Iray states that it's out of memory even though I have an abundance available- I believe the GPU memory gets fragmented after a while then defaults to CPU & Host Mem then crashes everything when full

Votes
1

Product:
iClone 7

Version:
7.82

Status:
Not Reproducible

Issue 7332

Iray states that it's out of memory even though I have an abundance available- I believe the GPU memory gets fragmented after a while then defaults to CPU & Host Mem then crashes everything when full

Using an RTX 2080 Ti with an i7, 32 GB host ram and 32 GB of virtual ram
I'm not using Smooth mesh by subdivision -you had discovered that as incompatible with Iray.
I am using cloth physics, is this causing the same incompatibility under the hood as if I wear using smooth mesh by subdivision?
Initially when I open the project fresh it renders fine with the GPU.
In the log I get the message: "1.14 POST rend warn : Postprocessing is running on the CPU, but Denoiser is requested. Denoising is not available on CPU." Maybe when there is an error it defaults/falls back to CPU only mode?
More errors: "1.14 IRAY rend error: CUDA device 0 (GeForce RTX 2080 Ti): an illegal memory access was encountered (while allocating host mapped memory buffer)
1.14 IRAY rend error: CUDA device 0 (GeForce RTX 2080 Ti): device cannot be used as it does not support mapped memory (canMapHostMemory)"

I can't accurately reproduce steps because it's intermittent, except for moving the camera often.. it seems it just has a meltdown..

My overall belief is that I think the GPU memory gets fragmented and that iClone needs to release the video memory after so many steps.. and how I know this is a comparison - it defaults to the CPU & host memory -right? and in my case I have a lot more time/host memory before Iray/iClone crashes.

OS: Windows 10

Attachment:

Iray With IBL -out of memory error.jpg

Bug Iray Render

Submitted byAscensi 2020/10/08 15:40:44

COMMENTS (3)

Ascensi 2021/09/29 17:37:34

Private Comment

Feedback Tracker Admin 2020/10/27 01:48:09

Hi Ascensi,

Can you provide your project file to us?

Molly
Reallusion

Feedback Tracker Admin 2020/10/20 04:45:00

Hi Ascensi,

Thank you for your feedback.Please provide your project file to help us identify the problem. Attach the file in your next comment. If the file size is larger than 10MB, you can use cloud services such as Google Drive, Dropbox or OneDrive to make the files available to us and put the link in your comment. You can use Private Comment if you want to keep it private.Your help is highly appreciated.

Molly
Reallusion