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

MayaUser

macrumors 68040
Nov 22, 2021
3,177
7,196
3x70=210W<270W and thats just from Gpu? can we find for the cpu to combine the whole package ?
 

theorist9

macrumors 68040
May 28, 2015
3,880
3,060
It seems that the M1 Ultra is more efficient than the A6000.
That was my expectation. So it seems my contention holds:
Assuming single-precision (FP 32) TFLOPS are a good meansure of general GPU compute performance, I'd say get a Mac if your primary concern is GPU performance/watt, or go PC-NVIDIA if your primary concern is GPU performance/$.
...with the added caveat that if you want to maximize GPU performance/$, get the GeForce RTX4000 series instead of the Quadro A series.
 

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
That was my expectation. So it seems my contention holds:
As performance depends on the scene, it is difficult to know which one is more efficient.

For example, Max Tech found that the M1 Ultra (1:06) took 4 times longer to render than the RTX 3090 (0:15), so in that scene the A6000 would be more efficient.
 

gradi

macrumors 6502
Feb 20, 2022
285
156
Ah, by the way, something about performance of Apple GPUs. This is 3dmark score for the iPhone 14 Pro: https://www.notebookcheck.net/Apple-A16-GPU-5-Core-GPU-Benchmarks-and-Specs.652746.0.html

It’s essentially approaching Nvidia’s dedicated MX450. That’s a phone we are talking about.

I see that according to this chart for 3Dmark the MX450 GPU is just 6% faster than the A16 GPU. And the M1 8-core GPU is 46% faster, M2 8-core GPU is 70% faster. Interesting info about the iphone GPU.

Screen Shot 2022-09-17 at 18.42.44.jpg
 

theorist9

macrumors 68040
May 28, 2015
3,880
3,060
As performance depends on the scene, it is difficult to know which one is more efficient.

For example, Max Tech found that the M1 Ultra (1:06) took 4 times longer to render than the RTX 3090 (0:15), so in that scene the A6000 would be more efficient.
But was this merely an exception? What matters is which is more efficient on average, since that's what determines your power bill.

What we can say for the Quadro A-series (and the GEForce RTX series) vs. the M-series is that, if you are doing ML then, for that category of tasks, the NVIDIA GPU's are more efficient.
 
  • Like
Reactions: Xiao_Xi
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.