So M1 Max with 32 Core GPU matches RTX 3080 Mobile.

[crazy dave]

I don’t think anandtech does GPU performance reviews, they do talk about architecture though.

They stopped because something happened to Ryan Smith their head editor and main GPU reviewer but they’re going to start again soon. He’s okay, I don’t know what happened - I read some stuff on the Anandtech forums but that was as far as I can tell only speculation, so I won’t spread it here only to say that, even in the speculation, Ryan himself was always okay. But regardless of the why, he says that he’s ready to start again. He actually did the GPU review here on the M1 Max.

 

[diamond.g]

They stopped because something happened to Ryan Smith their head editor and main GPU reviewer but they’re going to start again soon. He’s okay, I don’t know what happened - I read some stuff on the Anandtech forums but that was as far as I can tell only speculation, so I won’t spread it here only to say that, even in the speculation, Ryan himself was always okay. But regardless of the why, he says that he’s ready to start again. He actually did the GPU review here on the M1 Max.

Ah. I hope they get back into the groove of it. To be honest I am a fan of Gamers Nexus GPU reviews.

 

[crazy dave]

Cool. So his render probably would see performance increases from utilizing the hardware bvh units.

It should yeah. Though I’m not near enough to that field to comment on the differences between CPU and GPU renderers and the quality/speed of each. I also don’t know how such a hybrid chip as the M1 might be utilized differently. Some of the others here probably know more. Still though, looking forwards to his post about CUDA and Metal.

 

[diamond.g]

It should yeah. Though I’m not near enough to that field to comment on the differences between CPU and GPU renderers and the quality/speed of each. I also don’t know how such a hybrid chip as the M1 might be utilized differently. Some of the others here probably know more. Still though, looking forwards to his post about CUDA and Metal.

Me too.

 

[jeanlain]

specint/fp is like using MS-DOS snippets to benchmark in 2021.

If you say so...

 

[JMacHack]

Ah. I hope they get back into the groove of it. To be honest I am a fan of Gamers Nexus GPU reviews.

GN is the gold standard of PC parts reviews.

 

[mr_roboto]

If you say so...

It's hilarious that he thinks Phoronix, of all things, is a much better site to get data from than Anandtech.

Anandtech: Actually cares about benchmarking methodology and consistency.

Phoronix: If you want quantity but incredibly crap quality, it's great. You can't ever count on the numbers meaning a damn thing, but there sure are a lot of them to choose from.

Actually it makes a lot of sense he likes Phoronix - he can pull anything he likes out of the meaningless pile of poorly run, poorly controlled "tests".

 

[ElfinHilon]

Got mine. 32GB of RAM and 32 core GPU. My geekbench scores are around 58-59K compared to the 68K we saw with 64GB of ram. Very strange tbh.

 

[altaic]

Got mine. 32GB of RAM and 32 core GPU. My geekbench scores are around 58-59K compared to the 68K we saw with 64GB of ram. Very strange tbh.

Spotlight indexing? Also, I noticed a weird thing on my second boot— control center was using 8+ GB of ram (not virtual, which was 399+ GB…?), and after a reboot it stayed around 25 MB real memory usage which is a “bit” more reasonable. I think there are a few bugs in Monterey that need to be worked out.

 

[leman]

I’m using MetalKit which handles all the CVDisplayLink stuff automatically. My problems went away when I set the MetalKit view's preferredFramesPerSecond to 120, which then locked it to 60hz. Presumably there is some kind of double-waiting going on. I'm not doing anything too special, so I'm definitely blaming MetalKit

If you want to measure performance you have to render to an offscreen texture anyway. Display links will throttle your rendering.

 

[jmho]

If you want to measure performance you have to render to an offscreen texture anyway. Display links will throttle your rendering.

True, but it's an important part of real-world performance.

There are some incredibly weird things going on with this machine. I'm getting massive amounts of stutter and after multi-threading my particles it flops between 1000% CPU usage (i.e. all 10 cores going full out) @ 120fps and then drops down to 250% and starts dropping to sub 40 fps with huge stutters before going back up to 120 fps for a bit.

Previously, I've had both the M1 Air and the M1 iPad Pros sitting at 800% for 10+ minutes with no problems (both getting incredibly warm), so I have no idea what's going on with the M1 Max.

That's not to say that I'm not doing something wrong (or maybe there is something wrong with this machine), but it's disappointing that it's apparently so easy to tank your frame rate on the M1 Max, especially when the same code runs perfectly on an M1 Air, and all my boilerplate code is basically just Apple sample code.

 

[leman]

True, but it's an important part of real-world performance.

There are some incredibly weird things going on with this machine. I'm getting massive amounts of stutter and after multi-threading my particles it flops between 1000% CPU usage (i.e. all 10 cores going full out) @ 120fps and then drops down to 250% and starts dropping to sub 40 fps with huge stutters before going back up to 120 fps for a bit.

Previously, I've had both the M1 Air and the M1 iPad Pros sitting at 800% for 10+ minutes with no problems (both getting incredibly warm), so I have no idea what's going on with the M1 Max.

That's not to say that I'm not doing something wrong (or maybe there is something wrong with this machine), but it's disappointing that it's apparently so easy to tank your frame rate on the M1 Max, especially when the same code runs perfectly on an M1 Air, and all my boilerplate code is basically just Apple sample code.

Yeah it does sound weird. Would you be so kind to submit a bug report to Apple? Maybe there is something they have to fix on their end...

 

[Macintosh IIcx]

True, but it's an important part of real-world performance.

There are some incredibly weird things going on with this machine. I'm getting massive amounts of stutter and after multi-threading my particles it flops between 1000% CPU usage (i.e. all 10 cores going full out) @ 120fps and then drops down to 250% and starts dropping to sub 40 fps with huge stutters before going back up to 120 fps for a bit.

Previously, I've had both the M1 Air and the M1 iPad Pros sitting at 800% for 10+ minutes with no problems (both getting incredibly warm), so I have no idea what's going on with the M1 Max.

That's not to say that I'm not doing something wrong (or maybe there is something wrong with this machine), but it's disappointing that it's apparently so easy to tank your frame rate on the M1 Max, especially when the same code runs perfectly on an M1 Air, and all my boilerplate code is basically just Apple sample code.

Could you try at set your display refresh rate to 60hz manually. Not sure if that matters, but it looks like ProMotion is buggy. Just as a troubleshooting step of course.

 

[altaic]

True, but it's an important part of real-world performance.

There are some incredibly weird things going on with this machine. I'm getting massive amounts of stutter and after multi-threading my particles it flops between 1000% CPU usage (i.e. all 10 cores going full out) @ 120fps and then drops down to 250% and starts dropping to sub 40 fps with huge stutters before going back up to 120 fps for a bit.

Previously, I've had both the M1 Air and the M1 iPad Pros sitting at 800% for 10+ minutes with no problems (both getting incredibly warm), so I have no idea what's going on with the M1 Max.

That's not to say that I'm not doing something wrong (or maybe there is something wrong with this machine), but it's disappointing that it's apparently so easy to tank your frame rate on the M1 Max, especially when the same code runs perfectly on an M1 Air, and all my boilerplate code is basically just Apple sample code.

Please toss your code up on GitHub and give a link. That’d help me figure out what’s going on. We don’t have to guess any longer.

 

[jmho]

I've fixed it now. It was to do with the threading. I got the code for using DispatchQueue.concurrentPerform from a popular iOS development tutorial site which was dangerously incorrect.

Now I'm getting the kind of performance I'd expect from this machine. Single threaded I was losing frames at 200k particles, now I had to have 5 million to get it to drop from 120 fps, which is pretty incredible.

I still think ProMotion is mildly janky, but I'm relieved to find out that the biggest jank was my code

 

[Boil]

I've fixed it now. It was to do with the threading. I got the code for using DispatchQueue.concurrentPerform from a popular iOS development tutorial site which was dangerously incorrect.

Now I'm getting the kind of performance I'd expect from this machine. Single threaded I was losing frames at 200k particles, now I had to have 5 million to get it to drop from 120 fps, which is pretty incredible.

I still think ProMotion is mildly janky, but I'm relieved to find out that the biggest jank was my code

See what happens when you play with strange code off the interwebz...? ;^p

 

[ElfinHilon]

Spotlight indexing? Also, I noticed a weird thing on my second boot— control center was using 8+ GB of ram (not virtual, which was 399+ GB…?), and after a reboot it stayed around 25 MB real memory usage which is a “bit” more reasonable. I think there are a few bugs in Monterey that need to be worked out.

Didn't even think about that. I'll try again later. Cheers.

 

[mi7chy]

GN is the gold standard of PC parts reviews.

GN is too long winded and too much hair flipping. Prefer Hardware Unboxed which is more to the point.

 

[ElfinHilon]

So after nearly 40 pages, I think we might finally be done discussing this.

 

[Taz Mangus]

How about start discussing real world usage.

 

[Frankied22]

How about start discussing real world usage.

That is insane. These laptops and their chips really are a leap ahead. They are blowing away everything else in everything EXCEPT gaming. And even in gaming, they are holding up very well when you consider most games are running through Rosetta 2 or windows virtualization layers. I have the 16" M1 Max with 32 core GPU and it is a beast of a machine. I am still running it through several different games but for everything else it blows away my custom desktop pc.

I just compared shadow of the tomb raider on my M1 Max and my pc with a 1080ti and the Mac runs the game with more FPS at a higher settings than my pc. Insanity.

 

[Erasmus]

That is insane. These laptops and their chips really are a leap ahead. They are blowing away everything else in everything EXCEPT gaming. And even in gaming, they are holding up very well when you consider most games are running through Rosetta 2 or windows virtualization layers. I have the 16" M1 Max with 32 core GPU and it is a beast of a machine. I am still running it through several different games but for everything else it blows away my custom desktop pc.

I just compared shadow of the tomb raider on my M1 Max and my pc with a 1080ti and the Mac runs the game with more FPS at a higher settings than my pc. Insanity.

Could you provide more info? What other games have you tried, what are you using to test them (Rosetta, Crossover, Parallels, etc.), what frame rates are you getting?

I've also got a 1080ti in a desktop rig, and I'm likely going to get a 16" M1 Max (replacing my current base model 15" 2018 MBP) as my secondary computer. Certainly don't mean it to replace my desktop, but I'm often away from home, and it would be great to confirm whether it can function well for games.

 

[Frankied22]

Could you provide more info? What other games have you tried, what are you using to test them (Rosetta, Crossover, Parallels, etc.), what frame rates are you getting?

I've also got a 1080ti in a desktop rig, and I'm likely going to get a 16" M1 Max (replacing my current base model 15" 2018 MBP) as my secondary computer. Certainly don't mean it to replace my desktop, but I'm often away from home, and it would be great to confirm whether it can function well for games.

I have just started with the game testing but Shadow of the Tomb Raider I did the exact same settings on both machines (1440p, mix of High and Normal quality, bloom and motion blur turned off) and the Mac was getting about 15 more fps than my desktop. If I bump the quality up to all High on my Mac it gets about the same FPS as the pc with those mixed settings. I did some video rendering with Da Vinci and Affinity Photo stuff too and the Mac crushed my Ryzen

 

[Taz Mangus]

I have just started with the game testing but Shadow of the Tomb Raider I did the exact same settings on both machines (1440p, mix of High and Normal quality, bloom and motion blur turned off) and the Mac was getting about 15 more fps than my desktop. If I bump the quality up to all High on my Mac it gets about the same FPS as the pc with those mixed settings. I did some video rendering with Da Vinci and Affinity Photo stuff too and the Mac crushed my Ryzen

Apple has just turned the whole PC market on its ear. It has proved that you can do both performance and energy efficiency at the sam time. A question for you. When you were doing video rendering with Da Vinci with both systems, did you hear fan noise on both systems and how loud?

By the way, congrats on your new MacBook Pro M1 Max.

 

[Taz Mangus]

That is insane. These laptops and their chips really are a leap ahead. They are blowing away everything else in everything EXCEPT gaming. And even in gaming, they are holding up very well when you consider most games are running through Rosetta 2 or windows virtualization layers. I have the 16" M1 Max with 32 core GPU and it is a beast of a machine. I am still running it through several different games but for everything else it blows away my custom desktop pc.

I just compared shadow of the tomb raider on my M1 Max and my pc with a 1080ti and the Mac runs the game with more FPS at a higher settings than my pc. Insanity.

And that is probably running through Rosetta.