NVidia: Random crashes in 'DrvPresentBuffers' #85977
Labels
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
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
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
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
EEVEE & Viewport
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
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
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
21 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: blender/blender#85977
Loading…
Reference in New Issue
No description provided.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
System Information
Operating system: Windows-10-10.0.19041-SP0 64 Bits
Blender Version
Broken: version: 2.91.2, branch: master, commit date: 2021-01-19 16:15, hash:
5be9ef4177
Short description of error
On 2.91.2 I'm having random crashes, like 1 crash every hour or so. I'm not simulation or doing any havy computation, only modeling and UV, I can't reproduce them, they happen very randomly and are happening on different files.
I attach here the crash files.
ak47_210225_v01.crash.txt
gluegranade_210221_v01.crash.txt
Detonator_0224_v01.crash.txt
#030721_228OA_WB_FR_t8.crash.txt
Exact steps for others to reproduce the error
It seems like clicking quickly on many items causes the gpu to spike and Blender crashes at gpu 100% utilization (shown in the task manager) from selecting many different objects in the 3D viewport.
During normal work, the same crash happens just after saving or the file autosave when clicking anywhere in the 3D viewport.
Also, clicking any object causes the utilization to spike to 35%-55%, and the software becomes sluggish but doesn't crash.
Added subscriber: @MassimilianoPuliero
#88025 was marked as duplicate of this issue
#87958 was marked as duplicate of this issue
#87765 was marked as duplicate of this issue
#86378 was marked as duplicate of this issue
#87304 was marked as duplicate of this issue
#87224 was marked as duplicate of this issue
#86521 was marked as duplicate of this issue
#86393 was marked as duplicate of this issue
#85686 was marked as duplicate of this issue
#86502 was marked as duplicate of this issue
#86519 was marked as duplicate of this issue
#85953 was marked as duplicate of this issue
#86197 was marked as duplicate of this issue
Added subscriber: @filedescriptor
Changed status from 'Needs Triage' to: 'Needs User Info'
It would be great if we could narrow down the issue. Do you think it could be related to wireframe/x-ray mode? Do you remember what you were doing before the crashes happened?
Hi, well that's the problem, sometimes happens when I switch move, other time when I call the knife tool, other times when I switch between xray and shading.
It's totally random. I'm sorry that I can't be more specific than this.
thank you
Changed status from 'Needs User Info' to: 'Needs Triage'
Added subscriber: @mano-wii
Changed status from 'Needs Triage' to: 'Needs User Info'
All 3 crashes occurred on the NVidia graphics driver:
nvoglv64.dll :0x00007FFED4D636F0 DrvPresentBuffers
Two of them showed an error in the
asset_management_2_8
addon.This addon relies a lot on the drawing API.
Can you check if the crash also occurs if you disable the addon?
Also, can you check if you are using the latest version of it?
Thanks, I'll disable it from the time being.
Added subscriber: @gobb_blend
random crashesto Random crashes in 'DrvPresentBuffers' with RTX 20 SeriesA colleague of mine is experiencing the exact same crashes with his RTX 2060. He's not using the add-on in question.
The crashes even still occur after a factory reset while performing basic tasks such as moving vertices, sculpting, or even when switching modes.
Changed status from 'Needs User Info' to: 'Needs Triage'
Random crashes in 'DrvPresentBuffers' with RTX 20 Seriesto Random crashes in 'DrvPresentBuffers' with RTX SeriesAdded subscribers: @agx9101, @TyrMetHetZwaard
Managed to recreate the problem for me. It seems like clicking quickly on many items causes the gpu to spike and Blender crashes at gpu 100% utilization (shown in the task manager) from selecting many different objects in the 3D viewport. During normal work, the same crash happens just after saving or the file autosave when clicking anywhere in the 3D viewport.
Also, clicking any object causes the utilization to spike to 35%-55%, and the software becomes sluggish but doesn't crash.
#030721_228OA_WB_FR_t8.crash.txt
I can confirm this issue is repeatable and happens right after project saving. I changed the autosave intervals to every 30 minutes to lessen the number of crashes.
Changed status from 'Needs Triage' to: 'Confirmed'
I don't have the hardware to replicate the problem.
But given the repercussion, I believe it deserves a high priority.
However, idk if this is a resolvable problem on Blender side.
(Marking the #eevee_viewport team as this is apparently an OpenGL problem).
Added subscriber: @Juan-Urdaneta
Added subscribers: @ludovic_l, @Jaydead, @lichtwerk
What is the computer specifications for windows 10 developers use at the Blender Foundation to test stability?
Random crashes in 'DrvPresentBuffers' with RTX Seriesto NVidia: Random crashes in 'DrvPresentBuffers'Added subscribers: @phoenixart, @coasterherb
Added subscribers: @DarkkAndre, @obradk
Added subscribers: @pmoursnv, @brecht
CC @pmoursnv, @fclem.
Added subscriber: @fclem
Added subscriber: @bangyuan-zhou
Added subscriber: @APEC
I had the same "DrvPresentBuffers" crashes described in other task
https://developer.blender.org/T84894
In most of the crash reports files I have this text:
System Information
Operating system: Windows-10-10.0.19041-SP0 64 Bits
Graphics card: GeForce GTX 1650 SUPER/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 457.51
Added subscriber: @riverface
It happens roughly every 5 minutes for me, regardless of if I'm even touching the scene.
How far back do we have to roll for it to stop? I couldn't even find threads about this by searching. Why isn't it on the first page?
A lot of people use NVidia, this has to be happening a lot.
Edit: After switching to gameready and restarting, this may have been solved. Can people confirm whether they're using Studio or Game drivers?
Added subscriber: @V0vi
Quick update, I changed the Graphic card yesterday for a RTX 2080 ti and still had the problem until, I Updated the Drivers and Windows crashed. I had to Install Windows once again from scratch, and after serveral hours of working again Blender, I did not experienced the crashes anymore. Maybe It is not Blender, but a corrupted Windows that is causing the Problem. We are a team of 4 persons in the Office with the same configuration (all with RTX 2070S) and only I was having the problem.
Is someone brave enough with the same problem that would like to try reinstall windows from scratch and confirm?
Thanks.
@pmoursnv and @fclem have tried to reproduce this but not succeeded so far. We'll try with more hardware variations, scenes, drivers to find it, but some info would be helpful to pin it down. This merged report has an associated .blend file, might be worth testing that for a few minutes specifically:
https://developer.blender.org/T86519
For users, it would be helpful if you could narrow then when this happens exactly. Did it start happening with a specific driver version? A specific Blender version? Does removing certain types of objects or shaders from a scene stop the crash? Etc.
Added subscriber: @5L1D3
@brecht
Regarding your user question, in the hopes that this might narrow things down even a little.
Short description of error
Hard crash to desktop
Exact steps for others to reproduce the error
It started happening for me after switching over to 2.91.0 (I was using 2.83 beforehand) due to a switch of gpu from titan to msi rtx 3090.
The crashes started happening a number of times a day, and from what i can tell it's happening when rapidly using the interface/selecting either objects or pressing buttons "too" fast in succession, as mentioned in the original post here.
I sadly can't test with the Titan anymore since it's broken, and thus are kinda stuck with 2.91.
Pre-emptive apologies for the vague description above.
Blender version
2.91.0 official release / no addons
System Information
Operating System: Window 10 Home 64 bit
GPU: MSI RTX 3090 Gaming X Trio
Driver: Game Ready 461.09
The exact steps for my system occurs when saving scenes with more than 150,000 verts and then clicking anywhere in the viewport. The larger the scene the more frequent it occurs. It seems to happen immediately following autosave or user save.
You’re welcome to team view connect to my pc and try a file it happens in constantly.
My switch to game ready drivers didn't do squat.
I'm gonna do what another guy did and reinstall windows. My D:\ drive went down in recent months anyway and I've needed a fresh start.
The scene I'm experiencing this in is not over 150,000 verts, but is over 150,000 tris.
I have noticed that issue is gone for me when I deleted object with large image texture on it (8k x 8k).
Others with this issue, do you have objects with large textures on it that not hidden in viewport?
If you have, try to hide this object completely (icons eye, tv, camera off) and test your scene for crashes when you modeling.
Removed subscriber: @filedescriptor
This problem seems by its erratic frequency and non-reproducability to be linked to some "concurrent claiming resources" which is badly handled between the running processes in Blender and concentrating on the use of the NVidia cards and improvements done by NVidia in Blender. The Buffer handling from Nvidia is the prime suspect.
New fact: (I had it 3 times now) is for example: just before crashing the CTRL-B command (Box select) does not work anymore (in the timeline).
The problem escalates in frequency when heavily loading the systems (lots of objects, big textures, many quick commands). For this reason a reinstall of windows, a delete of objects etc... are only shortterm and not solving the cause of the problem. I am not savvy enough to further solve the bug, but as we have a great number of crash reports from different systems all pointing to the Nvidia Buffers it must be possible for an expert (from Nvidia ?) to help here.
Added subscriber: @Robert_B
The last versions of 2.93 crashed when rendering an animation. Individual pictures work well. The project renders very well in 2.92.
Okay now I've isolated when it happens for me, it mostly happens when I'm previewing in EEVEE.
If I keep it solid and don't mess with previewing it doesn't really seem to crash as much.
The last time it happened I was in the progress of toggling the overlays.
Latest crash
I had another crash tonight.
It happened while I was doing a box selection over multiple meshes.
Viewport was in wireframe mode.
My System:
Windows 10 Pro 64
AMD 1950x
64GB ram
(2) 1080ti
Log attached.EL-brutalistm-v013.crash.txt
I updated to the standard version (not dch) of the new Nvidia driver 462.31. The problematic file seems to be a problem no more. I ran DDU prior to installing the new driver. I'll try this week and confirm if the issue has been resolved on my machine. I'm still using Blender 2.92
Windows 10 20H2; GTX 970 Driver 466.11;
Crash when rendering an animation in EEVEE. 1 to 3 images are rendered, then crash.
I've just updated the drivers and it doesn't help me. Blender (2.92) has crashed while I was moving some objects in rendered preview (cycles) mode.
It appears to be a problem with the motion blur in EEVEE. When I turn off motion blur, it renders stable.
@brecht @fclem Still haven't reproduced it exactly. But based on the stack trace and exception info it looks like the problem is a broken VBO binding. The driver is trying to interpret a vertex attribute offset (that Blender set up via "glVertexAttribPointer" previously) as a CPU pointer, rather than a VBO offset, which happens if no valid VBO is bound (as defined in the OpenGL spec).
I can force pretty much the same crash to happen if I replace
glBindBuffer(GL_ARRAY_BUFFER, vbo_id_);
withglBindBuffer(GL_ARRAY_BUFFER, 0xdeadbeef);
in source\blender\gpu\opengl\gl_vertex_buffer.cc. This would indicate that at some point for some unknown reason a VBO ID breaks in Blender leading to the crash everybody above is experiencing. Maybe Blender tries to render a deletedGLVertBuf
object (and therefore accessing bogus data and binding a broken VBO ID)?Playing with VBO timeout, setting it to 6 seconds, means it crashes the same way in that time (during EEVEE preview render) since I start handling pieces of my armature or mesh. It still only happens in Preview though.
Is this a file coming from an earlier version? Or is it a new file? Because from a new file I haven't quite been able to replicate it. Only from a file that's already been saved with complex data.
Whatever other thing is happening VBOs do seem to be the biggest factor here. I say this theory is good.
EDIT:
Set VBO timeout to 0 for now, haven't crashed since.
EDIT 2:
The same instance of the same scene has been running perfectly for 4 hours.
No hiccups whatsoever.
Added subscribers: @Phred42, @krooooglov, @filedescriptor, @PratikPB2123
I found the problem with me. I found an object where the displacement modifier was not applied. The problem is displacement modifier with motion blur in EEVEE.
Added subscriber: @IneCmb
Changed status from 'Confirmed' to: 'Resolved'
I switched to 2.93 beta, changed nvidia drivers to studio, switched off all addons I dont use and set autosave function to 2 files in 1 minute interval. No crashes now.
Changed status from 'Resolved' to: 'Confirmed'
@Robert_B, your problems seems unrelated to this task, if you have further info please post it in #87464 (EEVEE Animation crashes with Motion Blur and Displacement modifier).
@krooooglov, this task contains many merged reports, we keep it open until we confirm it has been solved for most people. Feel free to unsubscribe from this task through if it's no longer relevant to you.
@fclem @pmoursnv I guess it's clear from @riverface's comment that this is related to VBO garbage collection, hopefully setting that to a very short interval makes it easy to repro, or at least provides a clue for which code to check for correctness.
Removed subscriber: @filedescriptor
So, I tried with 2.93.
I've also updated the Nvidia driver to the latest 462.31
At the moment of the crash I had two 3d viewports active, one on solid mode and the other with Cycles render view active.
While selecting some objects in the viewport on solid mode the crash happened.
Log attached head-projection-test.crash.txt
Added subscriber: @PlusUltra11
I updated to Studio Driver 462.31 and switched to Blender 2.83 LTS.
I've noticed I'm getting "Display driver nvlddmkm stopped responding and has successfully recovered." error more than the nvoglv64.dll on my Event Viewer when my blender crashes and I've tried adding TdrDelay on my regedit and didn't notice any startup crashes or on the new scene default cube and UI navigation. I'll still observe if it crashes on further working inside blender.
EDIT: I'm still experiencing silent crashes to desktop after startup but less frequently than before.
I'm trying to reproduce the crash on a rather old setup (Windows + GTX 960 + 466.11 Driver).
The only thing I could do to make it crash was to load a very big scene (like the 2.81 splash screen blend from the demo files page) and enable high precision shadows (lincreasing video memory usage).
It crashes because of out of video memory. But I'm pretty sure this is not what the crash reported here is about. It does crash just after a
glDraw***
function calls so it might also be the same issue.I tried using the file from #86519 but nothing crashed.
@riverface It would help a lot if you could share the blend file, since you seem to be able to reproduce the issue quite consistenlly.
I encourage anyone who has a crashing file to do the same.
Also if you can check your GPU memory usage to see if the crash comes from out of memory issue. The vram could very well be leaked by the garbage collection system.
@fclem, I get a quick crash when I disable any timeouts, and then go back and forth between Layout and UV editing workspaces (with factory startup).
Patch to disable timeouts: P2092
Backtrace: P2093
I am uh, a very niche artist. I don't think you want the scene. I could try to make one of comparable size with a bunch of random assets though, replicate it that way?
I still couldn't get a crash on either my linux workstation nor the nvidia one.
After looking at the code I can see a potential cause of this. Does P2097 still produce a crash?
@fclem, that patch fixes the crash for my simple test.
My file crashes a few times per hour. {F10043875}>>! In #85977#1153021, @fclem wrote:
Added subscriber: @arczii06
Added subscriber: @imasagittariusrising
Added subscriber: @TwinLoki
This issue was referenced by
96abe5ebbc
Changed status from 'Confirmed' to: 'Resolved'