Xeon 2698 v4 vs Corona 1.5.2

[pinchuk.don@icloud.com]

All good!
Questions-issues :
When Im rendering absolutely any scene, regardless of the count of Polygons or light or their type, Corona load CPU for 95% at a frequency of 1.88 GHz by core. Thinking that something is wrong with Max or corona settings, just launched corona benchmark 1.3 and got exactly the same result.
My system:
Intel Xeon 2698 v4 ES - 2.0 GHz (2.3 turbo mode)
Mother on the Intel X99
Memory 48 GB 2400 MHz 6 planks
Ps Smash his head, read everything, because Vray 3.40.1 and 3.40.3 tests show the useage of CPU in render time -  100% with a frequency of 2.28 GHz by core
Thank you all in advance for your attention)

[Juraj]

Hi, I wrote about this somewhere in the forum already long time ago :- ) I own a lot of those cpus.

It's because of 2 things:

1) Xeons v4 have AVX and non AVX clocks (and turbos). In case AVX instructions are used, Xeons lower their clock to avoid thermal limit ( which sucks because with proper cooling it would not matter, but these are server cpus so it's hard limitation).
   Intel claims that even underclocked, with AVX the net result is actually better.
   I tested it...and it's true :- ) But it's still lost potential.

2) ES processory have bigger disparity in turbos, and esp. AVX-turbos. That is why you see so much lowered clocks and CPU usage with ES xeons. The difference is 5-8 perc. instead of 2-5 with retail version cpus.

[pinchuk.don@icloud.com]

Hi, I wrote about this somewhere in the forum already long time ago :- ) I own a lot of those cpus.

It's because of 2 things:

1) Xeons v4 have AVX and non AVX clocks (and turbos). In case AVX instructions are used, Xeons lower their clock to avoid thermal limit ( which sucks because with proper cooling it would not matter, but these are server cpus so it's hard limitation).
   Intel claims that even underclocked, with AVX the net result is actually better.
   I tested it...and it's true :- ) But it's still lost potential.

2) ES processory have bigger disparity in turbos, and esp. AVX-turbos. That is why you see so much lowered clocks and CPU usage with ES xeons. The difference is 5-8 perc. instead of 2-5 with retail version cpus.

Tnx for answer)
there is no chance to solve this in future? in new wersions of Corona? or its a concept of Cocona's core?

[Dalton Watts]

Does this only happen with v4 chips?

[Juraj]

Does this only happen with v4 chips?

It should also with v3, both broadwell-e and haswell-e E5(E7) introduced separate AVX clocks. v4 should be smarter in that only cores executing AVX instructions underclock, in v3 all cores underclock. Though I am not sure if that is the case for ES versions.

[Dalton Watts]

One of my Xeon 2696v3 is ES and it also underclocks with AVX2 workload just like its non-ES brother.