Nvidia 5060 Not Working While Using Goldeye 25.10.1

TrueNAS General
, ,
1

Hi all,

I recently changed my update train to Goldeye from Fangtooth, and I understand that Goldeye is the answer for running 5000-series Nvidia GPUs. I have a 5060 that I would like to use to mess arouund for a personal LLM. However, I can’t seem to get Truenas (Scale) to recognize it.

nvidia-smi accurately displays my 5060, as well as driver 570.172.08, but it doesn’t show up in any apps as a GPU that I can pass through, nor does it show up as an option to isolate. I have ‘download Nvidia drivers’ checked in app settings, and I have unchecked and rechecked and restarted a number of times.

I’m wondering if I am missing something obvious here? Or does Goldeye not fully support 5000-series GPUs yet? I’ve tried to see if anyone else is having this problem, but I haven’t found a solution.

Ideas welcome.

2

Maybe check your revision? Is this the issue?

3

If @inertiax1000 can see their GPU in nvidia-smi then it shouldn’t be the same fault as in the other thread where GSP firmware wouldn’t load, because the driver won’t even init in that case.

Can I get a sudo lspci -k | grep -A4 -i nvidia please?

1 Like
4

Here is the output:

01:00.0 VGA compatible controller: NVIDIA Corporation Device 2d04 (rev a1)
Subsystem: ZOTAC International (MCO) Ltd. Device 1772
Kernel driver in use: vfio-pci
Kernel modules: nouveau, nvidia_drm, nvidia
01:00.1 Audio device: NVIDIA Corporation Device 22eb (rev a1)
Subsystem: NVIDIA Corporation Device 0000
Kernel driver in use: vfio-pci
Kernel modules: snd_hda_intel
02:00.0 Non-Volatile memory controller: Micron Technology Inc 2550 NVMe SSD (DRAM-less) (rev 01)
Subsystem: Micron Technology Inc 2550 NVMe SSD (DRAM-less)

07:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] (rev a1)
Subsystem: ASUSTeK Computer Inc. GP107 [GeForce GTX 1050 Ti]
Kernel modules: nouveau, nvidia_drm, nvidia
07:00.1 Audio device: NVIDIA Corporation GP107GL High Definition Audio Controller (rev a1)
Subsystem: ASUSTeK Computer Inc. GP107GL High Definition Audio Controller
Kernel driver in use: snd_hda_intel
Kernel modules: snd_hda_intel
08:00.0 Non-Volatile memory controller: Kingston Technology Company, Inc. NV2 NVMe SSD [SM2267XT] (DRAM-less) (rev 03)

5

Somehow not only is the passthrough vfio-pci driver still getting bound (you did uncheck it for isolation and reboot, yes?) but it’s also trying to bind nouveau which I was pretty sure we explicitly blacklist.

You’ve also got a 1050Ti in there but that one shouldn’t be getting attached to anything.

1 Like
6

Yeah I am a little confused as to what’s going on. Any suggestions to try?

7

Check System → Advanced → Kernel and make sure you aren’t somehow running the debug kernel.

8

My setting:

Enable Debug Kernel:Disabled

9

well there goes the easy fix

The nouveau driver should be way out of the way as it’s blacklisted from loading in our modprobe directory, so something’s up entirely outside the domain of that.

Did you install your 5060 before or after the upgrade from 25.04 to 25.10, and can you attach the file /var/log/nvidia-installer.log here in a reply?

10

I installed it before the upgrade. Perhaps uninstall and reinstall now that I have changed trains? Standby for the log.

11

I’m thinking something in 25.04 decided to fuse it to vfio-pci for whatever reason and that’s stuck in the kernel params.

Once you get the installer.log let’s see these (with sudo as needed)

midclt call app.gpu_choices | jq
midclt call system.advanced.config | jq
12

I can’t seem to find the .log file… I was just using shell to look for it, am I looking in the wrong spot?

For the other commands:

“midclt call app.gpu_choices | jq” returned: {}

“midclt call system.advanced.config | jq" returned:

{
“id”: 1,
“advancedmode”: false,
“autotune”: false,
“kdump_enabled”: false,
“boot_scrub”: 7,
“consolemenu”: true,
“consolemsg”: false,
“debugkernel”: false,
“fqdn_syslog”: false,
“motd”: “Welcome to TrueNAS”,
“login_banner”: “”,
“powerdaemon”: false,
“serialconsole”: false,
“serialport”: “ttyS0”,
“anonstats_token”: “”,
“serialspeed”: “9600”,
“overprovision”: null,
“traceback”: true,
“uploadcrash”: true,
“anonstats”: true,
“sed_user”: “USER”,
“sysloglevel”: “F_INFO”,
“syslogservers”: ,
“syslog_audit”: false,
“isolated_gpu_pci_ids”: ,
“kernel_extra_options”: “”
}

13

You’ve got a value of

“isolated_gpu_pci_ids”: [],

So the system shouldn’t be forcibly isolating anything.

Re: the logfile, that was where it was on 25.04 - and I haven’t used an NVIDIA since going 25.10 - but you’re right in that the file doesn’t appear to be there. What does your dmesg | grep -i nvidia say as far as if it’s attempting to load the closed vs open driver? :thinking:

14

@HoneyBadger

I got

dmesg: read kernel buffer failed: Operation not permitted

When I run sudo it just returns nothing.

15

Nothing at all in dmesg for nvidia is definitely unusual and/or wrong. At the very least it should be attempting to load and bailing out because it’s either already locked out by vfio-pci or nouveau.

I assume you’ve tried a longer cycle of “uncheck Install NVIDIA drivers, then reboot, then recheck install NVIDIA drivers”?

16

Yeah I have done the whole uncheck, reboot, recheck, reboot thing a few times.

I’m down to try anything at this point though.