first of all, I am relatively new to TrueNAS. I have a few years of general Linux experience, but please excuse me if I get things wrong.
I just got a good deal on a 16GB 5060 Ti. I already know that I probably won’t be able to get it to run with 25.04, so I tried the 25.10 nightly build.
After checking the box for installing NVDIA drivers, the card shows up for use in Apps and I was able to select it for Ollama.
However, when I try nvidia-smi in the TrueNAS as well as in the Ollama app shell, the command returns “no devices were found” (and I am pretty sure that the card does nothing when I used a test workload in Ollama).
Checking var/log/messages, everytime I use nvidia-smi, a line with “NRVM: The NVIDIA GPU installed in this system requires use of the NVIDIA open kernel modules.” appears.
The solution there is to install the 570 MIT/GPL driver. I understand this would not be an easy thing to realize in TrueNAS. Is there anything I can try in the short term or would it be most feasible to just wait for the October release of 25.10?
Thanks for the reply (replies :)), I wasn’t aware that isolating an unsupported GPU was possible. I’ll try that. Indeed I am just testing stuff right now, hence I am using a nightly build.
I am looking to replace an aging Synology NAS - I have already settled on a bunch of used server hardware, and now I am trying to evaluate what I can do with TrueNAS/other software. And one of my hopefully sometime productive use cases is LLM-based OCR for paperless-ngx with paperless-gpt and Ollama (e.g. for barely readable receipts).
@Chomp Thanks for the heads up. What’s the exact nightly build of TrueNAS you’re using? We just bumped from 570.153.02 to 570.172.08 to try to catch the vanilla 5060 support, so I want to see which driver this is on.
I recently upgraded to 25.04.2. I added a 5060Ti to my system as I saw the driver had been updated to 550.153.02 which supports it. However, when checking /var/log/messages, I still see it trying to use the 550.142 driver. Is that a similar issue to what he’s seeing with 25.10?
You are just a victim of bad timing with your purchase/install of that 5060Ti card in your TrueNAS server.
Per the 25.04 (Fangtooth) Version Notes, the version of the nVidia drives that is used by TrueNAS is 550.153.02 which were released in December, 2024. This is prior to the release of the entire nVidia 50xx series of cards.
When I run the command nvidia-smi, I see the following:
root@vaultron[~]# nvidia-smi
Fri Aug 1 19:00:06 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.142 Driver Version: 550.142 CUDA Version: 12.4 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce GTX 1080 Off | 00000000:01:00.0 Off | N/A |
| 0% 38C P8 7W / 200W | 3MiB / 8192MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
I actually acquired and installed a 5060 card on my secondary computer this past May mainly for video encoding purposes and put my 1080 card into my TrueNAS server. That 1080 card will be EOL for driver support in about three years.
Hi @HoneyBadger I have a ZOTAC RTX 5060 Ti 16GB (PCI ID 10DE:2D04, rev a1) on TrueNAS 25.10.2.1, BIOS 1825 (latest), ASUS PRO B760M-CT-CSM, Intel i5-14500.
lspci -k shows Kernel driver in use: nvidia — correct, not vfio-pci.
app.gpu_choices returns the card fully recognized with "error": null and "available_to_host": true.
isolated_gpu_pci_ids is empty.
However nvidia-smi returns “No devices were found” both on host and inside a CUDA container.
dmesg shows GSP firmware timeout at boot:
NVRM: _kgspProcessRpcEvent: Attempted to process RPC event from GPU0: 0x1022 (RECOVERY_ACTION) during bootup without API lock
NVRM: nvAssertOkFailedNoLog: Call timed out [NV_ERR_TIMEOUT] returned from kgspWaitForRmInitDone
NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x65:1859)
Firmware directory only contains gsp_ga10x.bin and gsp_tu10x.bin — no gsp_gb20x.bin.
I see another user on Reddit with the same card and same driver version (570.172.08) has a working nvidia-smi. What could be different? Any ideas?
