Will be interesting to see the battery life difference between the 30 and 38 cores GPU
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M2 Max 30-core Benchmarks
- Thread starter chrismu
- Start date
- Sort by reaction score
I have it. Still pretty preliminary and I don't generally run Geekbench but I did run 3dmark Wildlife Unlimited. It did 112fps compared to my 16" M1 Max 32 core 120fps, 14" M1 Pro 16 core 61.6fps and my Mac Studio M1 Ultra 48 core 172.5fps. Surprised me as I figured there would be some improvements per core so would be faster than my 16" but it doesn't seem to be. I ran this because it is slower in my AI Video upscaling tests too compare to my 16" M1 Max. However, the 14" M2 Max 30 temps are a bit lower than my 14" M1 Pro when doing the upscaling work so maybe the 14" won't throttle like the M1 Max variant did. I have a 38 core version coming next week. I'm expecting that one will see some improvements over my M1 Max 32 core but my expectations aren't as high as they were this morning before I ran some tests.
Edit: Just finished GFX bench. Same thing. The 16" M1 Max 32 core is running faster than the 14" M2 Max 30 core.
Last edited:
Its running cooler than my M1 Pro 16 core with the fan speeds maxed out while the M2 Max 30 core is running at 50%. Seemed to have beefed up the cooling.
A slight shame if the 30 core doesn't equal or beat 32 core m1, as the 30% claims would imply 20% due to extra cores and rest from improvements.
But assuming that thermals/battery are noticeably better, then I'll be happy.
But assuming that thermals/battery are noticeably better, then I'll be happy.
Like where? I am mainly looking for Final Cut and XCode performance, don’t really like their „style“ of videos either but I figured the test would be interesting.
I dunno there's plenty of thermal headroom. Here's screen shots maxing out the CPUs/GPUs with Luxmark compared to my 14" M1 Pro 16 core. Fans are max on both.
Attachments
That looks promising, 30 core 16" hopefully will keep chugging without fans needing to be raised too high then.
Yeah it should. Even with the 14" the system I was running AI upscaling for a few hours and the fans were sitting at about 2,400rpm and they were running at 6,500 on my 14" M1 Pro. They're only maxed out in my screen shot because I was showing there should be enough thermal head room for the maxed out version this time.
Good, that's one of the biggest desires for upgrading, more performance without the constant fan noise if doing anything that pushes the intel.
Hmmm looking at the 30 core model, its metal score on Geekbench is roughly around the same as my iMac with 5700XT. I do understand that the metal Geekbench score doesn’t push the apple GPU to its full potential however. I still think I’m willing to risk the 38 core Max in a 14” if they have beefed the cooling.
Very nice test of a 30 core 14 inch vs 32 core M1 Max 16 inch:
İf it's possible could you run Blender and render out classroom scene on your 14'' MBP MAX 30 GPU. I watched a benchmark video and Blender render times for base 14'' M2 MBPro looks very fast. On a YT video 14'' M2 MBPro 16 core GPU classroom scene 2:11, I render out same scene with and 48 core M1 Ultra 00:55. This is a huge improvement on GPU side for M2. There is another 14'' MBPro model with 30 GPU, with a perfect scaling if anyone could render out same scene it has to be rendered out more or less 1:10. I m not interested to buy any M2 MBP but I am considering upgrade my base Ultra to top spec Ultra, working with C4D and Redshift, ZBrush, basic 4/6 k video editing and cgi compositing most of the time. I m still using Base Ultra and a PC side by side everyday and I m trying to figured out is it worth to wait Mac Studio M2 Ultra or upgrade to full spec Ultra from base Ultra.
Yes would be good to know.
If someone can rund the blender benchmark on a 30 core would be nice too.
Blender - Open Data
Blender Open Data is a platform to collect, display and query the results of hardware and software performance tests - provided by the public.
Sorry, I ended up returning mine yesterday. I probably should have held onto it for a bit longer since I had 45 days for a return but it wasn't hitting the marks I needed it to in my work flow in order to keep it so opted to return it sooner rather than later. If someone else doesn't run the test for you before the end of next week I can run it on the 38 core model I have coming at the end of the week. We should be able to get a good idea of how the 30 core would perform based on that. Although, do we know if the GPU cores in the 14" clocked lower compared to the 16" too? If they are the 16" 30 core would be a better test though I suspect that setup would be pretty rare.
Thank you very much, I would like to see 38 core GPU actually, it could be better to compare with 48 core Ultra. as I wrote before I suspect about M2 GPU scaling and heat also. Your 38 core numbers will give us better idea about M2.
38 core beats the M1 Ultra 64 cores in Cycles rendering by 2%.
If my estimate is correct the 30 core might be under the 48 core Ultra however but well over the 32 core M1.
I am actually do not use Blender often. I try to get an estimation which I can translate to the Redshift renderer on C4D. Currently M1 Ultra works well with Redshift. But I could not say this for M1 Studio Max which I have used for 3 months and it did not work well at all. So same as here 16'' M2 MBP will not be ok with serious 3D workflows either. As long as MAX chip is not capable on M1 Studio it will not work on M2 MBP also. Rendering is not a big part of 3D works, one way or another required tasks for 3D crippled down by Max chip, this is what I experienced. Thanks again
Hope to see M2 38 core GPU scores.Have you seen this, however if I remember right from the M1 generation should he have not used a bigger bucket size.I am actually do not use Blender often. I try to get an estimation which I can translate to the Redshift renderer on C4D. Currently M1 Ultra works well with Redshift. But I could not say this for M1 Studio Max which I have used for 3 months and it did not work well at all. So same as here 16'' M2 MBP will not be ok with serious 3D workflows either. As long as MAX chip is not capable on M1 Studio it will not work on M2 MBP also. Rendering is not a big part of 3D works, one way or another required tasks for 3D crippled down by Max chip, this is what I experienced. Thanks again
Yes, I still read all this topic on Redshift forum everyday, this render time could happens by 32 GB Unified RAM which is not enough for this kind of benchmark and 2nd and most interesting part is M2 GPU is not Scale well as same on M1 GPU. On same forum you can find base model M1 Ultra benchmarks, if I remember it was slightly above 7 minutes. And yes anyone should go for 512 bucket size with 64 GB Unified AS Memory, it is faster. I have used Studio Max 32 GB ram and it was NOT OK for Redshift maybe it is fine for Blender where Apple puts a lot of effort last year. However Ultra is different kind of animal and base model Ultra is more than enough for any 3D workflow, this is what I experienced and still have today.
Also you can do the math easily:
( looks like Blender improvements do not works on Redshift too )
30 Core Max M2 GPU ends up 09:37
48 cores Ultra M1 GPU 07:17
the gap between this machines is 02:20 seconds.
M2 max GPU cores works x2,11 more or fast than M1 Ultra GPU cores, but end result shows M2 GPU either not scales as well as same with M1 GPU's or 32 GB Max chip is under memory pressure.
If there was %30 GPU improvement as Apple claims, this benchmark had to be ends neck to neck or there should be very little time gap.
))) Maybe all this math is wrong. I am really bad on doing this kind of math.
I thought the M2 max was running at 3.68ghz?
I have the 16" m2 with 30 cores and it only reading 3.49ghz. I that was only on the 14" or the M2 pro?
I have the 16" m2 with 30 cores and it only reading 3.49ghz. I that was only on the 14" or the M2 pro?
It definitely appears theres some factor effecting what top speed goes to at, browsing geekbench results I can see 38 core max's reporting both 3.5 and 3.7
Apple didn't say each gpu core is 30% better, just that gpu performance is 30% better, there's 25% more cores, so most of that 30% comes just from extra cores.
Moral of the story and same as it always is, if you have the last generation, pretty much not worth upgrading. If you had an older generation on the other hand…