54 minutes ago, Un1kk said:

That is, if I understand you correctly, keeping a static version before and after updating the system is not achievable

No, please don't get me wrong but I can't compile all the various driver version for all Unraid versions since the compilation takes about 6-10 minutes for each Nvidia driver, actually not the compilation is the part that takes the longest it's actually archiving the driver because it's semi multi core so to speak most of the time single core. This is really time consuming and please do keep in mind that in the worst case there will be four drivers be compiled for each Unraid version.

Please also don't forget that some drivers even do not compile against newer Kernel versions.

As @alturismo already said, if you are choosing a driver form the list that is listed below the latest, prduction/new feature/open source branch you should be good to go as long as the selected driver version is also available for the new Unraid version so to speak in your case for both versions 570.144 is a version that is available for both your linked versions:

58 minutes ago, Un1kk said:

https://github.com/unraid/unraid-nvidia-driver/releases/tag/6.6.78-Unraid

https://github.com/unraid/unraid-nvidia-driver/releases/tag/6.12.24-Unraid

Please do keep in mind if you are compiling your drivers on your own you have to come up with your own solution how to install them since the driver plugin could support it but you have to set the version in the configuration file manually, place the files manually and on top you also have to place the new files before the reboot in a specific directory.

However, wouldn't it be better that you simply write a bash script that checks the installed driver from the host or keeps track of the releases in the GitHub repository for the current running Kernel version and makes then sure to install the latest or the driver that you've possibly selected, even easier you could mount a symlink into the container from the Nvidia Driver settings file into your LXC container and from there you could determine which driver version is installed compare it against the current running Kernel version and upgrade the driver inside the container <- this would be way easier in my opinion.

1 hour ago, Rkpaxam said:

Ok so i think i found the issue, whilst my unraid is whitlisted in my AGH instance it seems to be having issues randomly so i have gone back to the data harvesters that are google and all is well.

Oh okay, so it was something with DNS or similar?

Anyways, glad that you've figured it out. ;)

17 minutes ago, ich777 said:

No, please don't get me wrong but I can't compile all the various driver version for all Unraid versions since the compilation takes about 6-10 minutes for each Nvidia driver

No, I didn't misunderstand, I made it clear from the start that I understood that and appreciated your efforts in the process.

Please let me express my claim in a different way. We always offer 5 versions now, and as far as I can see, the version numbers of all five options are rolling over quickly as the system is upgraded. But wouldn't it be nice if we could keep the version of one of the options stable (in my opinion, Production Branch would be a good fit for this, in line with its name), so we don't really need to spend extra time, and we can satisfy the demands of users who are looking for stability!

22 minutes ago, ich777 said:

Please also don't forget that some drivers even do not compile against newer Kernel versions.

Of course, it's perfectly understandable to push for a driver version number update in this case

23 minutes ago, ich777 said:

570.144 is a version that is available for both your linked versions

I see your point, but those version numbers I cited were just an example.

In other words, even if I chose 570.144 to maintain cross-version compatibility in the moment, that version would quickly become obsolete(no longer available) in future updates, So I'll be forced to update the next driver version again. wouldn't it?

26 minutes ago, ich777 said:

Please do keep in mind if you are compiling your drivers on your own you have to come up with your own solution how to install them since the driver plugin

know I'll have to figure this out on my own if I want to stick with one driver version for a long time. However, I'd like this to be part of the official solution if you can recognize my claim.

Edited June 6, 20251 yr by Un1kk

1 hour ago, alturismo said:

have some "spare" versions always which you can use, and meanwhile update your apps before the next major jump comes

I understand that, but the point I was trying to make is that I found these spare versions are still becoming obsolete too quickly. Sooner or later these options will no longer be avaliable when I update my Unraid system with several versions, and the system will in turn automatically choose a new driver version for me.

I'd like to have at least one of them really available for a long period of time so that I don't have to risk breaking the development environment by being forced to do frequent driver updates.

Edited June 6, 20251 yr by Un1kk

11 minutes ago, Un1kk said:

but the point I was trying to make is that these spare versions are still becoming obsolete too quickly, and I'd like to have at least one of them available for a long period

understood, honestly never followed how fast the cycle actually is ...

nevertheless, its a "either or" i guess, i cant imagine there will be any long term availability for "old" driver versions while there are new stable ones available, my way would be to keep my applications on track too, but as i dont know your workflow its hard to guess.

point is matched and clear, but as there are way more questions about "when will be the latest" ... and only some questions about very old legacy drivers for old cards, and when i think about none before this one im aware off to keep some driver for actual hardware longer available ... there is no real demand and usecases in a wide range, just my personal oppinion when i rethink about the questions here in this thread in the last years.

1 minute ago, alturismo said:

honestly never followed how fast the cycle actually is

Maybe if you open this link you'll see how fast it's been updated：

https://github.com/unraid/unraid-nvidia-driver/releases

In my opinion, maintaining a stable development environment is not an uncommon claim, as developers usually don't want to spend too much time configuring their environment. Nvidia is committed to providing 3 years support for their Long Term Support Branch (LTSB). So even though newer versions are available, there are still a lot of users using older versions of drivers

It seems to me that Unraid is updating kernel versions so aggressively because they assume that users won't be doing anything on bare systems and that everything required is done using Docker, VMs, and LXC. They assume that changes to the underlying system won't easily affect the upper tier applications. This was true to a certain extent until I realized that Nvidia Driver was also discarding old versions at such a rapid rate that I had to update my Usperspace driver, CUDA, CUDNN, Pytorch software stack, and destroy my longtime work environment in a accident.

Edited June 6, 20251 yr by Un1kk

3 hours ago, ich777 said:

Oh okay, so it was something with DNS or similar?

Anyways, glad that you've figured it out. ;)

Well thats my working theory anyway i have yet to see if im correct. Its odd as its been fine for months and months so i suspect its somthing on the backend of AGH

17 hours ago, Un1kk said:

It seems to me that Unraid is updating kernel versions so aggressively because they assume that users won't be doing anything on bare systems and that everything required is done using Docker, VMs, and LXC.

This is the root of the issue. Since the beginning of the year, there have been 6 releases of Unraid 7, each (I believe) with a different kernel. For those who wish for stability, upgrading once a month is not a winning strategy.

My main server is still on 6.12.10. I always stay away from software that ends in .0 (and x.0.0 is even worse). I was hoping that 7.1 would finally have most all of the bugs worked out, but it looks as 7.1.3 is still a mess. I really do wish to upgrade this server, but I will just watch from the sidelines until LT straightens out the mess.

Hi all,

I am having issues using this plugin and retaining a "connection" to the GPU. The GPU attached is an RTX 3060 (was a 1660 Super -> same issue) and is using driver version 575.57.08 on Unraid 7.1.2. The GPU is being passed through to Jellyfin in docker for transcoding and is not utilized anywhere else. Within 24h of the system being booted if a transcode attempts to occur, my syslog will be filled with:

Jun 12 01:54:01 Fennec kernel: NVRM: GPU 0000:0c:00.0: RmInitAdapter failed! (0x62:0x51:2584)

Jun 12 01:54:01 Fennec kernel: NVRM: GPU 0000:0c:00.0: rm_init_adapter failed, device minor number 0

This will cause my transcodes to not start and the media will fail to play. Things I have tried include:

1. Swapping the RAM

2. Downgrading to 7.0.1, 6.12.15, or 6.12.14 (Last OS I used before the issue started)

3. Downgrading the driver to version 570.153.02 or 565.77

4. Swapping the GPU (same issue occurs on a 1660 Super)

5. Moving the GPU to the secondary PCIE slot

6. Swapping the motherboard (B450 -> B550)

7. Changing the CPU (5600G -> 5800x)

I have not tried recreating the Unraid flash yet, although I am not opposed to doing this. Just though I would see if anyone can see anything in the logs. I took a look through the forum but the only other issues with the keywords "RmInitAdapter failed" were for individuals with a faulty GPU and the message was present on boot. This has only started in the last month and everything was fine up until then. Not to say that the GPU didn't go bad, but the 3060 was known working to test with and both still work consistently until the system has some uptime.

Diagnostics

Thanks

11 hours ago, tykateetee said:

I took a look through the forum but the only other issues with the keywords "RmInitAdapter failed" were for individuals with a faulty GPU and the message was present on boot.

Please disable C States in your BIOS and male sure that your BIOS is up to date.

2 hours ago, ich777 said:

Please disable C States in your BIOS and male sure that your BIOS is up to date.

Both of those changes have been applied so now I wait and see. Interestingly enough the B450 board that the server was originally running on did have its latest BIOS revision applied. It did, however, not have C States disabled. If you don't hear from me again, I'll go with C States being the issue.

Hi,

first of all: huge thanks to ich777 for providing everything in the context of the nvidia driver plugin, I wish I would be able to do the same/support you in doing so, but unfortunately I'm just a dangerous half knowledged semi-noob, so, sadly that will not work. -

Question for (better) understanding: there are mainly two different kinds of drivers, proprietary and open source, for GPUs like mine with Blackwell architecture, a RTX 5060Ti, I need to use the open source version to get at least a partly usable graphics card (recognized as generic NVIDIA Graphics Device). I already saw the next version 570.153.02 on github for kernel 6.12.30. Afaik things remain much or less the same? For full support of my card/GPU it would be necessary to advance the plugin to >575.51.02, but for this, a complete rewrite of the plugin would be necessary? I would like to use the GPU for Steam headless, right now most titles are unplayable bcs of very low fps, maybe bcs of partly usable 5060Ti? Thank you in advance

Chris

26 minutes ago, anticholium said:

but for this, a complete rewrite of the plugin would be necessary

No.

27 minutes ago, anticholium said:

575.51.02

On what Unraid version are you, please always provide Diagnostics if you need something in terms of a driver.

21 hours ago, ich777 said:

No.

On what Unraid version are you, please always provide Diagnostics if you need something in terms of a driver.

I just concluded because of this comment (see below), but was asking in my last post if I got right. Obviously no...

21 hours ago, ich777 said:

No.

On what Unraid version are you, please always provide Diagnostics if you need something in terms of a driver.

I'm on Unraid 7.1.3, using the latest Opensource Nvidia Driver (MSI Geforce RTX 5060Ti), latest Steam Headless Container and I'm trying to fire up some more up to date gameing titles and was hoping to streamline them with Sunshine/Moonlight to the more weak laptops of my son. Sadly, despite the quite powerful GPU, I simply can achieve just around 8-10fps even for older games like Counter Strike 2; hardware acceleration is enabled, the container is using the GPU (nvtop, nvidia-smi)... In other use cases like Handbrake video transcoding (NVENC AV1, 10bit, 30fps, CQ30) it works very well, for LLM-based text-toaudiobook-"conversion" it works pretty well too. So my thoghts were, that the current open source driver might only be partyl supporting the quite actual GPU gen? Do you know, when there might be a change in this? I will perform a diagnostics when using the scenario as described and will attach the zip, if that might be helpful. Thank you so much!

On 6/12/2025 at 9:47 PM, ich777 said:

Please disable C States in your BIOS and male sure that your BIOS is up to date.

This unfortunately did not work. The "RmInitAdapter failed!" message is still present in the syslog.

I'm open to any other suggestions that you or others may have.

New Diagnostics

Thanks

2 hours ago, anticholium said:

So my thoghts were, that the current open source driver might only be partyl supporting the quite actual GPU gen?

Nope, there is no partly support for GPUs it‘s either working or not.

Possibly Sunshine/Moonlight is the coolprit?

1 hour ago, tykateetee said:

I'm open to any other suggestions that you or others may have.

Initially transcoding is working or isn‘t it working at all?

At times I get reports about AMD CPUs/Motherboards having issues with Nvidia especially after a bit after initial boot.

What you can try is to also set the PCIe gen to 3.0 in your BIOS, make also sure you are really on the latest BIOS version.

Make sure to also disable ASPM for PCIe and all power saving features.

1 hour ago, ich777 said:

Initially transcoding is working or isn‘t it working at all?

At times I get reports about AMD CPUs/Motherboards having issues with Nvidia especially after a bit after initial boot.

What you can try is to also set the PCIe gen to 3.0 in your BIOS, make also sure you are really on the latest BIOS version.

Make sure to also disable ASPM for PCIe and all power saving features.

Everything works fine when the system boots. This includes transcoding. I have also notices that the error occurs in the syslog randomly, even when there is no transcode running. It seemingly just happens at random. It has happened twice more since the above diagnostics with the system being idle.

Your second point describes the scenario well. I don't have an Intel system to test with, although it is strange that it has now happened with 2 seperate AMD CPU/Motherboard combinations.

Setting the link speed to gen 3 does not solve the issue, as the B450 board is only gen 3 and it still had the issue. The latest BIOS images were downloaded from here and here. They both were up to date when the error occured. There is no PCIe riser involved.

All power saving methods are disabled in the BIOS. The issue is still present. It is also should be noted that with a "stock" bios and only DOCP enabled these issues did not occur for the better part of 2 years.

Thanks for you help so far. Any other thoughts are appreciated.

6 hours ago, tykateetee said:

I have also notices that the error occurs in the syslog randomly, even when there is no transcode running.

You have probably the Dashboard from Unraid open when you notice this, correct? GPU Statistics is also querying the state from the GPU and that‘s also what could produce that error.

7 hours ago, tykateetee said:

although it is strange that it has now happened with 2 seperate AMD CPU/Motherboard combinations.

AMD does funky stuff with PCIe at times and that could be the reason, however it could be also be the case that something is wrong with the card however I think that it has to do with the CPU/Motherboard.

7 hours ago, tykateetee said:

Setting the link speed to gen 3 does not solve the issue, as the B450 board is only gen 3 and it still had the issue.

This was just a refommendation since it helped other people in the past by setting it to a specific generation and not leaving it on Auto.

7 hours ago, tykateetee said:

DOCP

I would always recommend leaving everything at default in terms of overclocking and not enable overclocking or at least set the memory to it‘s non overclocked speed since stability is preferred on a Server.

One last thing that could cause this is the power supply if it’s not up to the task anymore.

Otherwise I‘m out of ideas, on my AMD test system I had to disable CPU C-States because it would also constanly randomly crash.

53 minutes ago, ich777 said:

You have probably the Dashboard from Unraid open when you notice this, correct? GPU Statistics is also querying the state from the GPU and that‘s also what could produce that error.

AMD does funky stuff with PCIe at times and that could be the reason, however it could be also be the case that something is wrong with the card however I think that it has to do with the CPU/Motherboard.

This was just a refommendation since it helped other people in the past by setting it to a specific generation and not leaving it on Auto.

I would always recommend leaving everything at default in terms of overclocking and not enable overclocking or at least set the memory to it‘s non overclocked speed since stability is preferred on a Server.

One last thing that could cause this is the power supply if it’s not up to the task anymore.

Otherwise I‘m out of ideas, on my AMD test system I had to disable CPU C-States because it would also constanly randomly crash.

The dashboard is open, and GPU Statistics is installed. This error still occured before I installed GPU Statistics for the first time. The error did show up one time at 3AM and I don't think a dashboard was open at that time, but I could be mistaken there.

Good to know that sometimes it is just funky. I really thought a complete hardware switch would fix it but as you mention the issue could be relating to AMD CPUs. I do have 2 other Unraid instances with AMD CPUs + Nvidia GPUs, and this includes one that has the exact same motherboard + CPU combo, just with a 2060 Super instead. I haven't needed to perform any trickery with power states on them (thankfully) and have never seen this issue.

Totally agree with the link speed tip - there was that and more I found reading through the forums to no avail.

While I agree with stock for important systems, my Unraid instance is primarily just for Jellyfin. Plus I have 3600Mhz RAM, so I just want to use 3600Mhz RAM :). I did try a week ago at stock frequency and the issue was still present.

The power supply is a 1200w unit from Asus. I don't think it's bad but hey - you never know.

Realistically, I'm going to reinstall from scratch tomorrow. It just seems like one of those things that is not worth the time to try to troubleshoot anymore and the OS is the last thing I have left to change.

Thank you for the support. I'll probably update here soon regarding if blowing everything up and starting over fixed it. While that tip won't be that helpful to most, it would at least rule out (for my situation at least) if it was funky hardware.

2 hours ago, tykateetee said:

The dashboard is open, and GPU Statistics is installed. This error still occured before I installed GPU Statistics for the first time. The error did show up one time at 3AM and I don't think a dashboard was open at that time, but I could be mistaken there.

But transcoding is working for some time after initial boot correct?

The error usually appears when the card can't be initialised after standby or initially on boot, the latter is mostly a firmware issue from the Motherboard.

10 hours ago, ich777 said:

But transcoding is working for some time after initial boot correct?

The error usually appears when the card can't be initialised after standby or initially on boot, the latter is mostly a firmware issue from the Motherboard.

Yes, transcoding is working for some time after initial boot. Even after the error occurs the card will re-attach and transcoding will work again.

I have seen that the error normally occurs when the card can't be initialised after standby or initially on boot. I do not have standby enabled and the error is never present on boot.

36 minutes ago, tykateetee said:

Even after the error occurs the card will re-attach and transcoding will work again.

That's weird, that shouldn't work at all...

2 minutes ago, ich777 said:

That's weird, that shouldn't work at all...

Welcome to my last month XD. I've thrown in some screenshots to show that I'm not fully crazy.

Just now, tykateetee said:

Welcome to my last month XD. I've thrown in some screenshots to show that I'm not fully crazy.

May I ask to what interval you've set the refresh from GPU Statistics?

However this seems like a firmware issue to me...