)

Claw 8 3-mo improvement

Claw battery drain (August)

Claw battery drain (June)

Claw 7 (Meteor Lake, Jun ’24)

Claw 8 Lunar Lake vs. Claw 7 Meteor Lake

AC Valhalla, 15-watt TDP 66 49 34.69% 20.5W (~3.9h) 23W (~3.5h) N/A N/A
20-watt TDP 78 59 32.20% 25.5W (~3.1h) 29W (~2.8h) 33 136.36%
25-watt TDP 84 69 21.74% 32W (~2.5h) 36W (~2.2h) 53 58.49%
30-watt TDP 85 79 7.59% 38W (~2.1h) 39.5W (~2h) 54 57.41%
Plugged in 86 83 3.61% N/A N/A 54 59.26%
Cyberpunk 2077, 15-watt TDP 51 48 6.25% 20W (~4h) 20.5W (~3.9h) N/A N/A
20-watt TDP 60 57 5.26% 25W (~3.2h) 26W (~3.1h) 32 87.50%
25-watt TDP 71 65 9.23% 31.5W (~2.5h) 31.5W (~2.5h) 49 44.90%
30-watt TDP 76 72 5.56% 37.5W (~2.1h) 37W (~2.2h) 41 85.37%
Plugged in 77 79 -2.53% N/A N/A 49 57.14%
DX: Mankind Divided, 15-watt TDP 79 66 19.70% 22.5W (~3.6h) 21.5W (~3.6h) N/A N/A
20-watt TDP 93 81 14.81% 29.5W (~2.7h) 27W (~3h) 45 106.67%
25-watt TDP 105 91 15.38% 31W (~2.6h) 32W (~2.5h) 49 114.29%
30-watt TDP 114 100 14.00% 37W (~2.2h) 37W (~2.2h) 48 137.50%
Plugged in 119 107 11.21% N/A N/A 58 105.17%
Returnal, 15-watt TDP 40 30 33.33% 20W (~4h) 22W (~3.6h) N/A N/A
20-watt TDP 48 36 33.33% 26.5W (~3h) 27W (~3h) 29 65.52%
25-watt TDP 52 41 26.83% 36W (~2.2h) 32W (~2.5h) 38 36.84%
30-watt TDP 54 43 25.58% 42.5W (~1.9h) 37.5W (~2.1h) 39 38.46%
Plugged in 56 44 27.27% N/A N/A 37 51.35%
Shadow of the Tomb Raider, 15-watt TDP 55 51 7.84% 21W (~3.8h) 23W (~3.5h) N/A N/A
20-watt TDP 66 60 10.00% 27W (~3h) 30W (~2.6h) 32 106.25%
25-watt TDP 73 68 7.35% 35W (~2.3h) 35.5W (~2.2h) 38 92.11%
30-watt TDP 81 74 9.46% 42W (~1.9h) 41W (~2h) 39 107.69%
Plugged in 81 75 8.00% N/A N/A 42 92.86%
HZD Remastered, 15-watt TDP 31 31 0.00% 21.5W (~3.6h) 24.5W (~3.3h) N/A N/A
20-watt TDP 37 37 0.00% 28W (~2.9h) 30W (~2.6h) Did not test N/A
25-watt TDP 42 41 2.44% 34.5W (~2.3h) 36W (~2.2h) Did not test N/A
30-watt TDP 44 42 4.76% 40W (~2h) 39W (~2.1h) Did not test N/A
Plugged in 46 45 2.22% N/A N/A Did not test N/A

Average frame rates at 720p and low settings, save Cyberpunk at “Steam Deck” setting.

And no, I can never, ever trust the Claw 8’s power button to keep it asleep. I always have to explicitly put the system into hibernation mode instead, which is thankfully the first option in that Game Bar widget. (Microsoft really needs to get out of its own way and let manufacturers set the power button to hibernate instead of Modern Standby.)

It’s also a little frustrating to go back to a world where I have to manually download the latest Intel graphics drivers because they never showed up in MSI Center’s updates tab. But once I installed those new graphics drivers and learned to avoid the power button, I was rewarded with some of the highest performance and the best battery life I’ve seen from a handheld yet. It’s just better at the whole turbo mode thing, with higher frame rates and higher TDP options than the Ally X with Windows, and it’s more power-efficient than the SteamOS Lenovo Legion Go S, even if Lenovo’s Steam handheld got higher frame rates in half my benchmarks.

Claw 8 vs. Ally X vs. Legion Go S vs. Steam Deck performance

Game and power mode

Claw 8 fps

ROG Ally X (Windows, Z1E) fps

Legion Go S (SteamOS, Z1E) fps

Steam Deck OLED fps

Claw 8 vs. Ally X

Claw 8 vs. Legion Go S

Claw 8 vs. Deck

AC Valhalla, 15-watt TDP 66 52 66 57 26.92% 0.00% 15.79%
20-watt TDP 78 71 84 N/A 9.86% -7.14% N/A
25-watt TDP 84 80 92 N/A 5.00% -8.70% N/A
30-watt TDP 85 89 96 N/A -4.49% -11.46% N/A
Plugged in 86 89 92 57 -3.37% -6.52% 50.88%
Cyberpunk 2077, 15-watt TDP 51 41 57 50 24.39% -10.53% 2.00%
20-watt TDP 60 59 73 N/A 1.69% -17.81% N/A
25-watt TDP 71 65 79 N/A 9.23% -10.13% N/A
30-watt TDP 76 71 82 N/A 7.04% -7.32% N/A
Plugged in 77 71 86 50 8.45% -10.47% 54.00%
DX: Mankind Divided, 15-watt TDP 79 59 74 61 33.90% 6.76% 29.51%
20-watt TDP 93 84 92 N/A 10.71% 1.09% N/A
25-watt TDP 105 91 99 N/A 15.38% 6.06% N/A
30-watt TDP 114 93 100 N/A 22.58% 14.00% N/A
Plugged in 119 93 102 61 27.96% 16.67% 95.08%
Returnal, 15-watt TDP 40 31 24 25 29.03% 66.67% 60.00%
20-watt TDP 48 40 30 N/A 20.00% 60.00% N/A
25-watt TDP 52 43 32 N/A 20.93% 62.50% N/A
30-watt TDP 54 46 33 N/A 17.39% 63.64% N/A
Plugged in 56 46 34 25 21.74% 64.71% 124.00%
Shadow of the Tomb Raider, 15-watt TDP 55 52 62 57 5.77% -11.29% -3.51%
20-watt TDP 66 65 83 N/A 1.54% -20.48% N/A
25-watt TDP 73 70 89 N/A 4.29% -17.98% N/A
30-watt TDP 81 76 93 N/A 6.58% -12.90% N/A
Plugged in 81 76 96 57 6.58% -15.63% 42.11%
HZD Remastered, 15-watt TDP 31 28 37 33 10.71% -16.22% -6.06%
20-watt TDP 37 30 47 N/A 23.33% -21.28% N/A
25-watt TDP 42 28 50 N/A 50.00% -16.00% N/A
30-watt TDP 44 34 52 N/A 29.41% -15.38% N/A
Plugged in 46 34 55 33 35.29% -16.36% 39.39%

Average frame rates at 720p and low settings, save Cyberpunk at “Steam Deck” setting.

We’re not talking about a step change in performance here: this pricy $1,000 handheld still doesn’t have near the frame rate of a similarly priced gaming laptop. It’s just enough extra power to make games feel smooth on this handheld that were borderline choppy on its peers, like Expedition 33 (even then, these handhelds need AI upscaling to get there).

But I didn’t have to sacrifice battery life to get that extra power. I often even push the Claw 8’s chip to 30 watts, a power mode that the Asus ROG Ally X doesn’t offer unless plugged in, for an extra frame rate advantage — while getting roughly the same battery life as the Ally X gets at its 25W unplugged maximum.

MSI Claw 8 vs. Asus ROG Ally X ports and width. The Claw still has handy raised dots on each port to find them by feel.

And though the Steam Deck has long reigned as the efficiency champ, I found the Claw 8 can even beat the Steam Deck on both performance and power use when set to the same 15-watt TDP, often draining its battery 2 or even 3 watts slower. When you combine that with its big 80-watt-hour pack, it handily beat competitors in my Dirt Rally drain test, lasting nearly 19 minutes longer than the Asus ROG Ally X with Bazzite, 28 minutes longer than the Steam Deck OLED, and 80 minutes longer than the Legion Go S with SteamOS and AMD’s Z1 Extreme chip.

Below, you can see how battery life might compare at each performance tier. For example: the Claw 8 and Legion Go S with SteamOS both offer the same 66 frames per second when you offer their chips 15 watts of electricity, but you’ll probably get 3.9 hours of battery from the Claw versus 2.3 hours from the Legion. That’s because MSI’s Intel chip is drawing less power from a larger battery.

Battery life vs. performance

Game and power mode

Claw 8 fps

Claw battery drain

ROG Ally X (Windows, Z1E) fps

Ally X battery drain

Legion Go S (SteamOS, Z1E) fps

Legion Go S battery drain

Steam Deck OLED fps

Deck battery drain

AC Valhalla, 15-watt TDP 66 20.5W (~3.9h) 52 24W (~3.3h) 66 24W (~2.3h) 57 23.5W (~2.1h)
20-watt TDP 78 25.5W (~3.1h) 71 30W (~2.6h) 84 30.5W (~1.8h) N/A N/A
25-watt TDP 84 32W (~2.5h) 80 36W (~2.2h) 92 36W (~1.5h) N/A N/A
30-watt TDP 85 38W (~2.1h) 89 N/A 96 43.5W (~1.3h) N/A N/A
Cyberpunk 2077, 15-watt TDP 51 20W (~4h) 41 22W (~3.6h) 57 25W (~2.2h) 50 23.5W (~2.1h)
20-watt TDP 60 25W (~3.2h) 59 29W (~2.7h) 73 32W (~1.7h) N/A N/A
25-watt TDP 71 31.5W (~2.5h) 65 35W (~2.2h) 79 38W (~1.4h) N/A N/A
30-watt TDP 76 37.5W (~2.1h) 71 N/A 82 44.5W (~1.2h) N/A N/A
DX: Mankind Divided, 15-watt TDP 79 22.5W (~3.6h) 59 22W (~3.6h) 74 25.5W (~2.2h) 61 22W (~2.2h)
20-watt TDP 93 29.5W (~2.7h) 84 30W (~2.6h) 92 32W (~1.7h) N/A N/A
25-watt TDP 105 31W (~2.6h) 91 36W (~2.2h) 99 38.5W (~1.4h) N/A N/A
30-watt TDP 114 37W (~2.2h) 93 N/A 100 45W (~1.2h) N/A N/A
Returnal, 15-watt TDP 40 20W (~4h) 31 23W (~3.5h) 24 25.5W (~2.2h) 25 23W (~2.1h)
20-watt TDP 48 26.5W (~3h) 40 30W (~2.6h) 30 32W (~1.7h) N/A N/A
25-watt TDP 52 36W (~2.2h) 43 36W (~2.2h) 32 38.5W (~1.4h) N/A N/A
30-watt TDP 54 42.5W (~1.9h) 46 N/A 33 45.5W (~1.2h) N/A N/A
Shadow of the Tomb Raider, 15-watt TDP 55 21W (~3.8h) 52 23W (~3.5h) 62 25W (~2.2h) 57 23.5W (~2.1h)
20-watt TDP 66 27W (~3h) 65 30W (~2.6h) 83 31.5W (~1.7h) N/A N/A
25-watt TDP 73 35W (~2.3h) 70 36W (~2.2h) 89 37.5W (~1.5h) N/A N/A
30-watt TDP 81 42W (~1.9h) 76 N/A 93 43.5W (~1.3h) N/A N/A
HZD Remastered, 15-watt TDP 31 21.5W (~3.6h) 28 23W (~3.5h) 37 25W (~2.2h) 33 23W (~2.1h)
20-watt TDP 37 28W (~2.9h) 30 30W (~2.6h) 47 32W (~1.7h) N/A N/A
25-watt TDP 42 34.5W (~2.3h) 28 36W (~2.2h) 50 37.5W (~1.5h) N/A N/A
30-watt TDP 44 40W (~2h) 34 N/A 52 44W (~1.3h) N/A N/A

Average frame rates at 720p and low settings, save Cyberpunk at “Steam Deck” setting.

All this said, you still can’t ease off the gas quite like with AMD chips, at least not out of the box. Officially, the Intel Core Ultra 7 258V is a 17W–37W chip, and while MSI lets you set the chip’s TDP as low as 8 watts, I still saw the Claw empty my battery at a rate of at least 11W in Balatro, the magic poker game I use as my best-case-scenario test. That means at minimum screen brightness, with wireless off, I’m getting around seven hours max — whereas the Steam Deck OLED can get nine hours and the Asus ROG Ally X can manage 10 in such lightweight games.

I have some quibbles with the Claw’s hardware. Though the speakers are above average, the rumble is annoying and weak. Hall effect joysticks and dedicated gyro modes are nice, but aiming felt sloppy out of the box, like MSI didn’t bother tuning either to a console controller standard, and I still haven’t quite nailed it with tweaks in either individual game settings or MSI Center. It’s also not the fastest handheld to charge or download games, despite its two Thunderbolt 4 ports and Wi-Fi 7, though not slow either.

But the real hurdles for the Claw 8 are that $1,000 price — and that MSI might only have one more whole month atop the Windows heap. Microsoft and Asus’ Xbox Ally is coming October 16th, with a revamped OS, and it could reshape the whole handheld market.

Handheld power, explained

You might have noticed I write about two different kind of wattage measurements (W) in my handheld reviews: 1) the TDP of each handheld’s chip, which basically translates to how much power you’re letting it use, and 2) the handheld’s total battery drain.

That’s because of a sea change in how portable gaming works. Unlike traditional laptops, today’s handhelds let you configure their processor’s TDP at a moment’s notice, even while you’re in the middle of a game, to give you more oomph. But when you do that — or when a manufacturer sets a higher default TDP so their handheld seems faster out of the box — it’ll drain your battery faster.

How much faster? You can find the answers in my charts, along with estimates of how quickly your battery will go from 100 percent to zero if you choose that TDP in a given game. And make no mistake, some games won’t run well on a handheld unless you choose a high TDP.

But the battery drain wattage is not the same as TDP, because it doesn’t account for all the rest of the handheld’s systems, including its storage and screen, that each game might push differently. Frame rate per watt drained is what to look for if you care about battery.

Photography by Sean Hollister / The Verge

0 Comments

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

Share.
Exit mobile version