Motherboard for Intel Xeon E5 2687W Engineering Sample

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

james-bond

Honorable
Mar 23, 2012
3
0
10,510
Hello,
Does anyone have a suggestion for a X79 motherboard that will work with this ES2 CPU? Doesn't have to be an Asus.
Thanks for the help.
 
OK, after one long night I think I know at least where the problem is:

- QPI and MRC driver (most likely)
- SB driver (maybe)

The first driver's responsibility is so-called "QPI and Memory Reference Code". OEMs use the reference source code provided by Intel and then subsequently change it to fit their motherboard design. The second one is, I suppose, related to the southbridge.

Why I suspect that it is the QPIandMrc driver? Well, this is the code where you implement various errata handling strategies related to the various silicon steppings. It could very well be that ASUS just removed older code paths handling specific behaviors of A1/B0/B1 silicon.

Thing is - you cannot just swap this driver from a board that works... If you try to do it, you will immediately be hit with 55 error during boot (CPU cannot see the memory), which suggests that this code is most likely customized to the specific board layout.

Funny thing is, swapping QpiAndMrc from Z9PE-D16 will work, but you will not solve the 5A issue. Swapping QpiAndMrc from P9X79 Pro or Supermicro x9dai will end up with error #55.

Another interesting thing I noticed when looking into those EFI drivers - it appears Supermicro QpiAndMrc driver is almost the same as Intel's W2600CR, including the debug symbols that are still in. ASUS drivers are much smaller in size, and do not have debug symbols. I suppose ASUS changed/extended this code heavily as they allow memory voltage tweaking unlike Intel or Supermicro.

So, all in all... this means that the only way ES1/ES2 silicon could work is that ASUS fixes this issue, or leaks some older BIOS image. Unfortunately I do not have the tooling necessary to debug this as this would require at least AMI DebugRx dongle and probably also JTAG connection to the board itself.
 
JFYI - it appears that the problem is limited to B1 samples only (e.g. QAMS/QAMT) - B0 (e.g. QA91/QA92) samples work with Z9PE-D8 WS.

B1 samples are almost identical to B0 with very few changes/fixes, so I suppose it is just handling bug in the MRC code which possibly does not cover B1 stepping at all.
 
I can confirm that B0 samples (QA91/QA92) work on Z9PE D8

Interesting enough, if the boot CPU (#1) is B0, and the CPU #2 is B1 (QAMT/QAMS) the board still boots...
 
I got 2 qa92 CPUs and they boot, however the eventually kill the copy of windows. No idea what the go is. Tried the new bios... Same deal. The weird thing is cinebench and windows internal benchmark fail. The windows one causes bsod and the cinebench locks at the start. Anyone run a qa92 in cinebench with success on this board? If so what bios version? Any setting suggestions?
 


That was with the Asus z9pe-d8 ws btw
 

Hi I am new to the forum
I also have two qa92 with the z9pe d8 motherboard and haven't received them yet so I would like to know if possible how your system is doing any problems so far ?
I will mainly use that system for video renderings so I am a bit anxious to know if you had any bsod or crashes etc etc and also have you tested the CPU's overnight using some sort of benchmark that will use the 100% of both CPU's for a significant amount of time because if you have and no problems ,errors , crashes, then maybe I don,t have to worry so much Thank you
 
Still no answer ,,, I was hopping my worries would go away if I had an answer from someone that already has the same hardware oh well maybe I have to wait and see I will post when I setup my new hardware and share my thoughts 🙂
 

I try that after flash bios to 5004 card wil not boot an no light in control led but when i put card in 3rd slot boot up at once no problem so far run better than ever.
Thanks for a real good help :)
Use 2 x 2687w es stepping C0