Hi, I'm running into a constant issue where my entire PC crashes with a black screen when using Karma in the Solaris viewport. It pretty consistently happens when I attempt to switch the delegate from XPU to another (VK, CPU, etc.) while the render is running however it also happens when just panning around the scene in XPU. The scene can literally be just a single geo primitive without a camera, materials, lights, ROP.. just using the display settings for the render view, though I have tried setting up more of a scene with a camera / Karma ROP and it's the same result. I also tried the 3 hour process of precompiling the render kernels, but that had no effect. I also don't believe that Houdini is producing a crash file (that I can find) as the entire machine is crashing.

My machine configuration:
Houdini 21.0.440
AMD 3970x
Asrock TRX40 Creator MB (most current BIOS)
128 GB DDR4
MSI Suprim Liquid X 4090 (used for compute - Set in Houdini settings as GPU and in Nvidia app as the only compute device)
Nvidia 2080Ti (used for display)

I'm not using any overclocking of CPU, GPU or memory (BIOS or apps). I tried removing the 2080Ti and only using the 4090 which gives me much faster viewport responsiveness for some reason, but didn't resolve the crashing.

I was able to film an example of the crash with my phone. Attached.

I generally render with Octane in the Obj context and never have any issues. Often times I will use both GPUs in the Octane settings to improve render performance. Occasionally Octane will crash because.. it's Octane, but the 2025 version has been much more stable.

I was hoping that H21 would resolve the Karma crashing issue as I was experiencing it in 20.5, but somehow I can't get past it and it makes Solaris completely unusable.

Any help would be greatly appreciated!
- Travis

This seems like a system/hardware issue. have you tried removing the 4090 (careful with those nasty powerconnectors) and only use the 2080. what happens if you only use embree, does it crash as well? while doing the stresstest, did you monitor the temperature of your gpu/cpu?

an unpopular approach is to install windows fresh (you can try to do that on another drive to see if it actually changes anything).

Hope this gives you some options to look into.

Hi Ronald, thanks for the suggestions, I appreciate it. It seems to have been an issue with the Nvidia USB 3.0 driver. I use the USB 3 port on my 2080 to connect one of my displays. After deleting and reinstalling it the issue went away

For anyone that comes across this post with a similar issue where the machine is crashing and you're not able to deduce the issue do to a lack of a log file. Windows produces dump files which it saves to C:\Windows\Minidump if you download the windows WinDbg app from windows you can analyze the file produced from the crash and it will give you information as to what triggered the crash. In my case it just pointed to a USB3 drivers being the issue, but later I deduced that it was the Nvidia USB3 driver. This video explains it well: https://youtu.be/wdHLuB4lkgg?si=kWBbyInN8oLAMYk6 [youtu.be]