CUDA error for big scenes #39268

New Issue

Lorenzo · 2014-03-19T19:05:47+01:00

Lorenzo commented

2014-03-19 19:05:47 +01:00

System Information
Windows 8.1 - 64 bit;
Nvidia 750m 4GB Vram;

Blender Version
Broken: 2.69.0 r60995 AND 2.7 rc hash 2525f9c

Short description of error
from the console:
invalid value in cuTexRefSetAddress(NULL, texref, cuda_deviceptr(mem.device_pointer), size)

Exact steps for others to reproduce the error
Sorry, it seems to me that when the scene requires around 2GB of Vram, cycles can't render it. Can't go deeper 'cause I'm a real noob. Sorry for that :p

Thanks!

**System Information** Windows 8.1 - 64 bit; Nvidia 750m 4GB Vram; **Blender Version** Broken: 2.69.0 r60995 AND 2.7 rc hash 2525f9c **Short description of error** from the console: invalid value in cuTexRefSetAddress(NULL, texref, cuda_deviceptr(mem.device_pointer), size) **Exact steps for others to reproduce the error** Sorry, it seems to me that when the scene requires around 2GB of Vram, cycles can't render it. Can't go deeper 'cause I'm a real noob. Sorry for that :p Thanks!

Lorenzo commented

2014-03-19 19:05:47 +01:00

Changed status to: 'Open'

Lorenzo commented

2014-03-19 19:05:47 +01:00

Added subscriber: @Mattakkione

Martijn Berger commented

2014-03-19 20:06:11 +01:00

Changed status from 'Open' to: 'Archived'

Martijn Berger closed this issue

2014-03-19 20:06:11 +01:00

Martijn Berger self-assigned this 2014-03-19 20:06:11 +01:00

Martijn Berger commented

2014-03-19 20:06:11 +01:00

Depending on your card and other things you have it doing at the time the actual amount of vram you need can be much greater then what we can report.
Nvidia disabled detailed memory usage access on non pro cards. I am sorry that we cannot report real usage better but it really looks like you need to either trim your scene or textures or use CPU rendering here.

Depending on your card and other things you have it doing at the time the actual amount of vram you need can be much greater then what we can report. Nvidia disabled detailed memory usage access on non pro cards. I am sorry that we cannot report real usage better but it really looks like you need to either trim your scene or textures or use CPU rendering here.

Lorenzo commented

2014-03-19 23:25:27 +01:00

Thanks juicy for the insight.

So as far as I can undestand blender (and others apps to monitor nvidia availabe resources as HWiNFO or Gpuz) can't tell the exact amount of Vram used. Consequentially I can imagine that the RAM used by the CPU for rendering the same scene it isn't at all the exact amount of Vram that the 750m would use for the same rendering, and that means the ratio between RAM/Vram usage is like almost 1/2 'cause for 2 GB RAM I'm using then the Vram filled is near 4GB, right?
But what card could be reliable on the Vram usage? Mybe I coul try to find someone to give me some real reading :)
Thanks for your very fast reply and sorry for my english :p

Thanks juicy for the insight. So as far as I can undestand blender (and others apps to monitor nvidia availabe resources as HWiNFO or Gpuz) can't tell the exact amount of Vram used. Consequentially I can imagine that the RAM used by the CPU for rendering the same scene it isn't at all the exact amount of Vram that the 750m would use for the same rendering, and that means the ratio between RAM/Vram usage is like almost 1/2 'cause for 2 GB RAM I'm using then the Vram filled is near 4GB, right? But what card could be reliable on the Vram usage? Mybe I coul try to find someone to give me some real reading :) Thanks for your very fast reply and sorry for my english :p

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

CUDA error for big scenes #39268