Failed to use GeForce GT1030 on Dell PowerEdge T110 with Debian 12 and other distributions

Graphics / Linux Linux
1

Hi,

I recently got an old Dell PowerEdge T110 (from year 2011) which has an onboard Matrox G200eW card.
To gain better video performances I bought a used GigaByte card with an nVidia GeForce GT1030 gpu, but I’m not able to get this one working on Linux. Specifically I’m using Debian 12, but I also tried with Xubuntu 24.04 and Ubuntu 18.04 (so kernels 5.x and 6.x have been tested) with the same results.

Basically:

lspci sees the cards:

01:00.0 VGA compatible controller: NVIDIA Corporation GP108 [GeForce GT 1030] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GP108 High Definition Audio Controller (rev a1)
[...]
06:03.0 VGA compatible controller: Matrox Electronics Systems Ltd. MGA G200eW WPCM450 (rev 0a)

but dmesg | grep -i nouveau reports:

[    1.956882] nouveau 0000:01:00.0: enabling device (0000 -> 0003)
[    1.957077] nouveau 0000:01:00.0: NVIDIA GP108 (138000a1)
[    2.074994] nouveau 0000:01:00.0: bios: version 86.08.24.00.15
[    2.075189] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/nvdec/scrubber.bin
[    2.075219] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/acr/bl.bin
[    2.075248] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/acr/ucode_load.bin
[    2.075278] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/acr/unload_bl.bin
[    2.075299] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/acr/ucode_unload.bin
[    2.075305] nouveau 0000:01:00.0: pmu: firmware unavailable
[    2.075557] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/fecs_bl.bin
[    2.075594] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/fecs_inst.bin
[    2.075618] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/fecs_data.bin
[    2.075635] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/fecs_sig.bin
[    2.075667] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/gpccs_bl.bin
[    2.075689] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/gpccs_inst.bin
[    2.075711] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/gpccs_data.bin
[    2.075729] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/gpccs_sig.bin
[    2.075755] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/sw_nonctx.bin
[    2.075780] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/sw_ctx.bin
[    2.075804] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/sw_bundle_init.bin
[    2.075833] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/gr/sw_method_init.bin
[    2.075876] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/sec2/sig.bin
[    2.075958] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/sec2/image.bin
[    2.076034] nouveau 0000:01:00.0: firmware: direct-loading firmware nvidia/gp108/sec2/desc.bin
[    2.212801] nouveau 0000:01:00.0: fb: 2048 MiB GDDR5
[    2.257234] nouveau 0000:01:00.0: bar: one-time init failed, -12
[    2.259020] nouveau 0000:01:00.0: init failed with -12
[    2.259043] nouveau: DRM-master:00000000:00000080: init failed with -12
[    2.260154] nouveau 0000:01:00.0: DRM-master: Device allocation failed: -12
[    2.260616] nouveau: probe of 0000:01:00.0 failed with error -12

while using the proprietary nVidia driver (I tried from the Debian repository and from the nVidia website, I also tried with 535 / 470 / 450) gave similar dmesg results with a lot of

RmInitAdapterFailed!

lines.

I read dozens of web pages and tried a lot of combinations, like:

  • making sure that no SecureBoot was on,
  • making sure to use the old BIOS in place of the UEFI,
  • temporarily disabling the onboard Matrox card,
  • temporarily changing the IRQ for the nVidia card,
  • temporarily booting with nomodeset GRUB kernel option, and also with other combinations of parameters I found in many other pages,
  • installing the driver from the website with and without the -m=kernel-open mode, with and without the options nvidia NVreg_OpenRmEnableUnsupportedGpus=1 line put in a specific /etc/modprobe.d/ configuration file,
    and something else I fail to remember right now :)

I don’t think the card itself is “broken”, because disabling the onboard Matrox card I was able to use the nVidia one in basic mode (1024x768 no acceleration etc.) with the basic X driver.

Before uploading other log information, I’d like to ask You one thing please.
I was going to blame the drivers, the kernel or a combination of boths, but I came across this page which is about a slightly different server version (T110 II, mine is T110) and reports about power limitations in the PCI bus.
Can it be that the card hasn’t got enough power? From what I read, every PCI bus is limited to 25W but I also read that the card needs 30W or more…
Or can it be that Dell made some trickery on the motherboard to prevent something from working as it should?

Thanks for any hint :)

MZ