RT GPU Craching on new rig

[3LP]

Hey there

We have got our new GPU node, looks awesome!

The only thing, it looks like it's crashing a lot.

here are some print screen that I did.


I've got planty of others.
It's crashing with a big scene, and with small scene as well.
A way to make it crash 99% of the time is just changing frame when I'm in a animated camera.

If I set the ray bundle size and rays per pixel a little to high, it crash as well.

I tried as well with only one GPU enable and not all 4, the drivers are crashing sometimes as well.
When rt crashes the driver, sometime the computer doesn't respond and restart.

It might not be RT the problem, it might be my setup, but do you guys now how I could benchmark efficiently this to figure out what component is not working properly?
PSU, Motherboard or GC.

I can't wait to have this baby working properly, but it looks more stable on my old GTX 570 for the moment...

Thanks

Stan

[vlado]

The 702 error means that the GPUs are working too long and that Windows UI tasks are not processed on time. Our latest builds tend to suck out all power out of GPUs, so this tends to happen more often. There are several things you can do:

a) Exclude the system GPU from participating in the rendering so that it is free to service Windows UI requests only; this alone should help a lot. If the MB has a built-in Intel video, it might be best to leave the UI to it. The system GPU is usually the first one;
b) Reduce the ray bundle size/rays per pixel;
c) Increase the Windows GPU timeout period (TdrDelay registry value) as described here: http://msdn.microsoft.com/en-us/libr...=vs.85%29.aspx - something like 10 seconds should work better;
d) Use older drivers - they seem a lot more stable in this regard;

From our experience a) should help a lot (of course, the drawback is that you lose one GPU).

Best regards,
Vlado