Additional examples of 8-GPU M2 Air outperforming 10-GPU model?

[blakespot]

Like many prospective M2 MacBook Air buyers, I have seen GregsGadgets' game benchmark video showing the M2 Air with 8 GPU cores outperforming the 10 GPU core unit, side by side -- presumably due to more significant thermal throttling in the 10-GPU core unit due to the additional cores.

Greg uses the in-built benchmark in the somewhat aged Shadow of the Tomb Raider, which is an Intel binary running under Rosetta 2 emulation. It is, however, apparently built utilizing Apple's Metal API.

I have a 10-GPU core M2 Air on order (24MB RAM, 1TB SSD) with which I am replacing my M1 Air.

I do fairly "normal" / typical things on the Air, and so GPU has not been a big concern for me. However, in addition to these typical things, I want to play No Man's Sky for the Mac (coming later this year) on the Air. (I am a big NMS fan and play on PC and, what with it coming as an Apple Silicon native game supporting Metal 3 and MetalFX scaling, it would be nice to play it on the go on the Mac.) So, all of a sudden gaming performance on the Mac matters to me again (it's been many years since that was the case).

I am trying to find other examples of the 8 vs 10 GPU core M2 Air running under heavy CPU and GPU load, to see if there is consistency in the 8-core outperforming the 10-core due to thermal issues. If this is the case, I will cancel and re-place my order, but for an 8-GPU Air. (And I am not interested in Max Tech's thermal-pad "fix," as it heats up the battery which can shorten its life / affect performance.)

Has anyone seen other examples demonstrating this? Does anyone have any experiences to add that speak to this? I would love to hear any such reports.

Thanks


bp

 

[camotwen]

I don't know about this specific comparison, but when less cores outperform more cores, it is very often because the software is not efficiently utilising multiple cores. I would be careful in taking decisions with long term effects based on something that is too situational or may be fixed in some update.

 

[jdb8167]

I don't know about this specific comparison, but when less cores outperform more cores, it is very often because the software is not efficiently utilising multiple cores. I would be careful in taking decisions with long term effects based on something that is too situational or may be fixed in some update.

There are several GPU benchmark applications that reviewers can use to test between the 8 core and the 10 core. I haven't seen them. I know, benchmarks, but at least when comparing identical M2 MacBook Airs besides of the number of GPU cores would be a straightforward comparison even if it isn't completely comprehensive.

 

[blakespot]

There are several GPU benchmark applications that reviewers can use to test between the 8 core and the 10 core. I haven't seen them. I know, benchmarks, but at least when comparing identical M2 MacBook Airs besides of the number of GPU cores would be a straightforward comparison even if it isn't completely comprehensive.

I agree. I have repeatedly reached out to the guy that did the original video, but he hasn't responded and doesn't seem to be posting further videos, which seems odd. He has a hot topic and two such machines differing in GPU count. I am just hoping he posts further videos.

Surely some other tester out there must have two such machines. I haven't found them, though.

bp

 

[Slartibart]

IMHO No Man’s Sky will most likely profit more from the additional computing power of the M2 - the M1 and the M2 in their respective MBAs will throttle both at some point in time anyway.
Does the Δt between M1 and M2 for when this happens in a gaming scenario really matter?
Isn’t the more important question: have the developers optimize their game for AS?
If Divinity: Original Sin 2 on an M1 iPP is any indicator you’ll probably much better off with the M2 - even throttled it offers more computing/gpu power.

 

[MayaUser]

for example my 10 core gpu , my WoW (its a game) friend that has the base base M2 mba , installed WoW on mine and the 10 core gpu showed constantly 9 extra fps than the 8 core gpu
Same in starcraft 2, constantly8-9 more fps
Maybe that app is a benchmark and not a creative/ or a real game , or i dont know but it is strange...maybe due that maximize everything and reach a certain heat point and beyond that cant get any higher than that no matter how many extra cores you have since you are limited by thermal envelope ..

 

[Zest28]

Better buy the M2 13” MBP instead of the Air.

Games are probably the most demanding “tasks” put on a computer and the MBA isn’t designed for that, which is why the 13” M2 MBP still exists.

 

[MayaUser]

i think the benchmarks are the most demanding "apps" that stress the cpu+gpu 100%, the entire SoC
Games in general can be gpu heavy or cpu heavy ..thats why in games the system shows the real potential of the gpu extra core
But i agree with Zest for gaming, just dont buy a mac, buy a windows device, or for casual gaming under the mac at least buy an M1 pro/Max device
I can test my Maya projects but when i have both machines again with me.

 

[blakespot]

Adding some comparative details -- I purchased Shadow of the Tomb Raider last night on Steam to run the bench on my M1 Air with 8 GPU cores and 16GB RAM @ 1440x900. Here are the results:

Screenshot of the results: https://blakespot.com/images/SotTR_M1Air_8GPU_16GB.jpg
( Frames Rendered: 5308 , Average FPS: 33 )

 

[jdb8167]

Adding some comparative details -- I purchased Shadow of the Tomb Raider last night on Steam to run the bench on my M1 Air with 8 GPU cores and 16GB RAM. Here are the results:

Screenshot of the results: https://blakespot.com/images/SotTR_M1Air_8GPU_16GB.jpg
( Frames Rendered: 3621 , Average FPS: 23 )

I'm at work so I can't watch the video. What was the frame rate on SotTR on the M2 8 and 10 Core MacBook Air?

 

[blakespot]

Better buy the M2 13” MBP instead of the Air.

Games are probably the most demanding “tasks” put on a computer and the MBA isn’t designed for that, which is why the 13” M2 MBP still exists.

I tried to make the point that I like the Air (have an M1) and use it for my regular stuff, but given the odd chance that No Man's Sky happens to be coming to the Mac and is indeed optimized for Apple Silicon (will use MetalFX upscaler, even), I want to be able to play it -- as a bonus, of sorts. On the go. (I play mainly on my Skylake i7-6700K quad core Intel PC w/ GTX-1080Ti, which destroys the M2 Air or my work issue MBP M1 Pro, as a gaming machine (as No Man's Sky is GPU bound)).

bp

 

[darngooddesign]

It may not matter too much if you are willing to accept slightly lower settings.

 

[blakespot]

So, comparing my Shadow of the Tomb Raider (which is an Intel game running under Rosetta 2, but it does utilize Metal) results with those of GregsGadgets, we have:

M2 MacBook Air - 10 core GPU @ 1440x960 resolution
Frames rendered: 6590
Average FPS: 42

M2 MacBook Air - 8 core GPU @ 1440x960 resolution
Frames rendered: 6629
Average FPS: 42

M1 MacBook Air - 8 core GPU @ 1440x900 resolution
Frames rendered: 5308
Average FPS: 33


bp

 

[blakespot]

I'm at work so I can't watch the video. What was the frame rate on SotTR on the M2 8 and 10 Core MacBook Air?

Please disregard (maybe delete?) my quote -- I edited my results post because I was initially running the M1 test at a higher res than the M2s he was testing (he runs at 1440x960 and I was running at 1920x1080 -- I swapped down to closest to his, 1440x900 and updated results).

 

[jdb8167]

So, comparing my Shadow of the Tomb Raider (which is an Intel game running under Rosetta 2, but it does utilize Metal) results with those of GregsGadgets, we have:

M2 MacBook Air - 10 core GPU @ 1440x960 resolution
Frames rendered: 6590
Average FPS: 42

M2 MacBook Air - 8 core GPU @ 1440x960 resolution
Frames rendered: 6629
Average FPS: 42

M1 MacBook Air - 8 core GPU @ 1440x900 resolution
Frames rendered: 5308
Average FPS: 33


bp

First question. If you are trying to test thermals, why would he only run the test for ~2.5 minutes? You would want both the 8 core and 10 core to be heat soaked before drawing any conclusions about GPU throttling. Run the test for 30 minutes and then compare results. Seems like an odd test. (Maybe with an agenda?)

Second question. If they both got 42 fps, why is it reported that the 8 core was faster? Slightly more total frames? They are close enough (~0.6%) that I would call it even or within the margin of error.

 

[blakespot]

First question. If you are trying to test thermals, why would only run the test for ~2.5 minutes? You would want both the 8 core and 10 core to be heat soaked before drawing any conclusions about GPU throttling. Run the test for 30 minutes and then compare results. Seems like an odd test. (Maybe with an agenda?)

Second question. If they both got 42 fps, why is it reported that the 8 core was faster? Slightly more total frames? They are close enough (~0.6%) that I would call it even or within the margin of error.

If you watch the linked video, you see that the 10-core unit starts out on top and then after about a minute it throttles such that it's FPS (indicated on screen as you watch the game) is lower. So it evened out in the end, but if the demo went on for any longer, the numbers would have drifted farther apart seems the conclusion I draw.

bp

 

[darngooddesign]

What I noticed from the video is that he loves saying "controversy". He must be jealous of MaxTech uncovering their SSD controversy and wants one of his own.

I don't think the numbers wouldn't have drifted further apart. The computer already stabilized at a higher temp so the performance at that higher temp would not have changed. The fps at the end of that benchmark would be the same as the fps at benchmark twice as long.

 

[jdb8167]

If you watch the linked video, you see that the 10-core unit starts out on top and then after about a minute it throttles such that it's FPS (indicated on screen as you watch the game) is lower. So it evened out in the end, but if the demo went on for any longer, the numbers would have drifted farther apart seems the conclusion I draw.

bp

But isn't that the behavior you would expect? The higher power draw 10 core does better at first but throttles earlier. Over time, the 8 core will throttle more too. It just takes a little longer. Eventually, the 10 core will probably remain slightly faster. Stopping the test because you want to show the 8 core is faster briefly (actually the same) is just cherry picking.

Thanks for the info, if I remember, I'll watch the video when I'm not at work.

 

[Gelam]

I agree. I have repeatedly reached out to the guy that did the original video, but he hasn't responded and doesn't seem to be posting further videos, which seems odd. He has a hot topic and two such machines differing in GPU count. I am just hoping he posts further videos.

Surely some other tester out there must have two such machines. I haven't found them, though.

bp

Be on the lookout for notebookcheck's review. They have done so for the M2 8 core GPU version and will be posting the review for 10 core GPU version "soon". Their reviews have several games included.

Interestingly, the 8 core GPU M2 Air is slower than the 8 and 7 core GPU M1 Air in gaming according to their reviews. Which is very odd so I assumed that it is because they tested the M2 Air during summer while the M1 Air was tested in winter (assuming they did not control the test room temperature).