MacBook Pro's M1 Max GPU is Over 3x Faster Than M1 in First Metal Benchmark

[hefeglass]

68k is the Metal score the others are OpenCL scores which of course are a bit lower. Pro scores will be worse!

you REALLY hope they look bad ..although bad is still lightyears beyond anything else out there. You are hoping these perform worse than a freaking gpu that runs on 90-100w..and is completely worthless when not plugged in. I have a laptop with a ryzen 5900hs and a rtx 3070..my m1 macbook pro already does blender simulations faster than the best AMD has to offer..soon with metal backend to take advantage of the gpu, it will run circles around my 3070 rendering with cycles, and will be able to do it unplugged..without raging the fan..without running its giant battery out in 30 minutes while unplugged and underpowered. These laptops processors blow anything intel/amd can currently offer and somehow you hate them. Its funny to watch you constantly tell us how BAD these chips are.

 

[hefeglass]

It's not rocket science. Apple claims M1 Max iGPU is 4x faster than M1 and there are a lot of data already out there for M1 so we can deduce the performance. If M1 takes 37.21 mins then 4x faster puts the M1 Max at ~9.3 mins.

https://www.cgdirector.com/redshift-benchmark-results/

Apple's M1 Max iGPU 4x faster than M1 claim:

https://www.apple.com/newsroom/2021...the-most-powerful-chips-apple-has-ever-built/

why are you comparing it to a list of DESKTOP GPUs?

the 1080ti is about 15% faster than a laptop 3080
seems like the 9 minute assumption is accurate and inline with the GPUs they compared to in their presentation..

with all your posting in here telling us all how it is..youd think you knew the difference between desktop gpus and laptop gpus..

 

[priamXus]

Has anyone considered that per configuration two results should exist because of the new high performance mode ?

Edit: three if you polish it everyday with the iCloth.

 

[Natrium]

you REALLY hope they look bad ..although bad is still lightyears beyond anything else out there. You are hoping these perform worse than a freaking gpu that runs on 90-100w..and is completely worthless when not plugged in. I have a laptop with a ryzen 5900hs and a rtx 3070..my m1 macbook pro already does blender simulations faster than the best AMD has to offer..soon with metal backend to take advantage of the gpu, it will run circles around my 3070 rendering with cycles, and will be able to do it unplugged..without raging the fan..without running its giant battery out in 30 minutes while unplugged and underpowered. These laptops processors blow anything intel/amd can currently offer and somehow you hate them. Its funny to watch you constantly tell us how BAD these chips are.

I never said I hate these chips or that they’re bad. If that’s what it looks like, let me clarify that I’m an Apple fan, own many Apple devices including one with the M1 chip, and I really love my devices and these new M chips. I think Apple is doing great with the M chips, they elevated all Macs to a new level and are simply put an extraordinary achievement.

 

[blazerunner]

What will be interesting is gaming performance in a Windows VM. The base M1 was pretty much useless for this because you could only allocate 3 high performance cpu cores (most games struggle with less than 4) and 4-8gb of ram to the VM. Combining that with GTX1050 equivalent 7-8 gpu cores make the entry M1 Macbooks a low end gaming machine in a VM... but 4 high performance cpu cores, 16-32gb of ram and 24-32 gpu cores? Now we're talking.

Yeah, but will it run at 60fps or more at native resolution? I still think we're a long way away from proper laptop gaming. I suppose hooking it up to a 1440p / 1080p monitor would help (if even doable?)

 

[priamXus]

Yeah, but will it run at 60fps or more at native resolution? I still think we're a long way away from proper laptop gaming. I suppose hooking it up to a 1440p / 1080p monitor would help (if even doable?)

Would be more than happy with stables 40-50fps at 1440p

 

[blazerunner]

Would be more than happy with stables 40-50fps at 1440p

Now I'm actually curious to see the benchmarks now actually..... never thought about gaming via VM on a MacBook... been planning on building a new hackintosh for this reason.

 

[priamXus]

Now I'm actually curious to see the benchmarks now actually..... never thought about gaming via VM on a MacBook... been planning on building a new hackintosh for this reason.

I mean cmon. It has been done since the M1 release. Think about this; the damn Air can run Overwatch via Parallels on Windows for Arm, virtualizing x86. (Low settings but playable)

 

[Spidder]

No, the latest games on highest settings will probably not run at 60fps in native resolution, but 60fps isn't necessary with adaptive refresh rate aka freesync aka g-sync, 40-50 will feel smooth and with dialed down settings, native resolution with good fps might be possible.

Go on Youtube and search for "M1 gaming", there are plenty of videos showing that you can run lots of AAA games with Crossover or in a VM, some even playable (at low res & low settings).

PS: And don't forget, you can game in HDR on the new MBP. Not some crappy HDR400 like lots of gaming monitors have, but real HDR!

 

[Drifter759]

Very impressive, no doubt, but I'm a little disappointed with the GPU performance. The OpenCL scores seem to be around 68k, while the 3080 Mobile (100W) seem to get 120k-130k.

The M1 Max beats my Thinkpad T15g Gen 2 (i9-11950H) in CPU by about 20% in both single core and multi, but my 3080 Mobile beats the M1 Max by almost 2x. I do a lot of 3D work, and I was thinking I might switch over, but this gives me pause.

The M1 Max completely outclasses it in battery life, obviously, but I was hoping it would be pretty close to the best of Windows-world in GPU power.

 

[sunny5]

It's not rocket science. Apple claims M1 Max iGPU is 4x faster than M1 and there are a lot of data already out there for M1 so we can deduce the performance. If M1 takes 37.21 mins then 4x faster puts the M1 Max at ~9.3 mins.

https://www.cgdirector.com/redshift-benchmark-results/

Apple's M1 Max iGPU 4x faster than M1 claim:

https://www.apple.com/newsroom/2021...the-most-powerful-chips-apple-has-ever-built/

We are keeping these results separate from the Nvidia / CUDA Results for now, as Redshift Metal isn’t as mature as Redshift CUDA yet, and the benchmark runs were done on eGPUs and/or beta macOS versions. Take these scores with a grain of salt. They’ll stabilize and improve over time.

Doesn't prove anything at all.

 

[hefeglass]

These benchmarks seem pretty accurate. Also seem to match the 3080 pretty evenly like apple said. (Notice how they are comparing to laptop gpus instead of the red chart that guy posted which is all desktop gpus) This is also using the best api for each test/gpu
seems about as fair a test as can be done

 

[hefeglass]

Doesn't prove anything at all.

He is comparing them to desktop gpus
Pointless comparison

 

[sunny5]

He is comparing them to desktop gpus
Pointless comparison

Not only that, the benchmark itself is the problem and they even mentioned that.

 

[DesertNomad]

Interesting. I have the exact same configuration and my fans a basically locked to full bore. One of the reasons I'm jumping on an M1 MAX now.

Weird. I have it (16" 2.4GHz i9, 64GB, 4TB, 5600M) hooked to an XDR Display and a 27" Thunderbolt and in a year the fans have come on once or twice. I use it mostly for software development in Xcode. It is warm above the '7' key, but not hot.

 

[-BigMac-]

How about a Mcclaren F1! New they cost under $900,000 and now one just went for $20 million so I would call that the opposite of your assertion.

Thats not what diminishing returns is, or what you were talking about.

 

[imbeee]

Compare benchmark Apple M1 Pro (8 Core) vs. Apple M1 Pro (10 Core) https://cpu-benchmark.org/compare/apple-m1-pro-8-core/apple-m1-pro-10-core/

 

[Natrium]

No, you are not right, those are the same results that have been there since yesterday. And like others have already pointed out, Geekbench is not reporting the number of compute units correctly - for example it reports the M1 with a 7-core GPU as 8 core GPU.

https://browser.geekbench.com/v5/compute/search?utf8=✓&q=MacBookPro18

Many more results now. All around the same scores. No way Apple handed out 24c versions to ALL reviewers. Means I was right and was attacked by you and others for stating my logical argument and conclusion.

 

[Serban55]

why are you comparing it to a list of DESKTOP GPUs?

the 1080ti is about 15% faster than a laptop 3080
seems like the 9 minute assumption is accurate and inline with the GPUs they compared to in their presentation..

with all your posting in here telling us all how it is..youd think you knew the difference between desktop gpus and laptop gpus..

outplayed on his own game

 

[Serban55]

Very impressive, no doubt, but I'm a little disappointed with the GPU performance. The OpenCL scores seem to be around 68k, while the 3080 Mobile (100W) seem to get 120k-130k.

The M1 Max beats my Thinkpad T15g Gen 2 (i9-11950H) in CPU by about 20% in both single core and multi, but my 3080 Mobile beats the M1 Max by almost 2x. I do a lot of 3D work, and I was thinking I might switch over, but this gives me pause.

The M1 Max completely outclasses it in battery life, obviously, but I was hoping it would be pretty close to the best of Windows-world in GPU power.

if you do 3d work than wait...these benchmark doest even come close to an actual 3d work...can't wait for some Maya and other cpu+gpu intensive tasks. In pro apps and situations that the gpu would eat even over 32 gb of vRam...these SoC will eats our windows mobile gpu that tops at 16gb vram

 

[PPietra]

https://browser.geekbench.com/v5/compute/search?utf8=✓&q=MacBookPro18

Many more results now. All around the same scores. No way Apple handed out 24c versions to ALL reviewers. Means I was right and was attacked by you and others for stating my logical argument and conclusion.

No it doesn’t mean a thing, you have no idea what are the machines being tested!
Furthermore the math doesn’t add up. You don’t go from linear scaling with 16 cores to 3/4 of the performance at 32, that is not how GPUs work, at least not when memory bandwidth, cache and theoretical TFlops also double.

 

[Serban55]

https://browser.geekbench.com/v5/compute/search?utf8=✓&q=MacBookPro18

Many more results now. All around the same scores. No way Apple handed out 24c versions to ALL reviewers. Means I was right and was attacked by you and others for stating my logical argument and conclusion.

like others said to you, that benchmark is buggy its showing a lot of false informations
so stop looking at that, and find something else..maybe another benchmark engine, or if you can find an comparasion between an dGpu and this M1 max running Maya

 

[Natrium]

get over it...like others said to you, that benchmark is buggy its showing a lot of false informations
so stop looking at that, and find something else..maybe another benchmark engine, or if you can find an comparasion between an dGpu and this M1 max running Maya

You get over it. Stop attacking me. I’m simply stating the obvious. Many scores now. Whatever core count these benchmarks state (32 btw) no way Apple provided reviewers with handicapped 24c versions. If you don’t agree that’s fine but stop downvoting me for stating my opinion based on reasonable assumptions and logic, while yours is based on nothing but “Geekbench is wrong”.

 

[Serban55]

You get over it. Stop attacking me. I’m simply stating the obvious. Many scores now. Whatever core count these benchmarks state (32 btw) no way Apple provided reviewers with handicapped 24c versions. If you don’t agree that’s fine but stop downvoting me for stating my opinion based on reasonable assumptions and logic, while yours is based on nothing but “Geekbench is wrong”.

whos attacking you? im saying stop misleading the users...until the reality is here

 

[StoneJack]

Now I'm actually curious to see the benchmarks now actually..... never thought about gaming via VM on a MacBook... been planning on building a new hackintosh for this reason.

I built a i9 10850 Hackintosh with 10 cores and RX580 GPU. Guess what, this notebook has faster Metal scores by 30%-50%.