Hi everyone,
I’m running into major issues with rendering in Houdini (20.5 Indie, macOS version). In my current project, the Render Manager queue stops after 5–10 frames, even though a full frame range is specified. This affects all AOV branches and happens regardless of whether I use Karma CPU or XPU.
When using “Render to Disk in Background”, the queue simply clears without completing the full range. If I try “Render to Disk”, rendering stops at the same point — except this time, my machine crashes due to low memory.
I’ve noticed that each subsequent frame takes longer to render, which makes me suspect that Solaris and/or the render engine is not releasing memory properly between frames. Over time, RAM usage just climbs until it hits a critical point.
Also important: this is not caused by a specific problematic frame. I can resume rendering from where it left off, and it will again stop after another 5–10 frames. This strongly suggests a cumulative issue, rather than one tied to faulty geometry, shading, or a particular frame.
My question:
Is there any way to tell Husk to fully reset or clear memory after each frame? Some kind of flag or workaround to avoid this accumulation?
System specs:
• Houdini 20.5 Indie (macOS)
• Mac Studio M3 Ultra, 256 GB RAM
Any suggestions, tips, or shared experiences would be greatly appreciated. Thanks!
Husk/Karma Render Queue Stops After Few Frames
415 1 1-
- gmarc
- Member
- 38 posts
- Joined: Sept. 2024
- Online
-
- gmarc
- Member
- 38 posts
- Joined: Sept. 2024
- Online
Solved: Motion Blur Node Causing Render Issues
After extensive debugging, I found that the issue was caused by the “Motion Blur” LOP. Replacing it with a “Cache” LOP led to the same problem, which pointed to the placement in the node graph as the actual culprit.
I had originally placed the Motion Blur node just before the Render Settings, which caused it to cache the entire stage—leading to a RAM overflow. Since all my scene objects (especially the scattered ones) already use velocity-based motion blur, I moved the Motion Blur LOP after the Camera LOP and merged only the camera motion into the main graph.
Now, with only the camera using actual transformation-based motion blur, the full frame range renders without any issues.
After extensive debugging, I found that the issue was caused by the “Motion Blur” LOP. Replacing it with a “Cache” LOP led to the same problem, which pointed to the placement in the node graph as the actual culprit.
I had originally placed the Motion Blur node just before the Render Settings, which caused it to cache the entire stage—leading to a RAM overflow. Since all my scene objects (especially the scattered ones) already use velocity-based motion blur, I moved the Motion Blur LOP after the Camera LOP and merged only the camera motion into the main graph.
Now, with only the camera using actual transformation-based motion blur, the full frame range renders without any issues.
-
- Quick Links