[Plugin] Nvidia-Driver


ich777

Recommended Posts

Hmm, okay so the direct host of the card needs a whole X environment right?, because the DebianBuster-Nvidia should also contain a X environment.

 

I wanted to set a much higher memory clock for mining, yes and also use the card for transcoding if needed. Anyway if it is not possible I will try HiveOS then, thank you for the hint :D. 

But I loved the idea that the miner is running within a container next to plex etc.

Link to comment
53 minutes ago, CK98 said:

Hmm, okay so the direct host of the card needs a whole X environment right?, because the DebianBuster-Nvidia should also contain a X environment.

It's actually a little more complicated but in really basic terms, yes.

 

54 minutes ago, CK98 said:

I wanted to set a much higher memory clock for mining, yes and also use the card for transcoding if needed.

That would be really difficult to achieve when you hand it over to a VM.

May I ask what kind of processor that you have? If it's a Intel one maybe the iGPU is a good choice.

 

55 minutes ago, CK98 said:

But I loved the idea that the miner is running within a container next to plex etc.

Yep, but that's really difficult to achieve on unRAID and would need further investigation (not the easy kind)...

Link to comment
7 hours ago, hahler2 said:

Just wondering if I can install this in my Unraid server and use it for transcodes with Plex and Tdarr?

You can, but keep in mind to use only the drivers up to the v470 series.

 

7 hours ago, hahler2 said:

Or is it too old a GPU to be any use for that?

This card is only capable of transcoding h264 and can't be used for h265 transcoding.

 

Something like a Nvidia T400 will do the job better, more efficient and you can get them new for about $ 130,-

Link to comment
47 minutes ago, letrain said:

It appears on 6.9.2 the Nvidia plugin breaks gpu passthrough to windows 10 vm.

Why did you install this plugin if you want to pass through your GPU to a VM (-> see first post bold text below the headline).

 

If you have two GPUs in your system and you want to pass one through to the VM and use one for Docker bind the one that you want to use in VMs to VFIO, that should do the trick.

 

Please post your Diagnostics for further help.

Link to comment

One for docker. Two for vm. My amd rx580 and 2070Suped are bound vfio. And I want my p2000 for docker. If I start without plugin vm boots fine- then if I install the plugin and just turn Dockers off and on it works fine. If I reboot however docker works fine with Nvidia,but then vm won't boot. The VM for some reason pins a few CPU cores to 100% allocates ram and stops booting windows and just sits there. It's an odd occurrence. I've done this before on a different setup (hp z800). Every since I switched to the current setup it's been finicky. 

tower-diagnostics-20211207-1138.zip

 

 

Sorry about the screenshot. Wasn't too sure how to show a VM that hangs as the log had no errors. And do I screenshot with the plugin installed? Or not? And with no plugin , no nvidia-smi. :) I can get some screenshots if you still need them. 

Edited by letrain
Admitting I didn't follow directions.
Link to comment
27 minutes ago, letrain said:

And I want my p2000 for docker.

May I ask why don't you use the iGPU for transcoding, in terms of quality it is a different world to NVENC (Pascal) and I think in terms of the parallel streams it is almost the same if I'm not mistaken.

 

30 minutes ago, letrain said:

My amd rx580 and 2070Suped are bound vfio.

Can you try to switch over from Q35 to i440fx and see if this makes a difference?

I also would recommend that you don't use a BIOS file for the Nvidia cards because Nvidia now "officially" "supports" consumer cards in VMs.

 

31 minutes ago, letrain said:

The VM for some reason pins a few CPU cores to 100% allocates ram and stops booting windows and just sits there.

I've had the same behavior on the box from my little son, it almost took two minutes to even see a output on the screen and the VM actually started, but I've also heard that on some systems the VM doesn't start regardless of how long they wait.

Link to comment
4 hours ago, ich777 said:

May I ask why don't you use the iGPU for transcoding, in terms of quality it is a different world to NVENC (Pascal) and I think in terms of the parallel streams it is almost the same if I'm not mistaken.

 

Can you try to switch over from Q35 to i440fx and see if this makes a difference?

I also would recommend that you don't use a BIOS file for the Nvidia cards because Nvidia now "officially" "supports" consumer cards in VMs.

 

I've had the same behavior on the box from my little son, it almost took two minutes to even see a output on the screen and the VM actually started, but I've also heard that on some systems the VM doesn't start regardless of how long they wait.

 

The igpu is the reason I upgraded my system from a z800. I have a lot of 4k hevc and had weird color issues when I first set this pc up using igpu. I'm trying it again now. I do transcode a bit of 4k because of tablets on the go for the kids. It does seem to handle it. The p2000 was out performing the igpu when I got the system. It seems to be ok now. Honestly I've been mining eth in trex miner docker for a while as it's just sitting there anyways might as well put it work. It was designed to be under load hours on end and it does well. I have no other use for the p2000 other than to set it to mining. 

 

I had lots of difficulty with the windows vm. I tried both. A bunch. Different versions etc. And couldn't get the card recognized in windows when it booted and couldn't get any drivers to install for amd or Nvidia. I actually put the SSD in another computer got windows updated, drivers, all working then put it all back and couldn't get it to boot up. Removed the plugin and the VM instantly booted. 

 

I'll try without the bios and try i440fx again and report back. If it boots I'll install your plugin and then see if it still boots. 

 

Yeah sometimes it pins a core 100% then after 1 minutes drops down and is fine. But with the plugin installed it just pinned and sat there for 10 minutes. 

 

It's always amazed me how 1 minute everything works fine (had this vm setup before I moved) and you don't touch anything, or change anything... Then all of a sudden it just all stops cooperating. 

Link to comment

Hi guys!

I'm having some trouble getting my new Quadro P2000 to show up properly. It shows up in Tools > System Devices as

[10de:1c30] 0a:00.0 VGA compatible controller: NVIDIA Corporation GP106GL [Quadro P2000] (rev a1)

but nowhere else. When I attempt to open Settings > Nvidia Driver, the page never loads. If I try to run the command nvidia-smi as instructed at the top of this post, nothing happens at all. The terminal just freezes. I've uninstalled and reinstalled the Nvidia Driver plugin. Here is my plugins page:image.thumb.png.1586312026120c569b3aeb9a91c036e6.png

 

And attached is my diagnostics file. Please advise, thanks!

 

Edit: One other thing that is probably important - I noticed once I installed the card into the HP Proliant DL380 G8 server, it stopped using the onboard VGA output for the local CLI. Maybe the machine itself is taking control of the GPU, but I'm not sure how to prevent that.

counterpart-diagnostics-20211207-1951.zip

Edited by poofysnoops
Clarification
Link to comment
7 hours ago, letrain said:

It's always amazed me how 1 minute everything works fine (had this vm setup before I moved) and you don't touch anything, or change anything... Then all of a sudden it just all stops cooperating. 

Yes I know, this is most of the time a real pain...

 

7 hours ago, letrain said:

I'll try without the bios and try i440fx again and report back. If it boots I'll install your plugin and then see if it still boots. 

What you can also do is to try and enable:

grafik.png.0eb879ae9972d6bf26970b06a63751d6.png

 

Had a user with a AMD system which has had also a weird issue, and when enabling Downstream it all began to work.

 

Oh I forgot do you boot in UEFI or Legacy? I would strongly recommend to boot with Legacy Mode and if you are already booting with Legacy Mode, try UEFI instead.

Link to comment
6 hours ago, poofysnoops said:

And attached is my diagnostics file. Please advise, thanks!

You have a Kernel Panic in your logs, I think it's caused when opening the plugin page (the plugin page calls nvidia-smi once to get everything necessary).

 

Is it possible that you try the card in a different system? If yes, please also install the drivers and hit it with a 3D load if possible.

Does the card get enough power?

 

From your Diagnostics file it looks like that you are booting with UEFI, can you try to boot with Legacy Mode please?

Link to comment
16 hours ago, ich777 said:

What you can also do is to try and enable:

grafik.png.0eb879ae9972d6bf26970b06a63751d6.png

 

Had a user with a AMD system which has had also a weird issue, and when enabling Downstream it all began to work.

 

Oh I forgot do you boot in UEFI or Legacy? I would strongly recommend to boot with Legacy Mode and if you are already booting with Legacy Mode, try UEFI instead.

 

I tried every which way. I turned on pcie acs both, vfio interrupts, i440fx, q35, multiple versions. Finally I got them both to boot to the same vm once with no plugin. Once.. Then I tried each card individually. That seemed to work best as they would both boot individually, but not together. Started turning things off in vm settings to nail down what I actually needed to have on, off, and settings. Eventually ended up at downstream, since they are both the same immou group I thought it best if I was planning on separate vms; q35, hyper v off...etc.. and the Nvidia plugin is fine, with my p2000 happily mining away. Happy to also report the 2070s is booting fine and restarting without rom file, I thought I saw in the forum unRAID didn't support yet. Only time I've had to use rom is when in testing I had to force stop vm and it wouldn't boot vm without it.

 

Then my amd vm started having issues, random lockups, having to force shutdown... Then I found your discussion here about rx cards....

 

I'm wondering if this reset bug has been my issue the whole time. As I said I had this working with both gpus in a VM from March to August this year and your plugin working. Then after moving it was a no go all of a sudden.. amazing how a driver update, windows update (I'm blaming windows 98%) breaks everything. 

 

I'm kinda at a loss on where exactly the issue was/is.  Not sure how your plugin affects me trying to boot both Nvidia and amd in the same vm.

 

I'm considering upgrading to rc1 as I don't really want to mess with the kernel. And use the reset plugin. I used the Nvidia modified unRAID version until your plugin came around. I prefer the plugin method for ease of use, and less hassle when upgrading versions, also troubleshooting is easier to remove and add.

 

I'm booting legacy mode as that was the preferred method for unRAID if you wanted gpu passthrough (and my previous system didn't boot uefi USB), and stub the pcie.

 

Every system and motherboard can have different things that make it unique. After passing through various gpus, pcie devices, and various vms I finally did some research on my HP z800 and found out that each CPU managed different pcie lanes and slots and completely explain why some vms ran like crap, or had issue with initializing cards. I'm sure this setup has its own as well. I'm going to keep playing around and see if I can get it all flawless here. I want it to be you bless operation and right now it's not. I'll try your suggestions. 

 

Thank you for the efforts and suggestions. 

  • Like 1
Link to comment
7 hours ago, letrain said:

Then I tried each card individually. That seemed to work best as they would both boot individually, but not together.

That seems like a issue with the Motherboard/BIOS that something is not properly implemented.

Have you turned on RBAR and/or above 4G Decoding in the BIOS?

I think I don't have to ask but anyways... Are you on the latest BIOS version? :)

 

7 hours ago, letrain said:

Happy to also report the 2070s is booting fine and restarting without rom file, I thought I saw in the forum unRAID didn't support yet. Only time I've had to use rom is when in testing I had to force stop vm and it wouldn't boot vm without it.

Yes, Nvidia "supports" this now "officially": Click

 

7 hours ago, letrain said:

I'm considering upgrading to rc1 as I don't really want to mess with the kernel.

I would also recommend to upgrade to RC2 since I'm on it since day one and have no problems whatsoever and I will also drop the Kernel Helper since you can add nearly everything with plugins.

 

7 hours ago, letrain said:

And use the reset plugin.

Please keep in mind the Vendor Reset Patch is just a workaround and will/can fail after n number of reboots from the VM.

 

7 hours ago, letrain said:

I'm booting legacy mode as that was the preferred method for unRAID if you wanted gpu passthrough (and my previous system didn't boot uefi USB), and stub the pcie.

Maybe try to boot into UEFI mode because then allocation of the memory regions and talking to the hardware works a little different than with CSM.

  • Like 1
Link to comment

Hello everyone,

 

I have a Quadro P5000 in my tower but the driver report it as a Quadro RTX 4000. Is there anyway to fix this? It still work fine, I'm just wondering if it's reporting other incorrect data since it's not the right card.

 

On another note, I saw some post up that you say iGPU transcoding does a better job then NVENC? Unfortunatly, I don't have iGPU on my Xeon but I was wondering where this is from cause I have that P5000 specifically for that to be used in emby with 4k hdr transcoding. I was doing it on a poor little i3-530 before, which is really struggling today.

 

 

Thank you!

Link to comment
28 minutes ago, Nodiaque said:

I have a Quadro P5000 in my tower but the driver report it as a Quadro RTX 4000. Is there anyway to fix this?

Please post your Diagnostics for further help with that.

That said, the Quadro P5000 (Pascal based) and RTX4000 (Turing based) should be different and maybe there is an error in the driver but there is nothing I can do about that.

 

28 minutes ago, Nodiaque said:

On another note, I saw some post up that you say iGPU transcoding does a better job then NVENC?

I don't want to start a war about this here, but in my opinion the Intel iGPUs do a way better job in terms of quality than NVENC (especially when transcoding something like Ski races or Formula 1 where the camera or better speaking the scenery moves quickly).

Also the Intel iGPU is very power efficient since it consumes way less power than a dGPU.

But keep in mind the Turing cards have a "newer" or better speaking improved NVENC that should do in terms of quality a similar job like the Intel iGPUs.

 

Haven't got my hands on a card with the new NVENC yet, since my Nvidia T400 has also the "old" Pascal based NVENC.

 

But in terms of simultaneous transcodes your P5000 (most users have a P2000) will outperform the Intel iGPU: Click

 

But that always depends on the input and output, for example see this thread:

 

28 minutes ago, Nodiaque said:

I was doing it on a poor little i3-530 before, which is really struggling today.

If you are interested here is also a overview for the feature set from each Intel iGPU: Click

Link to comment

hello, i have a problem with the nvidia plugin but i dont understand why. the problem is after a time working with two gpus on a docker trex-miner, one of the gpus just disaper from the plugin. i verify and the gpu shows on system devices but dont show on the plugin. if i restart the server a have both gpus on the plugin, but after a few days i get the same problem again.

what could be?

 

my gpus are gtx1060 and gtx1080 ti.

Link to comment

Thanks ich for the explanation. It wasn't a war I tried to start, really just understanding. In a perfect world, I wouldn't have transcoding and most of what I watch on my TV is direct play anyway. Transcoding will be mostly for mobile viewing and it's people normally watching 720p tv show thus a 4k re-encode shouldn't bother them to much.

 

As for my P5000, well, I'm sorry about the false reporting. my order included a P5000 and I have a RTX 4000, I'll have to go back to my seller.... My mistake

  • Like 1
Link to comment

Hi. I'm having a "sort of" issue and was looking for assistance. I've searched through the forums but haven't found this particular issue. I have a P400 Nvidia card installed along with both plugins "Nvidia Driver" and "Prometheus nvidia-smi Exporter". The plugins are working fine I think, as I can see and use the GPU within Emby. However on the Unraid DASHBOARD the GPU status area never shows any info. Even when Emby is actively encoding with the GPU. Any ideas?

 

Thank you,

 

Bobby

Link to comment
1 hour ago, luixmod said:

hello, i have a problem with the nvidia plugin but i dont understand why. the problem is after a time working with two gpus on a docker trex-miner, one of the gpus just disaper from the plugin. i verify and the gpu shows on system devices but dont show on the plugin. if i restart the server a have both gpus on the plugin, but after a few days i get the same problem again.

what could be?

Please share your Diagnostics after a fresh reboot and after the issue occurs.

Link to comment
17 minutes ago, BobbyDing said:

However on the Unraid DASHBOARD the GPU status area never shows any info. Even when Emby is actively encoding with the GPU. Any ideas?

I think the best would be to head over to the GPU Statistics support thread, but have you configured the GPU Statistics plugin so that it actually knows that it has to display the Nvidia statistics?

The GPU Statistics plugin can be used for AMD, Intel or Nvidia devices and has to be configured once.

Link to comment
1 hour ago, Nodiaque said:

It wasn't a war I tried to start

I think I wrote this a little wrong... Some people say that NVENC (up to Pascal) is the same as Intel QSV but it is not in my opinion and if they read this sentence the war starts... :D

 

1 hour ago, Nodiaque said:

Transcoding will be mostly for mobile viewing and it's people normally watching 720p tv show thus a 4k re-encode shouldn't bother them to much.

Then I think a Intel iGPU should get you covered too, but keep in mind that some Xeon Processors and even some low power Core i-Series are missing the QSV capabilities.

Link to comment
2 hours ago, ich777 said:

Then I think a Intel iGPU should get you covered too, but keep in mind that some Xeon Processors and even some low power Core i-Series are missing the QSV capabilities.

Ah but my Xeon doesn't have iGPU :( Bah, that NVENC will do the job anyway. Thanks again and sorry for the wrong error

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.