VM Freezes and Crashes Whilst Playing Games

[ekim]

I am having an issue with my Windows 10 VM where by I'll be playing a game via Steam Streaming and after a little while the game will eventually freeze and the VM gets paused.

 

I have attached two diagnostics, one straight after a crash and another when I try and reboot the VM. Specifically looking at the log from the VM I have the following:

 

2017-01-02T19:02:29.027723Z qemu-system-x86_64: vfio_err_notifier_handler(0000:03:00.1) Unrecoverable error detected. Please collect any data possible and then kill the guest
2017-01-02T19:02:29.031614Z qemu-system-x86_64: vfio_err_notifier_handler(0000:03:00.0) Unrecoverable error detected. Please collect any data possible and then kill the guest

 

03:00.0 and 03:00.1 is my GPU.

 

From the syslog after I reboot I have the following:

 

Jan 2 19:10:22 Tower kernel: vfio_cap_init: 0000:03:00.1 hiding cap 0xff
Jan 2 19:10:22 Tower kernel: vfio_cap_init: 0000:03:00.1 hiding cap 0xff
Jan 2 19:10:22 Tower kernel: vfio-pci 0000:03:00.1: Refused to change power state, currently in D3
Jan 2 19:10:23 Tower kernel: vfio-pci 0000:03:00.1: timed out waiting for pending transaction; performing function level reset anyway
Jan 2 19:10:23 Tower kernel: vfio-pci 0000:03:00.0: Refused to change power state, currently in D3

 

To get everything up and running again I have to power down the server (Power Down only, Reboot doesn't work). The freeze happens irregularly I can never guess how long it would take.

 

My guess is that I need to upgrade my PSU? I have a Seasonic X660 and am running the following:

 

i7 5820k

Geforce 210

Geforce 1070GTX

10 * WD 3tb Reds

1 * 500gb Samsung 850 Evo

13 * 120mm Case Fans

2 * LSI SAS Cards

 

Does anyone else have any other thoughts as to what the issue might be? Or is a PSU upgrade in order?

tower-diagnostics-20170102-1903.zip

tower-diagnostics-20170102-1910.zip

[jonp]

Very odd indeed.  A few questions:

 

Does the VM ever crash when playing WITHOUT steam in home streaming?

Does the VM ever crash when performing activities OTHER than playing a game?

Does the VM ever crash when idle?

[ekim]

Does the VM ever crash when playing WITHOUT steam in home streaming?

 

I haven't tried playing anything without streaming. Will give it a go, but will need to pull the server out.

 

Does the VM ever crash when performing activities OTHER than playing a game?

 

I've only made the VM for gaming, so haven't used it extensively for anything else.

 

Does the VM ever crash when idle?

 

No, it's sat up for a while (a few days) and never had any issues.

 

Only time I have ever had it crash on me is when I am playing a game.

 

I did a little extra research today. I think my PSU should be sufficient for what I have.

 

I read some people getting similar errors if the GPU isn't seated properly, or if there is dust in the connectors or similar. I can't pull my server out right now, but will have a look at the connections tomorrow to see if that is the issue.

 

Another suggestion I had is that it might not be playing nicely with the Geforce 210, although I think that is unlikely.

[ekim]

Ok, just tried playing connected directly to the server. Crashed after a while of playing BF1. The screen just turned off. Same errors as last time.

 

Just opened up the case the GPU seemed well connected to the motherboard. The PCI-Express power connector however whilst it was plugged in it wasn't clicked in to place. I've clicked it in to place now and will give it another go. Hopefully it was just that!

[ekim]

Managed to keep the game running for a much longer time. Thinking it was just the PCI Express power not being connected properly.

 

Have popped the server back in the cupboard and will keep an eye, but thinking this is solved now.