Installing Titan Xp on Raspberry Pi 5 via PCIe board, getting RmInitAdapter failed

Graphics / Linux Linux
1

Hi there! I’m trying to assemble the following system for a barebones, network-accessible GPU:

  • Raspberry Pi 5 8GB
  • NVIDIA Titan Xp GPU
  • GeeekPi P02 PCIe Slot for Raspberry Pi 5

I’ve installed the latest Raspberry Pi 64-bit OS Lite in headless mode (Linux pi5 6.6.74+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.6.74-1+rpt1 (2025-01-27) aarch64 GNU/Linux) + drivers (NVIDIA-Linux-aarch64-550.144.03.run). Build and install of drivers ran flawlessly. Rebooting I get the following:

rbates@pi5:~/tmp $ lspci
0000:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries BCM2712 PCIe Bridge (rev 21)
0000:01:00.0 VGA compatible controller: NVIDIA Corporation GP102 [TITAN Xp] (rev a1)
0000:01:00.1 Audio device: NVIDIA Corporation GP102 HDMI Audio Controller (rev a1)
0001:00:00.0 PCI bridge: Broadcom Inc. and subsidiaries BCM2712 PCIe Bridge (rev 21)
0001:01:00.0 Ethernet controller: Raspberry Pi Ltd RP1 PCIe 2.0 South Bridge
rbates@pi5:~/tmp $ lshw -c display
WARNING: you should run this program as super-user.
  *-display
       description: VGA compatible controller
       product: GP102 [TITAN Xp]
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:01:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: vga_controller bus_master cap_list rom
       configuration: driver=nvidia latency=0
       resources: iomemory:40-3f iomemory:40-3f irq:38 memory:1b80000000-1b80ffffff memory:1800000000-180fffffff memory:1810000000-1811ffffff memory:1b81000000-1b8107ffff
  *-graphics
       product: vc4drmfb
       physical id: 9
       logical name: /dev/fb0
       capabilities: fb
       configuration: depth=16 resolution=1920,1080

So it appears the driver is installed and associated with the GPU, but nvidia-smi shows no devices available and dmesg logs report:

NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x31:0xffff:2640)

Logs + debug output attached. Any help would be greatly appreciated!

nvidia-bug-report.log.gz (68.3 KB)
nvidia-installer.log (43.6 KB)

2

Looking at your debug log:

Feb 17 22:40:56 pi5 kernel: [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
Feb 17 22:40:56 pi5 kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 2
Feb 17 22:41:29 pi5 kernel: NVRM: Chipset not recognized (vendor ID 0x14e4, device ID 0x2712)
Feb 17 22:41:33 pi5 kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x31:0xffff:2640)
Feb 17 22:41:33 pi5 kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

it seems like the driver has an issue with recognising the Broadcom, (vendor ID 14E4) PCI Bridge.

Possibly a lack of imagination around installing Nvidia GPU’s on Pi’s, on the part of the driver devs.

3

Thanks for the insights! Do you think I might have better luck trying to use one of the OSS drivers, where at least maybe I could get support for the Pi’s PCI bridge added…?

4

Could be worth a try, although anything based on Nvidia’s Open GPU Kernel Modules only supports GPU’s from Turing onwards.

You could try filing a bug, although it’s unlikely to have a high priority.

5

Curious if any success has been had with pre-Turing cards on broadcom platforms. Were you able to find a work-around? Have the more recent driver/kernel packages been able to resolve the issues?

Some open kernel modules were forked for Aarch64 and we’re shown to work for CUDA, NVenc, and general 3D processing on the Pi5. Fingers crossed the requisite changes are upstreamed to the proprietary 580.xx driver branch before they finally kill it off (or that nvidia releases the source, since nonGSP processors are no longer being produced)