AMD Radeon S7150x2 - Not being seen by VMs
-
Just to be clear, when you try to install the drivers, from AMDs site, for these VMs, it says it can't detect a AMD GPU? Just making sure I know exactly what's going on, not quite sure on a fix but narrowing down the issue will help.
-
@planedrop Yes, that's right. It can't detect a gpu.
-
Gotcha, I would try to use another OS maybe for a VM and see if Ubuntu sees the GPU? Just to test to see if it's an issue with Windows complaining.
If you check device manager in the Windows VM does it show any display adapters that aren't installed properly?
-
@planedrop So what I have tried so far:
Hardware
CPU model Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
GPUs G200eR2, Tonga XT GL [FirePro S7150], Oland GL [FirePro W2100], Tonga XT GL [FirePro S7150]
Core (socket) 32 (2)
Hyper-threading (SMT) Enabled
Manufacturer info Dell Inc. (PowerEdge R730xd)
BIOS info Dell Inc. (2.12.1)Software
Version 8.2 XCP-NG
Windows 10, versions 1709 to 22H2
Ubuntu 18.04What's interesting is that in device manager there is a PCI card showing up but you cannot install the graphics card drivers to it. In Ubuntu I get an issue where I get logged out in 1 minute. Taking the graphics card out of the equation it runs fine. If I have the graphics card set to be passthrough only, it can be seen but then I have an over kill graphics card on one VM.
I think I'm going to abandon the idea as it's take too long to sort. It seems to me like a bug/setting/issue with the BIOS, hardware and software. But it's too broad to troubleshoot too much more.
Cheers
-
@cunrun Hmmmm, if device manager is seeing it then it sounds like it's working on a hypervisor level, though I'm no vGPU expert.
I'm sorry I don't have a better answer here, my experience mostly lies with passthrough rather than vGPU, though I did try vGPU once with an AMD RX series card (which I believe support vGPU as of recent) but ran into similar issues.
-
Just use official AMD drivers
https://www.amd.com/en/support/professional-graphics/firepro/firepro-s-series/firepro-s7150-x2
It works fine, but sometime Dom-0 "lose" graphic adapter and you should restart whole server....
I did find solution and AMD stopped supporting it.... -
@splastunov I've done that but it prompts that it can't see the graphics card. It works fine on normal pass through.
I think we are going to use pass though to just run a giant VM and chuck graphics related things at it. It was a bargain purchase at £150 for the graphics card so I'm going to keep it. I installed Windows Server 2019 DC and it was a similar result so it's a possibility that the firmware is out of date on the GFX card? Clutching at straws really.
-
@cunrun Are the XCP-ng host and the Windows 2019 Server VM booting in legacy/BIOS or UEFI? Since the FirePros was launched when the legacy/BIOS was still the standard, I'd try the this mode (if not yet).
-
@tuxen Oh excellent suggestion, totally forgot about that being the case.
-
All,
Very good suggestions. I will try them later today!
-
Deal all,
So, non of the above seemed to work. I'd imagine there's some sort of firmware update I need on the graphics card to get this to function properly.
-
Some further updates on this particular issue. I managed to see the graphics card on a Windows 2016 standard (older ISO) install. I installed it natively onto the Dell server. It worked for about 20 minutes before it crashed the server and then kept crashing each time Windows loaded. So it would seem that this may be possible if using Windows 2016 Server standard, and possibly if I can work out how to update the firmware of the graphics unit.
I do know that a Radeon FirePro W2100 and a Radeon PRO WX 3200 work for passthrough. I have both running on one server feeding a 3D globe VM and a 3d processing VM. It's a shame I cannot get the AMD Radeon S7150x2 working.
-
Hello,
Kinda stuck on the same step. MxGPU is visible in XO but not visible by the vm. Just getting started on troubleshooting.
Thanks,
-
@jorge-gbs Hello, I have done a fair amount of testing on this but haven't come up with much. Since the GFX card is from 2016 odd I did find that installing Windows 2016 natively onto the server hardware worked. But as soon as you update the driver you get a blue screen of death. I think this is going to be a firmware issue on the GFX card itself but I don't know how to update that and if it's going to be a pain at this stage, then it sounds like it will not be as reliable as one would hope.
I haven't tried creating a 2016 VM and passing it through. That might work, until it updates at least. You may have more luck than I did, but I still have the card in a server in case one of us figures it out.
Just to recap the things I've tried:
XCP-NG 8.1, 8.2 and 8.3 Beta - All have the same issue as you describe.
Passing through to - 2015 Windows 10 VM up to 2017 Windows 10. Windows 2019. Ubuntu. Nothing worked.
Dell R730xd server and a Dell R740. Same issue.On a side note, I tried the GIU USB pass through feature in 8.3 beta and I had the same results so there's a possibility it's something I haven't done with the Dom0 stuff.
I will ask, to any admins that are reading, that it might be a time to take down the S7150x2 from the documentation as it might be too old to work, now.
-
@cunrun @jorge-gbs any init errors in dom0
/var/log/kern.logre. GIM driver? Also, if you search some topics here covering this specific GPU, there were mixed results booting dom0 withpci=realloc,assign-busses. Maybe it worth a try.
