Very common hardware, really must be something wrong with your build tool chain. Checked elfutils? What’s the output of
ldd /usr/bin/eu-objdump
I appreciate all the help so far!
I am using the standard compiler and linker setup as it comes with openSUSE Tumbleweed. Right now I am on kernel 5.0.7-1 with gcc/g+±8.3.1
# ldd /usr/bin/eu-objdump
linux-vdso.so.1 (0x00007fff117f5000)
libasm.so.1 => /usr/lib64/libasm.so.1 (0x00007fd8262f9000)
libelf.so.1 => /usr/lib64/libelf.so.1 (0x00007fd8262de000)
libdl.so.2 => /lib64/libdl.so.2 (0x00007fd8262d9000)
libc.so.6 => /lib64/libc.so.6 (0x00007fd826114000)
libdw.so.1 => /usr/lib64/libdw.so.1 (0x00007fd8260c0000)
/lib64/ld-linux-x86-64.so.2 (0x00007fd826346000)
libz.so.1 => /lib64/libz.so.1 (0x00007fd8260a6000)
liblzma.so.5 => /usr/lib64/liblzma.so.5 (0x00007fd82606d000)
libbz2.so.1 => /usr/lib64/libbz2.so.1 (0x00007fd82604d000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fd82602c000)
Kernel 5.1.4-1 - NVidia driver 430.14-15.1
Still the same problem.
[url]https://build.opensuse.org/package/show/X11:Drivers:Video/nvidia-gfxG04[/url]
mentions this happening when using gold instead of standard ld.
That was right on the money!
foxhome:~ # update-alternatives --set ld /usr/bin/ld.bfd
update-alternatives: using /usr/bin/ld.bfd to provide /usr/bin/ld (ld) in manual mode
foxhome:~ # modprobe nvidia_uvm
foxhome:~ # dmesg | tail -n1
[42912.079816] nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 240
I still get no CUDA in Blender, but maybe I need to restart the system for that. I’ll give you an update once I did that.
Many many thanks for sticking with me! At least the module loads now, and it might just be Blender not supporting the 1660Ti yet. :-)
Should work with the right Blender and CUDA versions:
[url]https://www.phoronix.com/scan.php?page=news_item&px=Blender-Turing-CUDA-10-Support[/url]
One reboot later and it works again. Thanks so much for keeping an eye out for the solution!