Blender-hip segfaults when trying to render anything using the GPU

Help
1

I just installed the blender-hip package. Blender is detecting my GPU. I can select it in Blender preferences. But the second it tries to render anything, Blender crashes. My GPU is the iGPU in a Radeon 7940HS (a Radeon 780M).

Steps to reproduce:

  1. Create new Blender project
  2. Edit → Preferences → System, change hardware acceleration to HIP and make sure the GPU is checked
  3. In the right sidebar of Blender, in Render Properties, change the render engine to Cycles and the device to GPU Compute
  4. Change the viewport shading to Rendered
  5. Middle click and drag on the viewport to change the camera angle
  6. Blender exits with a segmentation fault trying to render the default cube

Sometimes it will render one frame before it crashes, but never more than that.

Here’s the crash log:

# Blender 4.2.3, Unknown revision
bpy.context.space_data.shading.type = 'RENDERED'  # Property
bpy.context.space_data.context = 'RENDER'  # Property
bpy.context.scene.render.engine = 'CYCLES'  # Property
bpy.context.scene.cycles.device = 'GPU'  # Property

# backtrace
/run/current-system/sw/bin/blender() [0xdf3b63]
/run/current-system/sw/bin/blender() [0x68bc30]
/nix/store/nqb2ns2d1lahnd5ncwmn6k84qfd7vx2k-glibc-2.40-36/lib/libc.so.6(+0x40620) [0x7f8d88e47620]
/run/current-system/sw/bin/blender() [0x2d6261a]
/run/current-system/sw/bin/blender() [0x2d7164b]
/run/current-system/sw/bin/blender() [0x2d7235e]
/run/current-system/sw/bin/blender() [0x2eae5cc]
/run/current-system/sw/bin/blender() [0x2eb03ab]
/run/current-system/sw/bin/blender() [0x2eb1008]
/run/current-system/sw/bin/blender() [0x2eb122f]
/run/current-system/sw/bin/blender() [0x321766e]
/nix/store/4gk773fqcsv4fh2rfkhs9bgfih86fdq8-gcc-13.3.0-lib/lib/libstdc++.so.6(+0xe86d3) [0x7f8d890e86d3]
/nix/store/nqb2ns2d1lahnd5ncwmn6k84qfd7vx2k-glibc-2.40-36/lib/libc.so.6(+0x90d02) [0x7f8d88e97d02]
/nix/store/nqb2ns2d1lahnd5ncwmn6k84qfd7vx2k-glibc-2.40-36/lib/libc.so.6(+0x1103ac) [0x7f8d88f173ac]

# Python backtrace

Please tell me this is something I’m doing wrong and not a bug I have to report upstream and wait for it to be fixed…

1 Like
2

I’m pretty sure you’re not the only one experiencing this. Only difference is that I get to fire 1 or 2 samples before it exits on life.

I’m currently on the unstable blender and I get a pretty similar segfault log, except that I have the /opt/rocm/lib/libamdhip64.so.6 lines:

# Blender 4.4.3, Unknown revision
bpy.context.space_data.context = 'RENDER'  # Property
bpy.context.scene.render.engine = 'CYCLES'  # Property
bpy.context.scene.cycles.device = 'GPU'  # Property

# backtrace
/etc/profiles/per-user/r/bin/blender() [0xf6cabe]
/etc/profiles/per-user/r/bin/blender() [0x736c92]
/nix/store/cg9s562sa33k78m63njfn1rw47dp9z0i-glibc-2.40-66/lib/libc.so.6(+0x41470) [0x7f901d241470]
/opt/rocm/lib/libamdhip64.so.6(+0x43022a) [0x7f8fa6e3022a]
/opt/rocm/lib/libamdhip64.so.6(+0x1e5779) [0x7f8fa6be5779]
/opt/rocm/lib/libamdhip64.so.6(+0x18cd26) [0x7f8fa6b8cd26]
/opt/rocm/lib/libamdhip64.so.6(+0x3381f8) [0x7f8fa6d381f8]
/opt/rocm/lib/libamdhip64.so.6(hipModuleLoadData+0x3f1) [0x7f8fa6cffd01]
/etc/profiles/per-user/r/bin/blender() [0x3c6981d]
/etc/profiles/per-user/r/bin/blender() [0x3abffe5]
/etc/profiles/per-user/r/bin/blender() [0x3c08b33]
/etc/profiles/per-user/r/bin/blender() [0x3c0a1cb]
/etc/profiles/per-user/r/bin/blender() [0x3c0ad40]
/etc/profiles/per-user/r/bin/blender() [0x3c0adff]
/etc/profiles/per-user/r/bin/blender() [0x3f807b2]
/nix/store/7c0v0kbrrdc2cqgisi78jdqxn73n3401-gcc-14.2.1.20250322-lib/lib/libstdc++.so.6(+0xed064) [0x7f901d6ed064]
/nix/store/cg9s562sa33k78m63njfn1rw47dp9z0i-glibc-2.40-66/lib/libc.so.6(+0x97e63) [0x7f901d297e63]
/nix/store/cg9s562sa33k78m63njfn1rw47dp9z0i-glibc-2.40-66/lib/libc.so.6(+0x11bdbc) [0x7f901d31bdbc]

I’m honestly clueless because it was working a version or two ago. Now it won’t. :frowning:

Sometimes I also use Steam for Blender and it still crashes, but some older versions work. :thinking:

Versions that I tested (Steam via Betas selection):

  • :x: 4.5 Alpha/Daily
  • :x: 4.4.3
  • :white_check_mark: 4.3.2
  • :white_check_mark: 4.2.10 LTS Candidate
  • :white_check_mark: 4.2.9 LTS
  • :white_check_mark: 4.1.1
  • :white_check_mark: 4.0.2
  • :white_check_mark: 3.6.22
  • :white_check_mark: 3.3.21 LTS

And I tested these from NixOS directly (nix-shell -p):

  • :grey_question: blender (4.2.3 LTS) No HIP devices, couldn’t test.
  • :white_check_mark: blender-hip(4.2.3 LTS) Blender 4.2.3 LTS.

And these from the unstable channel (nix-shell -p '(import <nixos-unstable> {}).program'):

  • :grey_question: blender (4.4.3), no HIP devices found, couldn’t test.
  • :x: blender-hip (4.4.3). Crashes, with the same errors as above.

During my testing I following all of the advice from AMD GPU - NixOS Wiki, except the overclocking/underclocking advice. Maybe i’ll try it but I don’t see why since it started since it works in other versions :thinking:. This will need investigating. My full NixOS config at the time of testing here.

3

Okay!! So I realized my mistake! I was running a legacy channel (24.11) and not the latest stable (25.05). Changing channels make it work like magic. I no longer get random crashes. I’m posting this in case anyone else finds thing in the future.