Crunching in a Virtual Machine

robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454552033
RAC: 3756

RE: That is great news!

Quote:
That is great news! Congratulations.


Thanks.

Quote:

Do you know how fast the card is outside a VM so you can compare speeds?

I do not have experience with this cards performance outside of a VM so I can't make a comparison.

I am going to scale back the number of GPU jobs running because this card has one BIG heat sync (no fan) so I want to get a feel for what its temps might be. With 3 jobs running it is running at 60C according to GPU-Z.

I installed p-sensor on the UBUNTU VM and it only reports CPU performance. It seems oblivious to the temps for CPU etc. on the mother board.

Logforme
Logforme
Joined: 13 Aug 10
Posts: 332
Credit: 1714373961
RAC: 0

Since you had such success, I

Since you had such success, I of course had to go back and try it once more. And got exactly as far as before :(

Graphics card looks fine in device manager but when I try to crunch with it boinc gives the error message "GPUs have become unusable; disabling tasks" in the log.

So I swore about the unfairness of the universe, tore the GPU out of the machine and set it back in my workstation. I've done this so many times I'm afraid I'm going to wear out the PCI connector :/

After a while I did what I should have done in the first place: Google it !
Found some links, among them this one. First reply there have this gem: "Any version of BOINC can't crunch on GPU as Microsoft's Remote Desktop replaces the video driver with non-CUDA generic video driver.". And guess what? I used remote desktop to access the VM, and you didn't! They even recommend using UltraVNC just like you use.

So .. Tomorrow I'll give it Yet Another Go. Wish me luck.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 539617522
RAC: 148149

Great you guys figured this

Great you guys figured this out! Now what we really need are comparisons to native performance.

MrS

Scanning for our furry friends since Jan 2002

robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454552033
RAC: 3756

RE: Since you had such

Quote:

Since you had such success, I of course had to go back and try it once more. And got exactly as far as before :(

Graphics card looks fine in device manager but when I try to crunch with it boinc gives the error message "GPUs have become unusable; disabling tasks" in the log.

So I swore about the unfairness of the universe, tore the GPU out of the machine and set it back in my workstation. I've done this so many times I'm afraid I'm going to wear out the PCI connector :/

I appreciate your frustration. I really do.

Here are some things that slammed me. I have an ASUS motherboard. They refer to virtualization in the document. It is defaulted to disable. So I enabled it. Still had problems, my MB document said nothing about VT-d. I even searched in an online document for this attribute and found nothing. Then one night I found a VT-d parameter in the BIOS so I set it. I also set graphics configuration to IGPU (I have built in graphics) and set "IGPU mult" to enabled. I throw this out because if just one bios setting is wrong your done before you start.

Also there is a requirement in Windows that you have installed Microsoft NET framework version 4 or better. Without this the drivers for ATI devices will not install successfully. Also I believe it was a requirement something else but I cannot remember at this point.

Quote:
After a while I did what I should have done in the first place: Google it !
Found some links, among them this one. First reply there have this gem: "Any version of BOINC can't crunch on GPU as Microsoft's Remote Desktop replaces the video driver with non-CUDA generic video driver.". And guess what? I used remote desktop to access the VM, and you didn't! They even recommend using UltraVNC just like you use.

Here is a link to the actual remote product i used. Get both the server and the client. Simple install. Free to try and you can continue to use it after its trial expires, but with less functionality. I can only thank Usoft for not providing me their remote desktop solution in the distro I have. It would have been like running low hurdles without a jock strap. [thwap, thwap]

Quote:
So .. Tomorrow I'll give it Yet Another Go. Wish me luck.

I wish you success. I have put together a README on XenServer on my site. It is not complete but later when I get it "done" if you want to give it a read maybe we can come up with a composite "readme" on how to do this painlessly. I should point out I am retired with time and that you are probably a working guy with family. So if you were willing great if not I understand. I would also be willing to cut/paste it here if that is acceptable by forum standards. It does not have to be on just my website. Of course having said all of this I am assuming that others would want to do this.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6537
Credit: 286443619
RAC: 97544

Cut & paste is fine ...

Cut & paste is fine ... congrats on UOTD too! :-)

[ FWIW on topic : virtualisation is often an exercise in disclosing hidden/differing precepts. There lies assumptionitis*! ]

Cheers, Mike.

* That's where your assumptions become inflamed and need surgical removal ie. assumptionectomy. ;-O

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454552033
RAC: 3756

RE: Cut & paste is fine ...

Quote:
Cut & paste is fine ... congrats on UOTD too! :-)

I will post it here, but give me some time. I need to "come up for some air".

Thanks :>). I had not noticed until later in the day. Talk about getting a "hug" when you needed it.

Quote:

[ FWIW on topic : virtualisation is often an exercise in disclosing hidden/differing precepts. There lies assumptionitis*! ]

Cheers, Mike.

* That's where your assumptions become inflamed and need surgical removal ie. assumptionectomy. ;-O


robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454552033
RAC: 3756

RE: Great you guys figured

Quote:

Great you guys figured this out! Now what we really need are comparisons to native performance.

MrS

I was holding off on responding to your post because wanted to see how these GPU jobs progressed. Like I had said earlier to Logforme I do no have experience with this card (AMD Radeon HD 6450)outside of a virtual machine.

Yesterday when I got my first GPU WUs I was optimistic but as of right now I am only ~33% of my way through the original three so my optimism is fading. Elapsed time is showing ~21:30 hours with a long way to go. I am crunching Binary Pulsar Search (Perseus Arm Survey) 1.39 (BRP5-opend-ati) WUs. I would like to believe it is this card and that it's performance would be the same in an "unvirtualized" world but I can't. Does any one here or in another project have a feel for how this card performs?

[Edit] I checked the stats for this machine at E@H and have not received any credit for these GPU work WUs. At this rate I am projecting 3 days to complete these 1st three GPU WUs.

GPU-Z provides the following:

GPU Core Clock: 625 MHZ
GPU Memory Clock: 666.6 (omininous :>) )
GPU Temp: 57.0C
GPU Load: 80-92%
Memory Usage (dedicated): 580 MB
Memory Usage (Dynamic): 45MB

I omitted fan speed and some other temp attributes.

Logforme
Logforme
Joined: 13 Aug 10
Posts: 332
Credit: 1714373961
RAC: 0

Finally got Boinc to use a

Finally got Boinc to use a GPU in a Virtual Machine.

It looks like my AMD 7970 loses 30% performance on the BRP4G tasks and 40% on the BRP5 tasks (only one data point on virtual BRP5 so far).

BRP4G 
Virtual	4114 4055 4041 4112 3930. Ave 4050
Native	3024 2972 3218 3217 3210. Ave 3128
Delta 1.3

BRP5
Virtual 15952
Native 11759 11148 11112. Ave 11340
Delta 1.4


Setup during this test was a virtual Windows Server 2008 machine with 4 vCPU running 3 simultaneous GPU tasks. At the same time another Debian VM was running a fairly CPU intense job.
Compared to native Windows 7 machine with a 4 core CPU with hyper threading running 3 simultaneous GPU tasks and 3 simultaneous iGPU tasks.
I'll let the effect of the Debian VM cancel out the iGPU jobs for simplicity :)

40% stings a bit. But at the moment I think I'll keep the GPU in the VM and play around with it.

Big thanks to Robl and the rest who helped me through this ordeal :)

robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454552033
RAC: 3756

RE: Finally got Boinc to

Quote:
Finally got Boinc to use a GPU in a Virtual Machine.

Congrats!!!

Quote:

It looks like my AMD 7970 loses 30% performance on the BRP4G tasks and 40% on the BRP5 tasks (only one data point on virtual BRP5 so far).

BRP4G
Virtual 4114 4055 4041 4112 3930. Ave 4050
Native 3024 2972 3218 3217 3210. Ave 3128
Delta 1.3


in the above example the Native data point 3024. What is it referring to? Where is this info located? [EDIT] is the 3024 a posted "run time"?

Quote:


BRP5
Virtual 15952
Native 11759 11148 11112. Ave 11340
Delta 1.4[/code]
Setup during this test was a virtual Windows Server 2008 machine with 4 vCPU running 3 simultaneous GPU tasks. At the same time another Debian VM was running a fairly CPU intense job.
Compared to native Windows 7 machine with a 4 core CPU with hyper threading running 3 simultaneous GPU tasks and 3 simultaneous iGPU tasks.
I'll let the effect of the Debian VM cancel out the iGPU jobs for simplicity :)

40% stings a bit. But at the moment I think I'll keep the GPU in the VM and play around with it.

Big thanks to Robl and the rest who helped me through this ordeal :)

:>)

The performance of my Radeon card is most disappointing. 30 hours into 3 jobs and only half way if I am to believe the "status/performance" bars. What I really need to do and the thought scares the crap out of me is to replace the radeon with a NVIDIA 650 TI. I have two. This way I could do a comparison across a Win7 VM and an Ubuntu box (not a VM). Courage. If a 650TI on a VM drops down significantly then I would have to rethink virtualization from a crunching perspective.

Logforme
Logforme
Joined: 13 Aug 10
Posts: 332
Credit: 1714373961
RAC: 0

RE: in the above example

Quote:
in the above example the Native data point 3024. What is it referring to? Where is this info located? [EDIT] is the 3024 a posted "run time"?


Yes. I looked at the Run Time (sec) column in my tasks page

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.