just wondering, becaus I tested MU72 SU0 9 months ago, default BIOS F23 work with C0. Of course, QWAN are better choices.
LGA4189 boards are still cheaper in my area. and he is talking about LGA4189-QV2F.
ES Xeon Discussion
- Thread starter britinpdx
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Okay do you know a regular motherboard for the QWAN chip atx or eatx i guess?
I saw Asrock W790 WS mentioned earlier in the thread for 8461V ES QYFU on a spreadsheet but can only find the R2.0 version of this board is that still compatible?
Motherboard ES CPU Compatibility List
Rev. 1.x, Rev. 2.x ??just wondering, becaus I tested MU72 SU0 9 months ago, default BIOS F23 work with C0.
we are at BIOS F30
Is the QYFU good without overclocking or is there a better processor for the R2.0 board?
there is any time a better CPU, depends on your budget. but good for a start.
The only negative is the expensive DDR5 memory.
Well I more wanted one of these xeon es just because they are cool total for the Asrock W790 WS R2.0 + QYFU is around £650
If there is a good MOBO + CPU bundle in that price range that is better I could get that
I dont mind buying the DDR5 also
the problem is you can only upgrade to Xeon W QS/PRD, QS/PRD scalables don't work.
there is currently a MS33-CPA-V10(OEM) board from Netherlands available, i can say more about BIOS etc. once mine arrives:
edit:
MS33-CPA-V10 BIOS F01 07/12/2023 (similar to F17 of MS33-AR0)
NO support from gigabyte for private consumer
you can use BMC FW from other MS33 board.
note that you need RDIMM, UDIMM don't work, UDIMM don't fit.
Last edited:
What about if I went with the ASUS W790-ACE is that a better option for upgradeability? i dont mind paying a bit more the issue is mainly motherboard size
W790
upgrade = leave SPR-ES(scalable), goto QS/PRD.
you can only upgrade to (expensive)Xeon W QS/PRD,
QS/PRD scalables don't work.
C741
you can go to any cheap acalable E0,E2,E3 ES or QS or final PRD unit.
So would that be something like the Supermicro X13SEI-F Motherboard LGA-4677 Intel C741? quite lacking in io but guess i could add a card for sata etc
I just saw on the spreadsheet there is Gigabyte MS03-CE0, this supports more chips do you know? I saw someone say it supports emerald rapids are the es for that already?
Gigabyte C741 boards like MS03-CE0 support all types of ES with latest BIOS so far.
update 23.10.24: ES2 A0 C06F1 do not work without BIOS modify.
- some rare exceptions.
EMR: there are EV-QS & QS but only a few real ES.
intel doesn’t throw out many EMR ES because the platform is already build,
note: EMR A0 & A1 share microcode but use diff. cpuid, the uncore is handled differently by the UEFI/BIOS.
EMR QDF A0 ES2:
Q2SP ES2 A0 C06F1 60C 5600 350W
Q2SR ES2 A0 C06F1 64C 5600 350W
Q2T7 ES2 A0 C06F1 52C 4800 300W
Q2T5 ES2 A0 C06F1 52C ????
Q2Y6 ES2 R0 C06F1 16C 4800
Q2Y7 ES2 R0 C06F1 32C 5600
Q2Y8 ES2 R0 C06F1 24C 4800 185W
Q30G ES2 A0 C06F1 48C 4800 330W
Q30H ES2 A0 C06F1 56C 5600 350W
Last edited:
Has anyone found/came up with any solutions to Intel's high core count parking issues and stuttering? Primarily the, "If you're not on high performance profile or linux equivalent"
Have found solutions to this, but only solution is BIOS adjustments but either it's 600+ watts or 300 watts at 1ghz all core. This happens whether ES or not. Same issues on W9 Xeons, but Intel/Microsoft feigns no known issues and it's littered everywhere, primarily on W11, and Server 2022... Have ran W10 with no issues, but W10 is EOL next year.
Have found solutions to this, but only solution is BIOS adjustments but either it's 600+ watts or 300 watts at 1ghz all core. This happens whether ES or not. Same issues on W9 Xeons, but Intel/Microsoft feigns no known issues and it's littered everywhere, primarily on W11, and Server 2022... Have ran W10 with no issues, but W10 is EOL next year.
in my experience this is not alone a parking issue. imho windows is misconfigurated for SPR.
Windows increases the uncore perf.state/clocks way too late.
e.g. 60% load is necessary for windows to incr. the uncore perf.state.
way too much for high core count processors.
and i guess the load scheduler wait's not long enough(very close to the knife edge) after wakeup a core, runs in timout and try's next lazy core until he is back to the first core now already wake and so on....
Last edited:
Right on the nail, SPR issue now... Do not experience this with Broadwell, Gen1/2/3 Skylake anymore now that I double check... used to at one point, and just set all towerrs to EEB/Bios perf settings... But Linux (lesser issue)/Windows/Server 22 can run balance setting just fine now on those today. Whenever they find solution for this on SPR I'll go back.
Be nice if more people running HEDT SPR complain more about it...
EPYC hcc was similar first, same with high core count Ice Lake.
i solved that by adjusting the min. perf.state to around 50% (let cores stay around 900mhz at near IDLE)
you can run a small miner in the back to keep some cores awake.
linux and Windows server use other thread scheduler. I think that's where the dog is buried.
in fact the issue is proportional to the core count.