You are not logged in.
Hello everyone,
I'm facing a persistent issue with DaVinci Resolve Studio (version 17.4.2) on a vanilla Arch Linux system. The application launches and seems to work, but the video preview (thumbnails in media pool, timeline viewer) shows heavy artifacts/garbage instead of frames. The audio plays fine. I get an error message: "The GPU failed to perform image processing because of an error. Error code: -59".
Hardware:
- GPU: AMD Radeon RX 550 (Polaris12, 2GB)
- CPU: Xeon 3460
- RAM: 16GB
- Kernel: 6.18.16-1-lts
Software & Drivers:
- Mesa 26.0.1-1 (with rusticl)
- OpenCL implementation: rusticl (confirmed working via clinfo)
- DaVinci Resolve Studio 17.4.2 (official .run installer from Blackmagic Design)
What I've done / Configuration:
OpenCL setup: I disabled the conflicting amdocl64.icd (moved to .disabled) and created /etc/OpenCL/vendors/rusticl.icd with libRusticlOpenCL.so.1.
Environment: I launch Resolve with RUSTICL_ENABLE=radeonsi GDK_BACKEND=x11 /opt/resolve/bin/resolve.
Resolve Settings: In Preferences -> System -> Memory and GPU, I've set:
GPU processing mode: OpenCL (manually, not Auto)
GPU selection: My AMD RX 550 is checked (the only device selected).
Troubleshooting steps tried:
Cleared render cache (Playback -> Delete Render Cache -> All).
Reset Resolve configs (renamed ~/.local/share/DaVinciResolve/configs and ~/.config/DaVinciResolve).
Killed any lingering Resolve processes before launch.
Tried RUSTICL_ENABLE=amdgpu and RUSTICL_ENABLE=all.
Tried adding RUSTICL_FEATURES=fp64 and ROC_ENABLE_PRE_VEGA=1 (the latter didn't change anything).
Verified that no other OpenCL implementations are active (clinfo --list shows only rusticl).
Ran with RUSTICL_DEBUG=1 - no additional errors appeared.
Current behavior:
- `clinfo` sees the GPU and reports OpenCL 3.0 capabilities.
- DaVinci Resolve starts, menus and interface are responsive.
- Imported H.264/H.265 files (and other formats) load correctly, and the video playback in the main viewer works fine - the video plays, audio is synced.
- However, the thumbnail previews in the media pool and the timeline viewer are heavily corrupted (artifacts, garbled image).
- The error -59 ("The GPU failed to perform image processing") appears consistently, seemingly related to the generation of these previews.
Relevant logs:
- Full clinfo output: clinfo_full.txt
- dmesg (amdgpu related): dmesg_amdgpu.txt
- glxinfo: glxinfo.txt
- List of installed packages (grep for mesa/opencl): packages.txt
Question:
Is this a known limitation of rusticl on Polaris cards with DaVinci Resolve? Are there any additional environment variables or configuration tweaks that could force proper OpenCL image handling? Or should I consider switching to the proprietary AMD OpenCL stack (like opencl-amd + amdgpu-pro-oglp) for this specific workload?
Any guidance would be greatly appreciated.
Offline
Do you use X11 or Wayland?
Try running without those variables.
Try full system update and try to install amdgpu-*.
Tarty
Offline
Thx for the suggestions.
I've done a full system update (pacman -Syu) and tested thoroughly:
- X11/Wayland: I tried launching Resolve both with GDK_BACKEND=x11and without it (native Wayland). In all cases, the behavior is identical: the main video plays fine, audio is OK, but the thumbnails/timeline previews are garbled, and the -59 error appears.
- Running without variables: I ran /opt/resolve/bin/resolve with no environment variables at all - same issue.
- With RUSTICL_ENABLE=radeonsi (current working OpenCL): same artifacts, same error.
- With RUSTICL_ENABLE=amdgpu or =all: Resolve starts but gives "unsupported GPU processing mode" and then behaves the same.
Could you clarify what you mean by install amdgpu-*? I already have the standard open‑source stack (mesa, xf86-video-amdgpu, linux-firmware). Do you mean the proprietary amdgpu-pro packages, or something else? I want to be sure before installing anything that might break the current setup
Offline
You could try to install opencl-amd from AUR and maybe try to install ocl-icd (pacman). after this install disable rusticl.idc so davinci uses offical amd.
Last edited by TartilupA (2026-03-12 17:15:02)
Tarty
Offline
I tried it: installed ocl-icd and opencl-amd from AUR, disabled rusticl. Now clinfo --list shows "AMD Accelerated Parallel Processing", so the proprietary driver is active.
But now DaVinci Resolve won't start properly - it gives an error:
"Unsupported GPU Processing Mode. Please review the GPU drivers and GPU configuration under preferences."
When I open Preferences from that error, the GPU section is completely empty - no "GPU processing mode" dropdown, no GPU selection list at all.
So the proprietary driver broke it completely. Any idea why Resolve doesn't see the GPU now? Should I try something else?
Offline
try to install rocm-opencl-runtime
Tarty
Offline
tried rocm-opencl-runtime. Now Resolve gives "Unsupported GPU Processing Mode" and GPU preferences empty. Doesn't work lol
Offline
hmm, try https://aur.archlinux.org/packages/open … amdgpu-pro
also try with this envar ROC_ENABLE_PRE_VEGA=1
Edit:
forgot mention, sorry, first make the system "clean" and then install that and also you might try with using that flag, although I dont think with that driver you'll need it, just in case...
Last edited by 5hridhyan (2026-03-12 18:10:07)
---
Offline
I tried the AUR package opencl-amdgpu-pro with ROC_ENABLE_PRE_VEGA=1.
Now the behavior changed:
1. After launch, I got: "GPU Configuration Reset. DaVinci Resolve has detected and reset an invalid GPU configuration. Please review the GPU drivers and GPU configuration under preferences."
2. I clicked through, and for the first time the GPU settings appeared in Preferences. I selected OpenCL and applied.
3. After restarting Resolve, it now crashes immediately during component loading.
Logs:
- Crash log (fresh coredump from today, PID 1290152 at 21:11): https://pastebin.com/Ha3EXfeY
- dmesg (amdgpu related, still relevant): https://pastebin.com/gfU3az9U
Last edited by InTeaReable (2026-03-12 18:18:22)
Offline
just a hunch
env QT_QPA_PLATFORM=xcb LD_PRELOAD=/usr/lib/libcrypt.so.1 ROC_ENABLE_PRE_VEGA=1 /opt/resolve/bin/resolveLast edited by 5hridhyan (2026-03-12 18:18:33)
---
Offline
I tried the command, but it still crashes. Here's the terminal output:
env QT_QPA_PLATFORM=xcb LD_PRELOAD=/usr/lib/libcrypt.so.1 ROC_ENABLE_PRE_VEGA=1 /opt/resolve/bin/resolve
ActCCMessage Already in Table: Code= c005, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c006, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c007, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= 2282, Mode= 0, Level= 0, CmdKey= 8, Option= 0
PnlMsgActionStringAdapter Already in Table: Code= 615e, Mode= 0, Level= 0, CmdKey= -1, Option= 0
log4cxx: No appender could be found for logger (BtCommon).
log4cxx: Please initialize the log4cxx system properly.
17.4.2.0009 Linux/Clang x86_64
Main thread starts: AF332580
0x7f07af332580 | Undefined | INFO | 2026-03-13 01:19:34,187 | --------------------------------------------------------------------------------
0x7f07af332580 | Undefined | INFO | 2026-03-13 01:19:34,187 | Loaded log config from /home/intea/.local/share/DaVinciResolve/configs/log-conf.xml
0x7f07af332580 | Undefined | INFO | 2026-03-13 01:19:34,187 | --------------------------------------------------------------------------------It seems to crash right after the log initialization.
Last edited by InTeaReable (2026-03-12 18:34:12)
Offline
...
can you please check with tail or strace?, to see exactly what is happening? like do you see "illegal instruction" or "OpenCL" etc??
---
Offline
I ran the strace as you asked. The last lines of the log show all child processes exiting with code 0, but the main Resolve process still crashes silently after printing the log messages. Here's the tail output:
216055 +++ exited with 0 +++
216054 +++ exited with 0 +++
216053 +++ exited with 0 +++
216052 +++ exited with 0 +++
216051 +++ exited with 0 +++
216050 +++ exited with 0 +++
216049 +++ exited with 0 +++
216048 +++ exited with 0 +++
216047 +++ exited with 0 +++
216046 +++ exited with 0 +++
216045 +++ exited with 0 +++
216044 +++ exited with 0 +++
216043 +++ exited with 0 +++
216042 +++ exited with 0 +++
216041 +++ exited with 0 +++
216040 +++ exited with 0 +++
216039 +++ exited with 0 +++
216038 <... futex resumed>) = ?
216037 +++ exited with 0 +++
216036 +++ exited with 0 +++
216035 +++ exited with 0 +++
216034 +++ exited with 0 +++
216028 +++ exited with 0 +++
216007 +++ exited with 0 +++
215993 +++ exited with 0 +++
215992 +++ exited with 0 +++
215991 +++ exited with 0 +++
215990 +++ exited with 0 +++
215989 +++ exited with 0 +++
215988 +++ exited with 0 +++
215987 +++ exited with 0 +++
215986 +++ exited with 0 +++
215985 +++ exited with 0 +++
215984 +++ exited with 0 +++
215983 +++ exited with 0 +++
215982 +++ exited with 0 +++
215981 +++ exited with 0 +++
216238 +++ exited with 0 +++
216235 +++ exited with 0 +++
216231 +++ exited with 0 +++
216223 +++ exited with 0 +++
216221 +++ exited with 0 +++
216219 +++ exited with 0 +++
216215 +++ exited with 0 +++
216213 +++ exited with 0 +++
216195 +++ exited with 0 +++
216194 +++ exited with 0 +++
216170 +++ exited with 0 +++
216038 +++ exited with 0 +++
215044 +++ exited with 0 +++No "illegal instruction" or OpenCL errors visible. The GUI window appears briefly, then it crashes during component loading. Should I try a different strace filter, like tracing signals (-e trace=signal)? Or maybe run it under gdb to get a backtrace?
Offline
thank you for the logs, but, wait lets see what is it actually seeing
/opt/resolve/bin/resolve --list-gpusLast edited by 5hridhyan (2026-03-12 18:45:10)
---
Offline
I ran /opt/resolve/bin/resolve --list-gpus as you asked. It crashes immediately with this output:
╰─? /opt/resolve/bin/resolve --list-gpus
ActCCMessage Already in Table: Code= c005, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c006, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= c007, Mode= 13, Level= 1, CmdKey= -1, Option= 0
ActCCMessage Already in Table: Code= 2282, Mode= 0, Level= 0, CmdKey= 8, Option= 0
PnlMsgActionStringAdapter Already in Table: Code= 615e, Mode= 0, Level= 0, CmdKey= -1, Option= 0
log4cxx: No appender could be found for logger (BtCommon).
log4cxx: Please initialize the log4cxx system properly.
17.4.2.0009 Linux/Clang x86_64
Main thread starts: 8CEC1580
Failed to load config file "--list-gpus".qt.qpa.plugin: Could not find the Qt platform plugin "wayland" in ""
resolve: /home/jenkins/jenkins/workspace/resolve@3/Resolve/Cyclone/Main/AppConfig.cpp:204: void AppConfig::LoadAllSiteInfo(): Assertion `m_SiteEnabledIdx > 0' failed.
==========[CRASH DUMP]==========
#TIME Fri Mar 13 01:46:36 2026 - Uptime 00:00:00 (hh:mm:ss)
#PROGRAM_NAME DaVinci Resolve Studio v17.4.2.0009 (Linux/Clang x86_64)
#BMD_ARCHITECTURE x86_64
#BMD_BUILD_UUID 0503b81e-313b-4db5-b10a-5463615a8e97
#BMD_GIT_COMMIT b0c9793b2f99a38d83804a46e65a7c5d5be948e0
#BMD_UTIL_VERSION 17.4.2.0009
#OS Linux
/opt/resolve/bin/resolve() [0x4123f29]
/opt/resolve/bin/resolve() [0x2148272]
/usr/lib/libc.so.6(+0x3e2d0) [0x7f4b8da4d2d0]
/usr/lib/libc.so.6(+0x98a2c) [0x7f4b8daa7a2c]
/usr/lib/libc.so.6(gsignal+0x20) [0x7f4b8da4d1a0]
/usr/lib/libc.so.6(abort+0x26) [0x7f4b8da345fe]
/usr/lib/libc.so.6(+0x26697) [0x7f4b8da35697]
/usr/lib/libc.so.6(__assert_fail+0xc2) [0x7f4b8da450b2]
/opt/resolve/bin/resolve() [0x217175d]
/opt/resolve/bin/resolve() [0x216fa68]
/opt/resolve/bin/resolve() [0x216f378]
/opt/resolve/bin/resolve() [0x3aefc48]
/opt/resolve/bin/resolve() [0x35a3a3d]
/opt/resolve/bin/resolve() [0x35a38ce]
/opt/resolve/bin/resolve() [0x218e826]
/opt/resolve/bin/resolve() [0x218e7a0]
/opt/resolve/bin/resolve() [0x2146542]
/opt/resolve/bin/resolve() [0x21424bf]
/usr/lib/libc.so.6(+0x276c1) [0x7f4b8da366c1]
/usr/lib/libc.so.6(__libc_start_main+0x89) [0x7f4b8da367f9]
/opt/resolve/bin/resolve() [0x214153b]
Signal Number = 6
================================
resolve: /home/jenkins/jenkins/workspace/resolve@3/Resolve/Cyclone/Main/AppConfig.cpp:204: void AppConfig::LoadAllSiteInfo(): Assertion `m_SiteEnabledIdx > 0' failed.
zsh: IOT instruction (core dumped) /opt/resolve/bin/resolve --list-gpusOffline
Mod note: Moving to AUR issues.
Sakura:-
Mobo: MSI MAG X570S TORPEDO MAX // Processor: AMD Ryzen 9 5950X @4.9GHz // GFX: AMD Radeon RX 5700 XT // RAM: 32GB (4x 8GB) Corsair DDR4 (@ 3000MHz) // Storage: 1x 3TB HDD, 6x 1TB SSD, 2x 120GB SSD, 1x 275GB M2 SSD
Making lemonade from lemons since 2015.
Offline
do you try downgrade rocm-opencl-runtime and hsa-rocr from 7.2.x to 7.1.1 yet ?
Just a tentacle in the pit.
Offline
I downgraded rocm-opencl-runtime and hsa-rocr to 7.1.1 as suggested.
Now clinfo -l shows only the platform, no devices.
ROC_ENABLE_PRE_VEGA=1 or HSA_OVERRIDE_GFX_VERSION do not help.
libamdocl64.so is present in /opt/rocm/lib and the ICD file points to it.
The other options (rusticl -> preview artifacts, opencl-amdgpu-pro -> crash) are not working.
Offline