mikotoiii

I'm running a similar setup (zen2 AMD) and I have 2 gpus; however, I'm passing 1 gpu to my windows vm and 1 gpu to my Linux system, and running my unraid in headless mode and managing it from another device (usually my phone or my laptop).

For the video card you should be able to pass it through to another VM, you will have to run your unraid headless, and manage it from a vm, or from another device. This quide is a little bit old, so some things might have changed a little, but here's how to change the size of a vm disk:

Glad to hear you came up with a solution.

Do you have 1912:0015 bound to vfio-pci as well? You're other screenshots that show both cards have it listed as not bound to vfio-pci. Is 1912:0015 what you're trying to run unraid from, and 1912:0014 what you're passing through to the vm? Can you follow the attached guide and make sure that all the ports on both cards aren't on the same busses, etc? It looks like 1912:0015 could be running on bus 7 and bus 8 from previous screenshots. Also if you can replicate the card going missing, you can run a dmesg | tail on your server via ssh, which might give you some more information into the cause of the problem. You can also capture diagnostic logs from tools->diagnostics before the reboot happens.

You should be able to change your graphics card from the passthrough to vnc. You may need to create a new vm template and point it to your vm file(s). I like to keep duplicate templates one for my passthrough and one for vnc for troubleshooting. I use an nvme drive passed through for my vms, so blasting away a template is fine since my os resides on a physical device. It also give me the option to boot my os as bare metal should the need arise.

I checked the manual and can't see anything that would cause your issue. I am wondering if it might have something to do with the breakout cable that you have. The only other thing that I saw was the it lists that server as capable of 2 video. https://www.dell.com/support/manuals/is/en/isdhs1/precision-t5810-workstation/precision_t5810_om_pub/technical-specifications?guid=guid-cb1a5aa6-1e70-44b9-b690-59507a3a9f31&lang=en-us For testing purposes, can you remove one of the video cards and plug the usb controller directly into the slot? I'd remove 1 of the video cards first, and then try to pass it though. If the behavior is the same, then try to plug the usb controller directly in. That will help you isolate if that extension is causing the issues.

Just to update folks, I did get the USB and the Audio to successfully passthrough to my Windows VM on my Threadripper 3970x setup. After sitting on it for the weekend, I came back to it fresh and made sure that I was only passing through the USB controller and adding the audio as a 2nd card (my Nvidia for GPU pasthrough is my 1st).

If you can, check the motherboard manual. It could be that your pci bus is sharing the same bus as the pci-e 2x slot. If you post up the model of your Dell server i can see if I can find anything.

Are you passing 1912:0014 or 1912:0015 through to the vm that's crashing. If it's 1912:0014 you'll probably have to pass through the rest of the iommu 32 group as well, or at least add it to your vfio-pci to prevent unraid from trying to access it.

Check your tools->system devices to see what the ids are on both your cards. They should also appear in separate iommu groups. If the ids are the same on both cards, then when you add them to the vfio-pci it will remove unraids ability to access the cards (but you could pass each of the cards to separate vms if you choose. If you map out what ports/controller/busses are on your onboard, then you can maybe shuffle things around to have unraid using the onboard and have the cards used by your vms. You might also be able to isolate it so that only some of the onboard ports are used by unraid and the other controllers can be passed to vms.

Are you passing through controllers to your vm for hot swapping, or just ports? For usb issues I'd follow the steps here. At the very least it will allow you to document each controller, port, and bus which should help you in figuring out if there's a conflict and what controllers are what.

I should also mention that there is a custom kernel for 6.8.3 that has those patches in place which can be found here:

There is a bug with amd usb and audio where it won't pass through properly (I've been using my arctis 7 headset in Windows as a workaround for now). Kernel 5.8 was supposed to contain a patch to fix that issue, so you can try 6.9 beta 29 and see if that fixes your issue. I had the same question regarding beta 29 and the audio passthrough fix located here:

Usually you would pass through a USB controller. so you can connect and disconnect devices at will. The guide here can help you more with that. https://forums.unraid.net/topic/35112-guide-passthrough-entire-pci-usb-controller/ You can have the vm powered on all the time if you plan to daily drive the VM, there is an autostart toggle on the VM page that will let the VM automatically power on when your Unraid fully loads. Otherwise you can start and stop VMs from webpage on another device like your phone. If you leave the VM on all the time, it will drain performance from the Unraid server, as those resources will be dedicated to your VM, and not your Unraid server. If you run Unraid headless you should be able to pass through the onboard graphics. My motherboard doesn't have onboard graphics, but I have 2 GTX 2060's and I am passing them to separate VMs running in headless mode. Here's a video that can give you some good information on GPU passthrough and running a system headless.

For gaming, you would pass the 1060 through to the VM. This will let the VM access it at the hardware level, and any monitors on that video card will be used by your VM. You will either need another video card (onboard may work) to run your Unraid from if you want local access to it, or you can run it "headless" aka without a monitor and control it from your phone, tablet, other computer, etc via the website.

Hopefully I'm posting this in the right place, Since 6.9 beta 29 uses the 5.8 kernel I wanted to try to get the USB and audio passthrough working, but so far haven't been able to, it just has the same behaviour as before where it hangs the host system and dmesg lists entires like this: vfio-pci 0000:03:00.0: not ready 1023ms after FLR; waiting. Has anyone been able to get the USB and audio passthrough to work on the 5.8 kernel / Beta 29? and if so is there something I need to append to my syslinux? FYI I'm using a Threadripper 3970x on an Asus ROG Zenith II Extreme (in case that matters). My ids match the 1022:1487 and 1022:148c entries that I've been seeing in relation to this issue. Thanks

I need IOMMU for the GPU and USB passthrough to my VM. BIOS is the latest one. I will give 6.9-beta1 a try. Should I install it onto a new key and copy over my config so I can roll back if needed?

The xfs_repair -v fixed the issues with the filesystem, so I started to install more games, and about 20GB into it disk 2 stopped working again. I've included the diagnostics from after the xfs repair, and after the disk stopped working again. obsidian-diagnostics-post-xfs-repair.zip obsidian-diagnostics-20200324-1354.zip

Good morning, So the rebuild succeeded; however, it's now saying that disk 2 is "Unmountable: No file system". I'm thinking that "xfs_repair -V -L /dev/md2" will fix that though? I've attached diagnostics from after the array rebuilt, and after I rebooted it. Once I get the file system sorted I'll be able to continue my testing and install some more games and see if the original issue is solved by swapping the sata cable and adding a new power rail. obsidian-diagnostics-post-rebuild.zip obsidian-diagnostics-post-rebuild-reboot.zip

I added a new line of power from my psu to disks 1 and 2, and swapped out the cable for disk 2. Brought my array back up and am currently rebuilding the array. I expect that to take 9h or so. All my hardware is brand new, so here's to hoping that swapping those things around fixes the issue. I'll post back tomorrow with an update.

Hey guys, The background: I'm new to Unraid. I have a new Unraid build with a Threadripper 3970x on the Asus ROG Zenith II Extreme MB. I'm using the onboard Sata. I'm running a Windows VM from a passed through a Samsung 970 Pro NVMe. I have 5x 6TB Seagate Barracuda Pro drives in my array, 1 of which is parity. The issue: In my Windows VM once I install Steam and start downloading games to my network mapped user share everything will be working great for a bit, then the drive will fail with write errors and remove itself from the array. If I remove the drive and add it back the the array and let it rebuild, everything will be fine until I download more from Steam. First it was disk 1 that was failing, so I excluded it from my user share, now disk 2 is failing when doing the same thing. Additional Information: I have copied over several ISO images via SMB from my old computer to the new unraid shares totaling about 14GB, and no issues seemed to occur. I am not married to any of the data on the drives, as this is a fresh build. I have attached my diagnostics. Hope someone can help. obsidian-diagnostics-20200323-1452.zip