Drag to reposition cover

ASUS Pro WS W790E-SAGE SE + Intel Xeon Sapphire Rapids SPR-SP

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
D

DHamov

Active Member
Jan 12, 2024
113
28
28
Thanks for the fast replies.
bayleyw said:
That seems about right - the 3495x is a similar die configuration and posts very bad Time Spy scores:
...
I wonder what's going on, the 3475X is XCC as well but scores almost double...except when stock, where it apparently performs like a Core i3?!
Click to expand...
So i guess all these weird results are normal (not ES related). But yes very very strange. Expecially the gigantic difference in on OC or stock. Maybe its really a benchmark issue. Did some weird run disabling half the cores, enabling 3.7Ghz but still low results.
nofull_load.PNG
But at least i know now that its a cpu related thing, and probably nothing wrong with the gpu. Seems the cpu cant load the gpu or so, for this benchmark. Other silly benchmarks like furmark i can heat it up to full load. Before i was also afraid its a pcie issue, related to the cpu. Also in the ws790 bios i cant find any pcie speed setting, But as long as it works its ok.
 
J

JosefHrib

Active Member
Jul 25, 2023
118
103
43
39
DHamov said:
Thanks for the fast replies.

So i guess all these weird results are normal (not ES related). But yes very very strange. Expecially the gigantic difference in on OC or stock. Maybe its really a benchmark issue. Did some weird run disabling half the cores, enabling 3.7Ghz but still low results.
View attachment 34268
But at least i know now that its a cpu related thing, and probably nothing wrong with the gpu. Seems the cpu cant load the gpu or so, for this benchmark. Other silly benchmarks like furmark i can heat it up to full load. Before i was also afraid its a pcie issue, related to the cpu. Also in the ws790 bios i cant find any pcie speed setting, But as long as it works its ok.
Click to expand...
In my bios (first version 0215) is option set PCIe speed. Yes, my results are with enabled all options in bios ReBAR,Above4G,etc.
 
R

RolloZ170

Well-Known Member
Apr 24, 2016
7,379
2,293
113
DHamov said:
Did some weird run disabling half the cores, enabling 3.7Ghz but still low results.
Click to expand...
still 2 Ghz away from good desktop processors.

to mayn cores: like emptying a fish pond with 48 people but only one bucket
 
J

JosefHrib

Active Member
Jul 25, 2023
118
103
43
39
RolloZ170 said:
still 2 Ghz away from good desktop processors.
Click to expand...
Yes, it is true.

DHamov said:
Thanks for the fast replies.

So i guess all these weird results are normal (not ES related). But yes very very strange. Expecially the gigantic difference in on OC or stock. Maybe its really a benchmark issue. Did some weird run disabling half the cores, enabling 3.7Ghz but still low results.
View attachment 34268
But at least i know now that its a cpu related thing, and probably nothing wrong with the gpu. Seems the cpu cant load the gpu or so, for this benchmark. Other silly benchmarks like furmark i can heat it up to full load. Before i was also afraid its a pcie issue, related to the cpu. Also in the ws790 bios i cant find any pcie speed setting, But as long as it works its ok.
Click to expand...
If it can help, I can post here my results with RTX3090Ti with my i9-13900KS (I created test in July 2023). I go find it.

Below you can find my stress test from July 2023 with my i9-13900KS+RTX3090Ti, and QYFS+RTX3080Ti,QYFS+RTX4090.

20230831 QYFS+RTX3080Ti 3DMark Speed Way 5240 points


20240104 QYFS+RTX4090 3DMark Speed Way 10184 points


20230715 i9-13900KS+RTX3090Ti 3DMark Speed Way 5951 points (best loop in stress test)
And as I wrote in my previous message, RTX3080Ti is very similar in performance with RTX3090 and RTX3090Ti.
RTX3080Ti GA102-225 10240 streams 12GB 384bit GDDR6X memory modules from one side.
RTX3090 GA102-300 10496 streams 24GB GDDR6X 384bit memory modules from two sides.
RTX3090Ti GA102-350 10752 streams 24GB GDDR6X 384bit memory modules from one side.


And below you will find today created tests.

20240204 QYFQ+RTX4090 3DMark Time Spy Extreme 14157 points


20240204 QYFQ+RTX4090 3DMark Speed Way10267 points
 
Last edited:
D

DHamov

Active Member
Jan 12, 2024
113
28
28
JosefHrib said:
20240204 QYFQ+RTX4090 3DMark Time Spy Extreme 14157 points
Click to expand...
Thanks. I dont have the paid version of 3DMark so i can only relate to these time spy results.
Your patterns look quite similar, rarely close to 100% gpu load. and if a 3090 must score here around 19000, the 4090 should be quite a bit higher. EDIT Sorry i confused time spy and time spy extreme.
Good to know, about this issue, but maybe it is also still just a particular benchmark issue, not so relevant for other loads.

Today i became aware that there exist windows 11 pro for workstations. and in the german website they say something like
"Expanded features for multitasking without delay" i dont know if it is just marketing bla bla, or maybe an OS that works better with so many cores as in these Sapphire Rapids. Did anyone try this? Because if i do not disable half of the cores, the cpu clocks were rarely 3.7GHz and mostly 2.6, in the time spy runs, while the benchmark did not use a lot of threads. So hypothetically it could clock up, but it feld, like it was not even using the speed it had available. If will try to find one benchmark that is available in linux as well as windows, to see if there is much difference, there from the OS.
 
Last edited:
D

DHamov

Active Member
Jan 12, 2024
113
28
28
JosefHrib said:
In my bios (first version 0215) is option set PCIe speed. Yes, my results are with enabled all options in bios ReBAR,Above4G,etc.
Click to expand...
Do you use bios 0215, because you have to, or because you want to? I mean Can you go to higher bios version on the Asus w790 SAGE if you wanted? I ask because i am considering maybe still returning my board and go for a SAGE, but if the bios is to new maybe my QYFS would not run.
Also if i keep the ACE, bios 0215 was not online at the Asus website any more so i could never try it. Would you have the installer? Is it the same bios for SAGE und ACE?
 
K

Kizune

Member
Dec 2, 2022
78
56
18
In 3DMark the Time Spy and Time Spy Extreme are pretty old benchmarks and they are known to misbehave on processors with large number of threads. Here are two comparison of two runs i done (QYFS, Asus ACE board, latest BIOS, RFTX 3090, both were made on the same system 3 minutes after Windows 11 started up):
3dmark-analyze.png

Note that first result is much better. The first result i did disabling 6 cores on each CCD having as a result 32 cores/64 threads instead of 56 cores/112 threads. Locking down cores lifted CPU score from 12144 to 15400.
 
  • Like
Reactions: JosefHrib
J

JosefHrib

Active Member
Jul 25, 2023
118
103
43
39
Kizune said:
In 3DMark the Time Spy and Time Spy Extreme are pretty old benchmarks and they are known to misbehave on processors with large number of threads. Here are two comparison of two runs i done (QYFS, Asus ACE board, latest BIOS, RFTX 3090, both were made on the same system 3 minutes after Windows 11 started up):
View attachment 34291

Note that first result is much better. The first result i did disabling 6 cores on each CCD having as a result 32 cores/64 threads instead of 56 cores/112 threads. Locking down cores lifted CPU score from 12144 to 15400.
Click to expand...
Yes, I agree with that on 100%. :)
Some older applications, games, or here older benchmark can works better with some deactivated cores.

@DHamov For my work flow and my few games is this not better, I prefer always more cores. But a hundred people a hundred opinions. Each person may prefer something different. Or in some situation it may be useful to deactivate or leave it on. As you wrote, if you deactivate for example half chip so some older benchmark can jumped for many percentages.
 
Last edited:
You must log in or register to reply here.
Top Bottom