Open prototype5885 opened 1 year ago
To make sure, could you try updating NVIDIA Flatpak driver?
flatpak update
To make sure, could you try updating NVIDIA Flatpak driver?
flatpak update
that's not the issue, I have the latest version and even having the same issue right after fresh Debian 12 install too
To make sure, could you try updating NVIDIA Flatpak driver?
flatpak update
that's not the issue, I have the latest version and even having the same issue right after fresh Debian 12 install too
It is just that sometime host and flatpak driver version miss match and cause weird problems. Can you confirm if both driver version are the same?
Also, non-flatpak blender is from Debian repo or binary from blender.org?
To make sure, could you try updating NVIDIA Flatpak driver?
flatpak update
that's not the issue, I have the latest version and even having the same issue right after fresh Debian 12 install too
It is just that sometime host and flatpak driver version miss match and cause weird problems. Can you confirm if both driver version are the same?
Also, non-flatpak blender is from Debian repo or binary from blender.org?
versions are nvidia-525-105-17 for both, also that blender is from blender.org
Try
flatpak run --device=all org.blender.Blender --verbose 5
Maybe there is more info
flatpak run --device=all org.blender.Blender --verbose 5
i tried this yesterday but i dont think it was much more helpful, here is the result:
flatpak run --device=all org.blender.Blender --verbose 5 Read prefs: "/home/prototype585/.var/app/org.blender.Blender/config/blender/3.6/config/userpref.blend" Failed to open dir (No such file or directory): /run/user/1000/gvfs/ I0722 13:18:55.837390 2 device.cpp:32] CUEW initialization succeeded I0722 13:18:55.837424 2 device.cpp:34] Found precompiled kernels CUDA cuInit: Unknown error
generally flatpak uses gpu just fine as godot and steam games run without issue, it obviously seems to have some issues loading cuda/optix
edit: as a side note there has been a debian update recently upgrading the nvidia driver to 525.125.06, did the same update on flatpak to make the versions mach, but the same issue as before stays
it seems like the flatpak version of blender wont initialize cuda, as i installed cuda toolkit + pycuda, and after running the "import pycuda.autoinit" command in python, the flatpak version of blender sees the gpu now,
but after restarting the system obviously the issue persists
What is your flatpak version?
flatpak --version
What is your flatpak version?
flatpak --version
1.14.4
default one from Debian 12 repo
I have the same version as well. What is your GPU model?
I have the same version as well. What is your GPU model?
MSI RTX 2070 Ventus
As a workaround i just compiled an executable using nvidia cuda compiler that prints "hello world" using cuda and running it when the system starts, so cuda gets used at least once, because without this flatpak blender refuses to use cuda
I have no clue. I could not replicate your issue and the error message is not helpful.
@kevinsmia1939 Hello! I have the exactly same issue.
I'm using Debian 12, Nvidia RTX 2070, Driver Version: 525.125.06, CUDA Version: 12.0. The CUDA driver is the official Debian version.
After a fresh reboot, flatpak Blender cannot detect my GPU, so cannot enable CUDA or OptiX in Cycles. But after starting once the downloaded version of Blender from blender.org (through blender-3.6.5-linux-x64/blender-launcher
which can use CUDA and OptiX normally), the flatpak Blender can also detect the GPU then.
The following is the version information of the downloaded Blender. The flatpak Blender is also 3.6.5 and have the same build hash.
.local/opt/blender-3.6.5-linux-x64/blender-launcher --version
Blender 3.6.5
build date: 2023-10-18
build time: 23:31:25
build commit date: 2023-10-16
build commit time: 14:30
build hash: cf1e1ed46b7e
build platform: Linux
build type: release
build c flags: -Wall -Werror=implicit-function-declaration -Werror=return-type -Werror=vla -Wstrict-prototypes -Wmissing-prototypes -Wno-char-subscripts -Wno-unknown-pragmas -Wpointer-arith -Wunused-parameter -Wwrite-strings -Wlogical-op -Wundef -Winit-self -Wmissing-include-dirs -Wno-div-by-zero -Wtype-limits -Wformat-signedness -Wrestrict -Wno-stringop-overread -Wno-stringop-overflow -Wnonnull -Wabsolute-value -Wuninitialized -Wredundant-decls -Wshadow -Wno-error=unused-but-set-variable -Wimplicit-fallthrough=5 -std=gnu11 -msse -pipe -fPIC -funsigned-char -fno-strict-aliasing -ffp-contract=off -msse2
build c++ flags: -Wuninitialized -Wredundant-decls -Wall -Wno-invalid-offsetof -Wno-sign-compare -Wlogical-op -Winit-self -Wmissing-include-dirs -Wno-div-by-zero -Wtype-limits -Werror=return-type -Wno-char-subscripts -Wno-unknown-pragmas -Wpointer-arith -Wunused-parameter -Wwrite-strings -Wundef -Wformat-signedness -Wrestrict -Wno-suggest-override -Wuninitialized -Wno-stringop-overread -Wno-stringop-overflow -Wundef -Wmissing-declarations -Wimplicit-fallthrough=5 -msse -pipe -fPIC -funsigned-char -fno-strict-aliasing -ffp-contract=off -msse2
build link flags: -Wl,--version-script='/home/blender/git/blender-v360/blender.git/source/creator/symbols_unix.map'
build system: CMake
terminal says:
CUDA cuInit: Unknown error
But strangely if I start non flatpak Blender (which sees nvidia just fine), after that the flatpak version will see the gpu too, but after restarting the PC the issue happens again