Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
Interestingly I’m finding very little difference compiling projects in Xcode between my M1 Ultra and my M1 Mac Mini. Current project has the ultra compiling at 90 seconds, m1 mini at 97.

Compiling is one of those things where the specifics of a project can greatly affect the benefit of core count. Any place where your compile bottlenecks can create bubbles where the machine’s resources go under-utilized. So this isn’t too surprising for me.

At least on a team I used to be on, we were large enough that we would get a hold of hardware for testing to see what offered the best perf/dollar for our builds to drive upgrade decisions. Despite the size of the codebase, the end results were often surprising, and diminishing returns was pretty common.
 

Rickroller

macrumors regular
May 21, 2021
114
45
Melbourne, Australia
It seems M1 Ultra is using only one M1 Max GPU instead of 2 so far.
I think you might be right…maybe not for everything, but a lot of the testing hasn’t pegged the GPU past 60w…which also happens to be what you would expect from a 32 core MAX model. I guess we have to wait for software updates to really figure out what the true potential of the Ultra is. Hopefully the next version of Geekbench makes the changes needed to push Apple GPUs to their full power.
 

Kpjoslee

macrumors 6502
Sep 11, 2007
417
269
It seems M1 Ultra is using only one M1 Max GPU instead of 2 so far.

I think it is simply being underutilized. Software just sees M1 Ultra GPU as one GPU so I don't think it is possible to make it utilize just one M1 Max GPU.
 

powerslave65

macrumors 6502
Mar 21, 2011
394
211
Sherman Oaks CA
I don't have a Geekbench license but tested my Mac Studio Ultra base model that I picked up Friday and got this score. Premiere is really snappy on this bad boy.
 

Attachments

  • Screen Shot 2022-03-18 at 5.03.54 PM.png
    Screen Shot 2022-03-18 at 5.03.54 PM.png
    60.7 KB · Views: 75
  • Like
Reactions: Pressure

Malus120

macrumors 6502a
Jun 28, 2002
696
1,456
There's a lot to unpack both here and all over the internet surrounding Mac Studio performance. I don't have, nor do I plan on purchasing a Mac Studio (perfectly happy with my M1 Max 14" MBP,) but i do want to share a few thoughts regarding performance and optimization.

1. Apple's repeatedly made bold claims regarding Apple Silicon performance, comparing the Max GPU to the 3080M and the Ultra to the 3090. Naturally, everyone is thus trying to "verify" these claims but this is actually very difficult because:
2. We're still VERY early in the Apple Silicon transition, a transition not only to a new CPU architecture, but ALSO to a bespoke GPU architecture and unified memory subsystem all with their own unique strengths, weakness, and quirks.
(It also differs in that x86 was already, at the time of the PPC -> X86 switch, the dominant platform around which developers were building/optimizing apps for, whereas ARM64, while prominent in mobile is rare on the desktop )
3. While "optimization" for Apple Silicon is often boiled down to the existence of a native app, the reality is a native app is just a starting point, and doesn't necessarily mean an app has been "optimally optimized" for AS.
(Side note, I don't think anyone should be taking Rosetta 2 benchmarks, even "GPU limited" ones as any indication of what AS can do, without a HUGE grain of salt)
4. Case in a point, Capture One Pro (primary competitor to Adobe Lightroom) released an Apple Silicon App in May of last year (2021) It performed well, with the M1 Max comparing favorably to i9/3080M laptops, matching Apple's claims*. One might assume it was already "well optimized" for AS. However an upcoming version is going to improve the performance of many operations on Apple Silicon by as much as 2-4x. That is a massive improvement to what was already impressive performance in line with Apple's claims, and shows just how much was being left on the table even with a well regarded native app.

All of which is to say that, as much as everyone would like a clear cut answer on the M1 Ultra's performance, particularly GPU performance vs the 3080/3090/6900XT, it's simply far too early to make definitive claims. I think it's safe to say Apple's claims show where their aspirations were/are for performance in the kinds of workloads Apple cares about. Does that mean it's as fast as those GPUs at everything? No. But I do think it is fair to say the performance we see today may only be the tip of the iceberg for many types of applications.

Does that make it harder to make purchasing decisions? Yes. If you're workload is massively underperforming vs an Nvidia powered PC (or a Mac Pro with an RX 6900 or a 6800 Duo) you may be better off buying one of those if you need something today. But for a lot of people I think it makes sense to sit on the fence for a while and see how things shake out.
 
Last edited:

Appletoni

Suspended
Mar 26, 2021
443
177
There's a lot to unpack both here and all over the internet surrounding Mac Studio performance. I don't have, nor do I plan on purchasing a Mac Studio (perfectly happy with my M1 Max 14" MBP,) but i do want to share a few thoughts regarding performance and optimization.

1. Apple's repeatedly made bold claims regarding Apple Silicon performance, comparing the Max GPU to the 3080M and the Ultra to the 3090. Naturally, everyone is thus trying to "verify" these claims but this is actually very difficult because:
2. We're still VERY early in the Apple Silicon transition, a transition not only to a new CPU architecture, but ALSO to a bespoke GPU architecture and unified memory subsystem all with their own unique strengths, weakness, and quirks.
(It also differs in that x86 was already, at the time of the PPC -> X86 switch, the dominant platform around which developers were building/optimizing apps for, whereas ARM64, while prominent in mobile is rare on the desktop )
3. While "optimization" for Apple Silicon is often boiled down to the existence of a native app, the reality is a native app is just a starting point, and doesn't necessarily mean an app has been "optimally optimized" for AS.
(Side note, I don't think anyone should be taking Rosetta 2 benchmarks, even "GPU limited" ones as any indication of what AS can do, without a HUGE grain of salt)
4. Case in a point, Capture One Pro (primary competitor to Adobe Lightroom) released an Apple Silicon App in May of last year (2021) It performed well, with the M1 Max comparing favorably to i9/3080M laptops, matching Apple's claims*. One might assume it was already "well optimized" for AS. However an upcoming version is going to improve the performance of many operations on Apple Silicon by as much as 2-4x. That is a massive improvement to what was already impressive performance in line with Apple's claims, and shows just how much was being left on the table even with a well regarded native app.

All of which is to say that, as much as everyone would like a clear cut answer on the M1 Ultra's performance, particularly GPU performance vs the 3080/3090/6900XT, it's simply far too early to make definitive claims. I think it's safe to say Apple's claims show where their aspirations were/are for performance in the kinds of workloads Apple cares about. Does that mean it's as fast as those GPUs at everything? No. But I do think it is fair to say the performance we see today may only be the tip of the iceberg for many types of applications.

Does that make it harder to make purchasing decisions? Yes. If you're workload is massively underperforming vs an Nvidia powered PC (or a Mac Pro with an RX 6900 or a 6800 Duo) you may be better off buying one of those if you need something today. But for a lot of people I think it makes sense to sit on the fence for a while and see how things shake out.
Yes a lot of claims from Apple.
Looking at other benchmarks, like chess, you can see that Apple M1 Ultra CPU is as slow as a turtle.
 

Malus120

macrumors 6502a
Jun 28, 2002
696
1,456
Yes a lot of claims from Apple.
Looking at other benchmarks, like chess, you can see that Apple M1 Ultra CPU is as slow as a turtle.
Ah yes the Stockfish guy! Should've known you were here somewhere.
Guess I better sell my MBP because one obscure chess benchmark, Stockfish is faster on x86! Whatever shall I do! Probably disregard the fact it runs circles around x86 based systems in software I (and most other normal people on this forum) actually use, while maintaining god-tier battery life ?

Seriously man, we get that Stockfish could be faster. Maybe go bug the developers about that? Or perhaps you'd like to show some real world usage (you know something other than... Chess) where Apple Silicon struggles?
Given that Apple Silicon turns out great results in both synthetic benchmarks (SPEC, Cinebench, GeekBench) and many real workloads from companies like Apple, Adobe, PhaseOne, etc I think we can firmly say it's an outlier.

Not saying M1/Apple Silicon is the best for everything but saying it's "slow as a turtle" is just ignorant.
 

Malus120

macrumors 6502a
Jun 28, 2002
696
1,456
Completely off topic but while copying files from my M1 Max MBP to get ready for the Studio it ran 24 hours copying files and still had 53% battery life. That is some Star Trek atomic level **** right there!
Hot damn ?
That’s insane, and also not a irrelevant at all, the efficiency and battery life of Apple Silicon is as much of part of the performance profile as number crunching speed.

I’m assuming you’re on the 16 inch, (as good as the 14 inch is I don’t think the battery life is quite that good,) but I’m curious was that with a stock config or did you modify the system software to be more efficient?

I remember the first time I did photo import, Preview generation and basic processing on my old M1 MacBook Air… I think that shoot was close to 1000 photos and it didn’t even crack 10%.
 

powerslave65

macrumors 6502
Mar 21, 2011
394
211
Sherman Oaks CA
Hot damn ?
That’s insane, and also not a irrelevant at all, the efficiency and battery life of Apple Silicon is as much of part of the performance profile as number crunching speed.

I’m assuming you’re on the 16 inch, (as good as the 14 inch is I don’t think the battery life is quite that good,) but I’m curious was that with a stock config or did you modify the system software to be more efficient?

I remember the first time I did photo import, Preview generation and basic processing on my old M1 MacBook Air… I think that shoot was close to 1000 photos and it didn’t even crack 10%.
Yes the 16 MBP and the only thing special was my screen was set to turn off quickly maybe that helped. But still! I have never had anything that good.
 

dieselm

macrumors regular
Jun 9, 2009
195
125
I remember the first time I did photo import, Preview generation and basic processing on my old M1 MacBook Air… I think that shoot was close to 1000 photos and it didn’t even crack 10%.
I remember when I got my MBA and did a reinstall of the whole system and set of apps on the battery. Dead silent, cool, and the battery was down like 8%. Previously, I'd always plug-in for things like that and the fans would go crazy.
 

JimmyjamesEU

Suspended
Jun 28, 2018
397
426
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.