Apple M2 Ultra SoC isn’t faster than AMD and Intel last year desktop CPUs

Post by solomatov » Sat, Jun 10, 2023, 7:06 PM UTC

sosborn wrote: post #95

solomatov wrote: post #25

Earlier quoted context omitted.

Not always, if you are connected to a source of electricity and doing something limited by speed performance is critical. For me, Desktop use is almost perfect on Apple due to battery life and perf but professional use is much better on Intel/AMD+NVidia. Also you could get much more perf for $ on such machines

> professional use I hate this terminology. How would anyone define "professional use"

Post by sunflowerfly » Sat, Jun 10, 2023, 7:12 PM UTC

microtonal wrote: post #121

adastra22 wrote: post #2

50% slower than the absolute best of class NVIDIA discrete GPU offering that consumes massively more wattage is hardly a damning statement.

Maybe, but most of these comparisons are based on GPU performance (ie. games). For other workloads like machine learning, Tensor cores on NVIDIA GPUs will blow Apple Silicon GPUs out of the water. The M2 Ultra is 27 TFLOPS. The 4080't Tensor Cores are 48.7 TFLOPS in FP32, 194.9 TFLOPS in FP16, 389.9 TFLOPS in FP8 with FP16 accumulate. (IIRC on Apple Silicon GPUs FP16 performance is roughly the same as FP32.) (There i…

Post by imbnwa » Sat, Jun 10, 2023, 7:14 PM UTC

perrohunter wrote: post #106

brucethemoose2 wrote: post #88

Earlier quoted context omitted.

Take anything from wccf with a grain of salt. And yeah, as others point out, this is Apples to oranges. x86 desktops are great at some things, M2 Ultras are great at others, and the overlap that really matters is pretty small... Like, you have to be crazy to buy an M SoC for gaming, or buy a Nvidia GPU for workloads that won't fit in VRAM.

I tested the new game porting kit on a Mac mini M2 and was surprised I was able to run Cyberpunk 2077 and it was running really well, I can imagine that more powerful M processors will be awesome, so the argument of games might go away soon

Post by CamperBob2 » Sat, Jun 10, 2023, 7:15 PM UTC

microtonal wrote: post #121

adastra22 wrote: post #2

50% slower than the absolute best of class NVIDIA discrete GPU offering that consumes massively more wattage is hardly a damning statement.

Maybe, but most of these comparisons are based on GPU performance (ie. games). For other workloads like machine learning, Tensor cores on NVIDIA GPUs will blow Apple Silicon GPUs out of the water. The M2 Ultra is 27 TFLOPS. The 4080't Tensor Cores are 48.7 TFLOPS in FP32, 194.9 TFLOPS in FP16, 389.9 TFLOPS in FP8 with FP16 accumulate. (IIRC on Apple Silicon GPUs FP16 performance is roughly the same as FP32.) (There i…

Post by Demmme » Sat, Jun 10, 2023, 7:17 PM UTC

llm_nerd wrote: post #123

The M2 has been a bit underwhelming in general through all of its iterations: Apple jumped into such a lead with the M1 [1], so it was disappointing when they slowly iterated while Intel and AMD have made enormous strides catching up. Everyone keeps citing TDP, but given that we're talking about desktops that just isn't a huge factor. Having said that, the 7950X was released late February, and the 13900KS was release…