Test EVGA Nvidia RTX 3080 Ti on Raspberry Pi 5

[geerlingguy]

I have tested the 3080 Ti on my Raspberry Pi 5 using Raspberry Pi's prototype PCIe adapter. It required an external power supply to feed that adapter and the card itself, and I had to install Nvidia's latest arm64 driver for the 3080 Ti, but after all that, I got the dreaded RmInitAdapter failed! message.

[geerlingguy]

Testing with nouveau now...

pi@pi5:~ $ dmesg | grep nouveau
[    2.938104] nouveau 0000:01:00.0: enabling device (0000 -> 0002)
[    2.938329] nouveau 0000:01:00.0: NVIDIA GA102 (b72000a1)
[    3.404851] nouveau 0000:01:00.0: bios: version 94.02.71.80.74
[    3.426761] nouveau 0000:01:00.0: bios: M0203E type 0a
[    3.426780] nouveau 0000:01:00.0: fb: 12288 MiB of unknown memory type
[    3.461589] nouveau 0000:01:00.0: DRM: VRAM: 12288 MiB
[    3.461602] nouveau 0000:01:00.0: DRM: GART: 536870912 MiB
[    3.461609] nouveau 0000:01:00.0: DRM: BIT table 'A' not found
[    3.461613] nouveau 0000:01:00.0: DRM: BIT table 'L' not found
[    3.461617] nouveau 0000:01:00.0: DRM: TMDS table version 2.0
[    3.461621] nouveau 0000:01:00.0: DRM: DCB version 4.1
[    3.461625] nouveau 0000:01:00.0: DRM: DCB outp 00: 04800f86 04600020
[    3.461630] nouveau 0000:01:00.0: DRM: DCB outp 01: 04000f82 00020020
[    3.461634] nouveau 0000:01:00.0: DRM: DCB outp 02: 04811f76 04600010
[    3.461637] nouveau 0000:01:00.0: DRM: DCB outp 03: 04011f72 00020010
[    3.461641] nouveau 0000:01:00.0: DRM: DCB outp 04: 02822f66 04600020
[    3.461645] nouveau 0000:01:00.0: DRM: DCB outp 05: 02022f62 00020020
[    3.461650] nouveau 0000:01:00.0: DRM: DCB outp 07: 02033f52 00020010
[    3.461655] nouveau 0000:01:00.0: DRM: DCB conn 00: 02000046
[    3.461658] nouveau 0000:01:00.0: DRM: DCB conn 01: 01000146
[    3.461661] nouveau 0000:01:00.0: DRM: DCB conn 02: 00020246
[    3.461665] nouveau 0000:01:00.0: DRM: DCB conn 03: 00010361
[    3.461748] nouveau 0000:01:00.0: DRM: failed to initialise sync subsystem, -22
[    3.495783] nouveau: probe of 0000:01:00.0 failed with error -22

[geerlingguy]

Trying proprietary driver again. First disabling nouveau by creating a file /etc/modprobe.d/blacklist-nouveau.conf with the content:

blacklist nouveau
options nouveau modeset=0

Then after a reboot, it was not loading the nouveau module.

May I interject debugging is SO much more fun when the whole system doesn't freeze every time requiring me to reset power?

[geerlingguy]

Downloading the Nvidia driver from https://www.nvidia.com/en-us/geforce/drivers/results/214101/

wget https://us.download.nvidia.com/XFree86/aarch64/545.29.02/NVIDIA-Linux-aarch64-545.29.02.run
chmod +x NVIDIA-Linux-aarch64-545.29.02.run
sudo ./NVIDIA-Linux-aarch64-545.29.02.run

After installation was complete, I rebooted, and dmesg shows:

[    3.756735] nvidia: loading out-of-tree module taints kernel.
[    3.756757] nvidia: module license 'NVIDIA' taints kernel.
[    3.756760] Disabling lock debugging due to kernel taint
[    3.840532] nvidia-nvlink: Nvlink Core is being initialized, major device number 509

[    3.845903] nvidia 0000:01:00.0: enabling device (0000 -> 0002)
[    3.845947] nvidia 0000:01:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
[    3.890053] NVRM: loading NVIDIA UNIX aarch64 Kernel Module  545.29.02  Thu Oct 26 21:50:44 UTC 2023
[    3.918561] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  545.29.02  Thu Oct 26 20:50:54 UTC 2023
[    3.925431] [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
[    3.925439] [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 2
[    6.148023] NVRM: Chipset not recognized (vendor ID 0x14e4, device ID 0x2712)
[    6.148030] The NVIDIA GPU driver for AArch64 has not been qualified on this platform
               and therefore it is not recommended or intended for use in any production
               environment.
[   10.688237] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x25:0x65:1470)
[   10.688306] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
[   14.949123] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x25:0x65:1470)
[   14.949180] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

That fun, and dreaded RmInitAdapter failed!.

[geerlingguy]

See possibly related issue: NVIDIA/open-gpu-kernel-modules#514

I need to test with the open-gpu-kernel-modules bits.

(The quickest way to uninstall Nvidia's proprietary driver is sudo nvidia-uninstall.)

(Also, the 3080 Ti gets HOT.)

[Radiicall]

Have you tested with the new nouveau vulkan driver? It might have more luck initializing than the normal nouveau gl driver (I dont really know but its just an idea)