GPU Crash out of Render. #75200

New Issue

Paul Hartsuyker · 2020-03-30T02:26:01+02:00

Paul Hartsuyker commented

2020-03-30 02:26:01 +02:00

System Information
Operating system: Windows-10-10.0.18362-SP0 64 Bits
Graphics card: GeForce GTX 1070/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 442.74

Blender Version
Broken: version: 2.83 (sub 11), branch: master, commit date: 2020-03-28 14:08, hash: a882debee0
Worked: (this has been a problem for awhile, but I can't say which version worked. I will try the file on 2.82A as a confirmation)

Short description of error
[Opened working scene, changed to GPU to try and speed up Render, F12, pause while Initializing, then full crash to empty screen]
I consistently use the Clean Install of NVidia Graphics Drivers, always updating after difficulties.
Thank you again for all of the hard work.

**System Information** Operating system: Windows-10-10.0.18362-SP0 64 Bits Graphics card: GeForce GTX 1070/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 442.74 **Blender Version** Broken: version: 2.83 (sub 11), branch: master, commit date: 2020-03-28 14:08, hash: `a882debee0` Worked: (this has been a problem for awhile, but I can't say which version worked. I will try the file on 2.82A as a confirmation) **Short description of error** [Opened working scene, changed to GPU to try and speed up Render, F12, pause while Initializing, then full crash to empty screen] I consistently use the Clean Install of NVidia Graphics Drivers, always updating after difficulties. Thank you again for all of the hard work.

Paul Hartsuyker commented

2020-03-30 02:26:01 +02:00

Added subscriber: @PaulHartsuyker

Richard Antalik commented

2020-03-30 10:37:13 +02:00

Added subscriber: @iss

Richard Antalik commented

2020-03-30 10:37:13 +02:00

Changed status from 'Needs Triage' to: 'Needs User Info'

Richard Antalik commented

2020-03-30 10:37:13 +02:00

Please go to directory, where blender.exe is located and run blender_debug_log.cmd after crash please upload files that will be created. Also upload simplest possible blend file that does cause crash.

Please go to directory, where blender.exe is located and run `blender_debug_log.cmd` after crash please upload files that will be created. Also upload simplest possible blend file that does cause crash.

Paul Hartsuyker commented

2020-03-30 23:16:01 +02:00

Change title, new info. I was not able to reproduce the Crash as noted, however, Blender left two(2) pieces in Memory, when I checked the Task Manager, after that crash. I had been working for a long period, so some accumulation of 'Cruff' may have triggered the Crash??
However, I had noted that the Render Passes are strange lately, as I usually save out multiple passes to comp together, and this no longer is a working choice with 2.83.11, 03-30-2020.
I don't have the version numbers, just the dates of the renders. Note that the three on the dates for 03-30-2020 are very different with the first one being an earlier build.
The expected look is the 03-18 first panel. Note the bizarre difference with the Shadow and Emit Passes. I have gone back to 2.81, to try and re-establish that look, and I will further investigate my file for any lighting changes, as I had to move the file into other applications for a better chance to find polygon dupes and vertice excess, and brought it back as an ABXC Alembic file.
RenderVariations.zip

Change title, new info. I was not able to reproduce the Crash as noted, however, Blender left two(2) pieces in Memory, when I checked the Task Manager, after that crash. I had been working for a long period, so some accumulation of 'Cruff' may have triggered the Crash?? However, I had noted that the Render Passes are strange lately, as I usually save out multiple passes to comp together, and this no longer is a working choice with 2.83.11, 03-30-2020. I don't have the version numbers, just the dates of the renders. Note that the three on the dates for 03-30-2020 are very different with the first one being an earlier build. The expected look is the 03-18 first panel. Note the bizarre difference with the Shadow and Emit Passes. I have gone back to 2.81, to try and re-establish that look, and I will further investigate my file for any lighting changes, as I had to move the file into other applications for a better chance to find polygon dupes and vertice excess, and brought it back as an ABXC Alembic file. ![Combined_Collage.png](https://archive.blender.org/developer/F8437582/Combined_Collage.png) [RenderVariations.zip](https://archive.blender.org/developer/F8437585/RenderVariations.zip)

Paul Hartsuyker commented

2020-03-31 03:10:56 +02:00

Just checked today's version, blender-2.83-b555b8dedce0-windows64, and Rendered a Cycles pass and a GPU pass for reference. The Shadow Pass for the Cycles is better now that I 'turned on' the Layer Collection (DUH>???) but the GPU is still strange, but doesn't crash. Mystifying. Is anybody else getting strange results?? Try again tomorrow.

Just checked today's version, blender-2.83-b555b8dedce0-windows64, and Rendered a Cycles pass and a GPU pass for reference. The Shadow Pass for the Cycles is better now that I 'turned on' the Layer Collection (DUH>???) but the GPU is still strange, but doesn't crash. Mystifying. Is anybody else getting strange results?? Try again tomorrow. ![03-30-20_Cycles_comp_v4.png](https://archive.blender.org/developer/F8437870/03-30-20_Cycles_comp_v4.png)

Evan Wilson commented

2020-03-31 04:22:04 +02:00

Added subscribers: @brecht, @EAW

Evan Wilson commented

2020-03-31 04:22:04 +02:00

The green in the different passes where it doesn't belong? Yes I am. And I am rendering on my CPU.

Without bisecting, I would guess that ada28d3c65, 3033b2a044, and/or 6777956005 that @brecht committed on the 20th is causing a channel to get bumped into a completely different pass.

This issue should be it's own report though. I will attempt to bisect to get the right commit and file a new report on this issue.

The green in the different passes where it doesn't belong? Yes I am. And I am rendering on my CPU. Without bisecting, I would guess that ada28d3c65, 3033b2a044, and/or 6777956005 that @brecht committed on the 20th is causing a channel to get bumped into a completely different pass. This issue should be it's own report though. I will attempt to bisect to get the right commit and file a new report on this issue.

Richard Antalik commented

2020-03-31 07:51:38 +02:00

@PaulHartsuyker I am not changing subject of report completely. I can close this report if you can't reproduce crash, and you don't want to investigate it further.

Paul Hartsuyker commented

2020-03-31 08:01:23 +02:00

I can not repeat the crash that started this thread, but the renders that I continue to produce are repeatable and constitute a bug. I cannot investigate it further, as it has not repeated. I have had other Bug Reports renamed before to something more appropriate, which is all that I was requesting. Do what you will. I will continue my efforts, but it sound like my reports aren't needed?? Best to you.

Richard Antalik commented

2020-03-31 08:22:12 +02:00

Changed status from 'Needs User Info' to: 'Archived'

Richard Antalik closed this issue

2020-03-31 08:22:12 +02:00

Richard Antalik self-assigned this 2020-03-31 08:22:12 +02:00

Richard Antalik commented

2020-03-31 08:22:12 +02:00

I would rename, if investigation will reveal that cause is something different that was originally reported.
I won't rename this one, in case you experience crash again and you find how to reproduce so we can reopen issue.

Technically it doesn't really matter, but it is better to create new report for different subject so if some dev will read whole thread he don't have to filter through off-topic stuff. You can edit your own reports by the way, but keep in mind that it is better to net deviate from original too much.

I would just recommend either @EAW or @PaulHartsuyker leave here note that you opened new report for new issue so we don't end up with 2 reports

I would rename, if investigation will reveal that cause is something different that was originally reported. I won't rename this one, in case you experience crash again and you find how to reproduce so we can reopen issue. Technically it doesn't really matter, but it is better to create new report for different subject so if some dev will read whole thread he don't have to filter through off-topic stuff. You can edit your own reports by the way, but keep in mind that it is better to net deviate from original too much. I would just recommend either @EAW or @PaulHartsuyker leave here note that you opened new report for new issue so we don't end up with 2 reports

Evan Wilson commented

2020-04-01 18:04:50 +02:00

@PaulHartsuyker doing my first bisecting build, will post the new task and tag you in it after lunch.

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

GPU Crash out of Render. #75200