OS Security updates is my primary reason for patching.
I’m currently rebuilding OMV in VMWare Workstation on my desktop box and will see how much I hate it or not. Will miss the easier app store vs setting up all my stuff in Portainer, but that’s a one and done.
I am much more comfortable in the long run with OMV and the fact that it uses APT for package management rather than the proprietary junk that TrueNAS is using.
The only reason I upgraded to 25.10.1 (via the “unsupported” installer that puts in legacy NVIDIA support) was to get a slightly newer graphics driver that kept NVENC working properly with the latest release of Handbrake Web - which fixes a number of bugs that annoyed me, and also included the latest Handbrake engine under the hood.
Just for context, there are multiple apps with over 20,000 active deployments, as you can see on https://apps.truenas.com/. Plex, Jellyfin, Immich, and maybe 1 or 2 others are actually over the 30,000 threshold, but we haven’t updated the site badges for that yet. While the apps system isn’t perfect, I think “hundreds” of forum threads is still probably giving a slanted view of its reliability. People don’t tend to start threads to say everything is working fine.
Yeah - apps working just fine here (12 of them). Mostly with host paths.
The problems that I have noticed boil mostly down to permissions with a few no default gateway / DNS issues. There will always be an occaisonal outlier though.
To be fair, permissions do seem to be something of a ballache
So I am considering a 5060Ti for my Immich server hosted on this latest TN. It sounds like this card would be perfectly supported and a great fit for its AI workload. Any suggestions?
Big sigh. I didn’t read this notice either and now all my containers that used GPU are dead. Was running Plex and Immich with a Tesla P4. I can’t upgrade, because basically every GPU is $300+ dollars right now, excluding Intel, but I don’t need that much power for basic transcoding.
I manage GPU clusters all day for work, but lets see how hard they make it to install the driver I need since it’s an “appliance“. Who the heck is even running Blackwell in their NAS? Why not be a normal person and just let your storage be storage and then pull data over NFS to an actual compute node? It’s kinda what every single AI datacenter on the planet is doing.
@HoneyBadger i saw the closed feature request…. the key here is we are not asking ixsystems to ship GRID or Legacy drivers, we are asking for a way to let us install them when we need them
maybe the fix here is to allow us to more easily create sysexts - be it for nvidia, hailo8 drivers, whatever?
Glad to see the Nvidia drivers got updated to support 50xx series cards. I have a RTX 5060TI and while this card shows as supported for the 570 drivers, my PNY version fails to load the GSP firmware which prevents me from using the card with any apps like Frigate.
From my research on this, it seems that the 570 version of drivers isn’t compatible with newer models of the 5060ti and it needs version 575 or higher.
It would be very helpful if we are able to load drivers ourselves otherwise we have to wait until they are included in the next release. Are there any plans on updating the drivers to a later version in the next release?
The 5060 with or without “Ti” should be supported - we’re missing a couple of newer drops like the 5050, RTX PRO 2000 and RTX PRO 5000 72GB, but the 5060 should be covered.
What’s the VID/PID that your card is returning in lspci or the nvidia module init?
As far as major release, 26.04 is currently targeting 580.105.08 - note that the NVENC preset changes in the 590 release are a pretty major hurdle to clear for downstream workloads as well, so we want to make sure all of the commonly used packages/apps have had a chance to adapt before we throw those out there.
Our own @zzzhouuu has already done exactly that, building a custom sysext with the older (non-free) NV module -
The truly “legacy” drivers from the 400-series won’t work with the newer kernel, so that’s a non-starter as it is now and will be even less possible with 26.04 when we switch to the 6.18 base.
With regards to GRID/vGPU, I assume you’ve seen the threads and posts from @ChanningHe about this:
Kernel compatibility is of course still an issue, but for vGPU you also have to ensure that you don’t do anything that will earn you a nasty letter from NVIDIA.
10de:2d04 is listed as a valid pair in the supported card list. Can you paste the error that was shown for the GSP firmware failing to load?
I know some early RTX 5060/Ti cards had a firmware bug that manifested as “black screen until the Windows driver was loaded” on non-UEFI systems, and since the Windows driver doesn’t use the nvidia-open module it wouldn’t depend on GSP, so this might be a case of hardware needing an update. NVIDIA’s got an updater, but of course it requires Windows …
I do have a way to use Windows outside of TrueNAS on the same system with the 5060ti so I can check the firmware. The Nvidia Tool linked said there was no update available and the firmware was already applied. As for the error I am getting, here is way shows up in dmesg at startup.
sudo dmesg | grep NVRM
[ 32.102921] NVRM: loading NVIDIA UNIX Open Kernel Module for x86_64 570.172.08 Release Build (dvs-builder@U22-I3-AF01-21-3) Tue Jul 8 18:08:21 UTC 2025
[ 122.067699] NVRM: _kgspProcessRpcEvent: Attempted to process RPC event from GPU0: 0x1022 (RECOVERY_ACTION) during bootup without API lock
[ 122.067717] NVRM: nvAssertFailedNoLog: Assertion failed: 0 @ kernel_gsp.c:1443
[ 126.065459] NVRM: nvAssertOkFailedNoLog: Assertion failed: Call timed out [NV_ERR_TIMEOUT] (0x00000065) returned from RPC_HDR->rpc_result @ kernel_gsp.c:4751
[ 126.065482] NVRM: nvAssertOkFailedNoLog: Assertion failed: Call timed out [NV_ERR_TIMEOUT] (0x00000065) returned from kgspWaitForRmInitDone(pGpu, pKernelGsp) @ kernel_gsp_gh100.c:928
[ 126.065787] NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
[ 126.067338] NVRM: iovaspaceDestruct_IMPL: 1 left-over mappings in IOVAS 0x100
[ 126.067350] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x65:1859)
[ 126.068513] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
[ 126.108168] NVRM: _kgspBootGspRm: unexpected WPR2 already up, cannot proceed with booting GSP
[ 126.108185] NVRM: _kgspBootGspRm: (the GPU is likely in a bad state and may need to be reset)
[ 126.108206] NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
[ 126.109585] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x40:1859)
[ 126.111173] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
[ 131.013040] NVRM: nvAssertOkFailedNoLog: Assertion failed: Failure: Generic Error [NV_ERR_GENERIC] (0x0000FFFF) returned from RPC_HDR->rpc_result @ kernel_gsp.c:4751
[ 131.013063] NVRM: nvAssertOkFailedNoLog: Assertion failed: Failure: Generic Error [NV_ERR_GENERIC] (0x0000FFFF) returned from kgspWaitForRmInitDone(pGpu, pKernelGsp) @ kernel_gsp_gh100.c:928
[ 131.013139] NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
[ 131.014723] NVRM: iovaspaceDestruct_IMPL: 1 left-over mappings in IOVAS 0x100
[ 131.014736] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0xffff:1859)
[ 131.015876] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
[ 219.468931] NVRM: loading NVIDIA UNIX Open Kernel Module for x86_64 570.172.08 Release Build (dvs-builder@U22-I3-AF01-21-3) Tue Jul 8 18:08:21 UTC 2025
[ 219.552506] NVRM: _kgspBootGspRm: unexpected WPR2 already up, cannot proceed with booting GSP
[ 219.552524] NVRM: _kgspBootGspRm: (the GPU is likely in a bad state and may need to be reset)
[ 219.552542] NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
[ 219.553957] NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x40:1859)
[ 219.555020] NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
[ 219.552506] NVRM: _kgspBootGspRm: unexpected WPR2 already up, cannot proceed with booting GSP
[ 219.552524] NVRM: _kgspBootGspRm: (the GPU is likely in a bad state and may need to be reset)
There’s a couple of internal NVIDIA bug tickets that call out these errors about the GSP being in a weird state and not taking a firmware load. Some are related to laptops with hybrid graphics but I wonder if that’s bleeding over into desktop setups where they aren’t the primary GPU.
Are you able to try out a nightly build of 26.04 just long enough to see if it picks up the GPU? We’re on 580 there but IIRC NVIDIA developed 570/580 in parallel, so if this is something fixed in 575 it might need 585 on that side as well.
Interesting how this problem happens with laptops and hybrid graphics. I am using an Asus B650M-A Prime AX II with a Ryzen 5 7600X and I do have the iGPU enabled with set to primary. I did try disabling the iGPU and boot directly off the RTX 5060TI however this made no change to having the card being usable in TrueNAS.
I also have Resize BAR and Above 4G enabled in the BIOS with UEFI only enabled. Changing these settings also had no change.
As for trying out a nightly build, I don’t feel comfortable doing this on my current setup since it is being used as a production system. I use it for 24/7 video surveillance and it need it running as reliable as possible. If there are other things I can try on the early adopter build or gather more information, I can certainly do that.
Yeah, I can’t repro this with my 5070Ti as a primary or secondary (to a Radeon) GPU unfortunately.
Can you link to any discussions of this? I wasn’t aware of a new hardware or chip revision on the 5060 series but I wouldn’t be surprised if NVIDIA did a sneaky on me.
