cuMEmcpyHtoD_v2 timeout after render was (almost) completed #105490

Closed
opened 2023-03-06 15:56:38 +01:00 by Ivo van der Lans · 4 comments

System Information
Operating system: Windows-10-10.0.19044-SP0 64 Bits
Graphics card: Quadro M2000M/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 516.25

Blender Version
Broken: version: 3.4.1, branch: blender-v3.4-release, commit date: 2022-12-19 17:00, hash: rB55485cb379f7
Worked: (Blender 3.3.4 failed too)

Short description of error
Having loaded the Blender 3.3 LTS splash demo file, I changed to max_samples = 30 and got a timeout on cuMEmcpyHtoD_v2 after the final sample which turned the result image to black. Thereafter, the GPU kept yielding errors.

Exact steps for others to reproduce the error

  • I have an NVidia GPU like this: image.
  • I downloaded the Blender 3.3 LTS splash demo file.
  • I started Blender and loaded this demo file
  • I changed render.max_samples from 300 into 30 and started rendering by pressing F12:
    image
  • While rendering, the screen updates were fine:
    image
    image
    image
  • After the final sample, I got a timeout on cuMEmcpyHtoD_v2 which turned the result image to black:
    image
  • After re-starting the render using F12, I got this:
    image
  • Further attempts to re-render failed with various error codes.

Blender 3.3.4 failed too:
image
image

**System Information** Operating system: Windows-10-10.0.19044-SP0 64 Bits Graphics card: Quadro M2000M/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 516.25 **Blender Version** Broken: version: 3.4.1, branch: blender-v3.4-release, commit date: 2022-12-19 17:00, hash: `rB55485cb379f7` Worked: (Blender 3.3.4 failed too) **Short description of error** Having loaded the [Blender 3.3 LTS splash](https://download.blender.org/demo/splash/blender_3_3_lts_splash_by_piotr_krynski.blend) demo file, I changed to max_samples = 30 and got a timeout on cuMEmcpyHtoD_v2 after the final sample which turned the result image to black. Thereafter, the GPU kept yielding errors. **Exact steps for others to reproduce the error** - I have an NVidia GPU like this: ![image](/attachments/95d18772-7afe-43b6-836e-12a737f02b11). - I downloaded the [Blender 3.3 LTS splash](https://download.blender.org/demo/splash/blender_3_3_lts_splash_by_piotr_krynski.blend) demo file. - I started Blender and loaded this demo file - I changed render.max_samples from 300 into 30 and started rendering by pressing F12: ![image](/attachments/c19adabe-7096-4c62-84d6-3f3ea29e6238) - While rendering, the screen updates were fine: ![image](/attachments/adb234dd-7282-4310-b0f0-f998b17628bf) ![image](/attachments/9ee185c2-b291-4ff6-9f64-53ee71d9b5a1) ![image](/attachments/2de1abcb-5280-4a6f-8b1b-b686d0d22741) - After the final sample, I got a timeout on cuMEmcpyHtoD_v2 which turned the result image to black: ![image](/attachments/cdb01061-be92-4a21-b8ba-33f7f3e722e1) - After re-starting the render using F12, I got this: ![image](/attachments/980cdc0f-5347-4a04-afc5-1af84389789f) - Further attempts to re-render failed with various error codes. Blender 3.3.4 failed too: ![image](/attachments/8e31655d-e5d4-4468-9214-fb9df704a203) ![image](/attachments/ed6d95dd-d687-4c54-9748-f587700c58ee)
Ivo van der Lans added the
Severity
Normal
Status
Needs Triage
Type
Report
labels 2023-03-06 15:56:39 +01:00
Iliya Katushenock added the
Interest
Render & Cycles
label 2023-03-06 16:18:51 +01:00
Member

Hi, thanks for the report. Unable to confirm the issue locally.
Does it work with newer GPU drivers?: https://docs.blender.org/manual/en/dev/troubleshooting/gpu/windows/nvidia.html
Is this a CUDA only issue? Can you replicate it with Optix preference?
Maybe "increase the timeout" is the solution here. Can you try that?: https://docs.blender.org/manual/en/dev/render/cycles/gpu_rendering.html#the-nvidia-opengl-driver-lost-connection-with-the-display-driver

**System Information**
Operating system: Windows-10-10.0.22000-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3050 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 512.78```
Hi, thanks for the report. Unable to confirm the issue locally. Does it work with newer GPU drivers?: https://docs.blender.org/manual/en/dev/troubleshooting/gpu/windows/nvidia.html Is this a CUDA only issue? Can you replicate it with Optix preference? Maybe "increase the timeout" is the solution here. Can you try that?: https://docs.blender.org/manual/en/dev/render/cycles/gpu_rendering.html#the-nvidia-opengl-driver-lost-connection-with-the-display-driver ``` **System Information** Operating system: Windows-10-10.0.22000-SP0 64 Bits Graphics card: NVIDIA GeForce RTX 3050 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 512.78```
Pratik Borhade added
Status
Needs Information from User
and removed
Status
Needs Triage
labels 2023-03-09 02:52:11 +01:00

Hi Patrik,

I updated my NVIDIA drivers.

Using Optix does work; the issue is related to CUDA only as it appears. Interestingly, if I use CUDA with the tiling option and set the tile size to 512, then the rendering completes without a problem. image.

I am working on a company laptop. I cannot change the timeout without breaking company-rules.

It is understandable that Cycles is limited by the transfer-rate of the hardware and given time-outs. Working around such time-out by providing alternative transfer mechanisms might be too complex a solution, since using tiles is a viable solution. Maybe, it helps to append the hardware time-outs related error messages with the a suggestion: "Try again using smaller render tiles." The usability of encountering such error message is however quite low since one would receive the error message after a possibly very long rendering time. Could Cycles test for possible time-outs before doing the actual rendering? Then the error message would read: "We identified that your render configuration cannot support this scene. We suggest to use render-tiles of size 512 by 512 and try again."

Yet, it remains a mystery why retrieving the final render result causes an error while the intermediate render results are displayed without a problem.

Maybe, it is helpful to have a system-wide limitation on tile-size to avoid time-outs in general. So on my system I would then configure Cycles to be limited to 512 by 512 tiles.

Hi Patrik, I updated my NVIDIA drivers. Using Optix does work; the issue is related to CUDA only as it appears. Interestingly, if I use CUDA with the tiling option and set the tile size to 512, then the rendering completes without a problem. ![image](/attachments/d1f79923-4b24-422a-9a21-7509eb418464). I am working on a company laptop. I cannot change the timeout without breaking company-rules. It is understandable that Cycles is limited by the transfer-rate of the hardware and given time-outs. Working around such time-out by providing alternative transfer mechanisms might be too complex a solution, since using tiles is a viable solution. Maybe, it helps to append the hardware time-outs related error messages with the a suggestion: "Try again using smaller render tiles." The usability of encountering such error message is however quite low since one would receive the error message after a possibly very long rendering time. Could Cycles test for possible time-outs *before* doing the actual rendering? Then the error message would read: "We identified that your render configuration cannot support this scene. We suggest to use render-tiles of size 512 by 512 and try again." Yet, it remains a mystery why retrieving the final render result causes an error while the intermediate render results are displayed without a problem. Maybe, it is helpful to have a system-wide limitation on tile-size to avoid time-outs in general. So on my system I would then configure Cycles to be limited to 512 by 512 tiles.
4.3 KiB
Member

Hi, will close this ticket as one of the workaround i.e "reducing tile size" has fixed the issue at your end.

Maybe, it helps to append the hardware time-outs related error messages with the a suggestion:


Could Cycles test for possible time-outs before doing the actual rendering?

This improvements would be good but they are not handled on bug-tracker.

Please submit the request on RCS.

Hi, will close this ticket as one of the workaround i.e "reducing tile size" has fixed the issue at your end. > Maybe, it helps to append the hardware time-outs related error messages with the a suggestion: - - - > Could Cycles test for possible time-outs before doing the actual rendering? This improvements would be good but they are not handled on bug-tracker. Please submit the request on RCS.
Blender Bot added
Status
Archived
and removed
Status
Needs Information from User
labels 2023-03-15 02:59:16 +01:00

I agree. Thanks for your effort :).

Please do consider to put my suggested changes on the backlog.

Ivo

I agree. Thanks for your effort :). Please do consider to put my suggested changes on the backlog. Ivo
Sign in to join this conversation.
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset System
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Code Documentation
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Viewport & EEVEE
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Asset Browser Project
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Module
Viewport & EEVEE
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Severity
High
Severity
Low
Severity
Normal
Severity
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
2 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#105490
No description provided.