AVX-512 Performance With 256-bit vs. 512-bit Data Path For AMD EPYC 9005 CPUs

[coder]

An example of a program that is limited by the memory bandwidth on Zen 5, so it cannot use the enhanced computational throughput of 512-bit AVX-512 is y-cruncher.

See the explanation at:

http://www.numberworld.org/blogs/202...x512_teardown/

IMO, the key question isn't whether you can find some cases where it's memory-bottlenecked, but rather to what extent such cases represent outliers.

I think it's too bold of that author to decry Zen 5 as memory-bottlenecked, on the basis of that one workload.

[coder]

The SVT-AX1 benchmarks perfectly illustrate why I say these high core count CPUs from AMD are a scam.

4k Bosphorus with preset=3 encode is done in 13-14 fps; the source is 4k 120 fps, meaning that a top of the line 128C/256T EPYC is epically slow encoding video at 1/8 real time.

Quit your BS shilling. Nobody claimed video encoding scaled particularly well with respect to CPU cores or that you should buy such a big CPU for that purpose. What GPUs do is instead offload most of the work to special-purpose and fixed-function hardware.

If someone bought a 128-core CPU just for doing encodes of one or two videos at a time, I'd agree that would be a poor decision. However, that's not what these CPUs are marketed or sold for, just what Michael decided to test.

Meanwhile I can buy a cheap $100 Intel Arc video card, put it into a cheap low end computer and encode in real time with much lower power consumption.

You noted the preset. So, please show us where the quality differences were quantified, since you're posing it as an alternate solution.

What an EPIC failure.

Oh, and it also runs like crap on a Xeon. So, I guess that'd make Xeons a scam, too?

If cheaper EPIC runs it faster than Xeons, and EPIC is a failure, what would that make Xeon? An utter calamity?