Meteor Lake seemingly struggles against AMD's Phoenix APU in early MSI Claw review
Performance issues could be due to unoptimized software
An early review for the Meteor Lake-powered MSI Claw gaming handheld is out, and its performance is clearly behind the ROG Ally and its Phoenix-based Z1 Extreme (via VideoCardz). Meteor Lake is significantly less performant than Phoenix in a handheld form factor according to the reviewer. The comparison is especially unfavorable for Intel when using lower TDPs, where the Ally's Z1 Extreme could be up to twice as fast as the Core 7 Ultra 155H in the Claw.
While Intel was mostly focused on AI with the development of Meteor Lake, it does feature significantly more powerful graphics than prior Intel CPUs. Its Xe-LPG-based integrated graphics have a complete tile to itself and on paper are pretty powerful. Seemingly a big improvement over the aging Xe Iris graphics that first debuted with Tiger Lake in 2020. There was even some hope that Intel could catch up to AMD in iGPU performance, which is generally something that AMD always does well in.
Unfortunately for Intel, according to the early review by Please, Xiao Fengfeng it looks like Meteor Lake doesn't have what it takes to go head-to-head with AMD in gaming handhelds. The reviewer put the Core 7 Ultra 155H version of the MSI Claw up against the Z1 Extreme model of the ROG Ally, both of which are the top-end variants.
At their default TDPs (35 to 40 watts for the Claw and 30 for the Ally), the 155H was behind in every single game by significant amounts. Unfortunately, only footage of the benchmarks was provided rather than an average framerate, but at every point the Z1 Extreme had at least a 10% performance lead, and often closer to a 30% to 40% higher framerate. For one brief moment in the Resident Evil 4 remake, the Z1 Extreme had over double the framerate. That's not just lower performance, but lower performance while using more power.
Row 0 - Cell 0 | MSI Claw (Core 7 Ultra 155H) | ROG Ally (Z1 Extreme) |
25 Watts | 47 | 54 |
20 Watts | 40 | 53 |
15 Watts | 24 | 42 |
The margins got even worse for the 155H when comparing the two chips at 25 watts, 20 watts, and 15 watts in Shadow of the Tomb Raider, where we finally have average framerate figures. Meteor Lake isn't so far off at 25 watts, but it loses a substantial amount of performance at 20 watts while Phoenix doesn't. At 15 watts however, the 155H is just about half as fast as the Z1 Extreme. Just like with the default TDPs, the review indicates that Meteor Lake's graphics are much more inefficient than Phoenix's, and that's especially true as the TDP decreases.
While it may be possible that some of this performance and efficiency disparity is down to the Claw, which hasn't even launched yet and may not be fully optimized on a software and firmware level, we already knew Meteor Lake's integrated graphics weren't perfect. Benchmarks from December showed that Meteor Lake's graphics efficiency declined dramatically at lower TDPs, which lead to a conclusion that Intel's AI-focused CPU probably wouldn't be ideal for a handheld. Intel remains in second place with its integrated graphics, though the gap has shrunk with Meteor Lake.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Matthew Connatser is a freelancing writer for Tom's Hardware US. He writes articles about CPUs, GPUs, SSDs, and computers in general.
-
salgado18 It's amazing how can vendors still use Intel hardware for power-limited scenarios. Time and again it has been proved that AMD has more efficient CPUs, but do they listen? Noo, a very cheap chip (it has to have a big discount) is too tempting to pass. But now we are stuck with the majority of notebook designs using Intel, and lacking options woth AMD hardware.Reply -
Notton Just 2 months ago, you had a news article about Meteor Lake being 10% faster than 780M at 28W TDP.Reply
https://www.tomshardware.com/pc-components/gpus/intel-shows-first-integrated-arc-gpu-benchmarks-for-meteor-lake-up-to-twice-the-graphics-performance-compared-to-i7-1370p
What changed since then?
And I checked, the only difference between 155H and 165H is a 50Mhz bump in GPU clock speed. -
peachpuff
Those are intel supplied numbers and are in a laptop not a handheld, that's the issue.Notton said:
What changed since then? -
salgado18
That's according to Intel, in unspecified scenarios. Cyberpunk 2077 (according to them) is 4% faster, but we don't know the specific system they used. Did they use a highly ventilated notebook which let the turbo boosts go wild, while keeping the claim of 28W with many caveats?Notton said:Just 2 months ago, you had a news article about Meteor Lake being 10% faster than 780M at 28W TDP.
https://www.tomshardware.com/pc-components/gpus/intel-shows-first-integrated-arc-gpu-benchmarks-for-meteor-lake-up-to-twice-the-graphics-performance-compared-to-i7-1370p
What changed since then?
And I checked, the only difference between 155H and 165H is a 50Mhz bump in GPU clock speed.
Want to be sure? Wait for third party benchmarks. But there is no miracle: Intel stuff today is hot and loud. -
Pierce2623
What changed is those number were in synthetic benchmarking and everybody knows Intel tunes their stuff to score unrepresentatively good in benchmarks.Notton said:Just 2 months ago, you had a news article about Meteor Lake being 10% faster than 780M at 28W TDP.
https://www.tomshardware.com/pc-components/gpus/intel-shows-first-integrated-arc-gpu-benchmarks-for-meteor-lake-up-to-twice-the-graphics-performance-compared-to-i7-1370p
What changed since then?
And I checked, the only difference between 155H and 165H is a 50Mhz bump in GPU clock speed. -
DaveLTX
And also, cranking up power limits absolutely helps when the arch is less efficientPierce2623 said:What changed is those number were in synthetic benchmarking and everybody knows Intel tunes their stuff to score unrepresentatively good in benchmarks.
🤣
No surprise meteor lake is less impressive than Hawk or Phoenix. Then tack on how varying Arc is. It's not outright broken at times like it was but it still goes from broken to working across titles. Especially if you run a older title
And also, what on earth is "unoptimized software" when it's running windows and x86?