jbartlett

Might have spoke too soon. The video card does indeed utilize but the P2000 is reporting that it found problems. Still mucking with it.

The card is not compatible with UNRAID's "BIND" via vfio-pci.cfg

A video card in PCIe 1 (primary video) can be passed through to a VM if you dump the cards BIOS and pass it in.

Haven't messed with the OC yet.

Might be possible based on a Spaceinvader One video with passing through the card's ROM to it. I'm going to be looking into that because I'd rather have the two Quadro cards on separate nodes.

Pick a port. Any port. Except ones you pass through to a VM, don't pick those.

Here's my current build. This is a VM box so SSD's for hosting the VMs (one per VM) and a spinner for backups of those VMs. I just realized that I haven't had two Win 10 VM's running at the same time. Gotta test that soon! PCIe 1: GeForce GT 1030 (OS) PCIe 2: High Point 4-Port 4-Controller USB 3.0 PCI-Express 2.0 x 4 HBA RocketU 1144D PCIe 3: Quadro P2000 (Win 10) PCIe 4: Empty PCIe 5: Quadro P620 (Win 10)

I watched a LTT video last night that said the Noctua NH-U12A out performs AIO water coolers in both cooling & sound - but custom loops can still out perform the Nactua given a good pump & large reservoir. I put a Noctua NH-U14S on mine. Don't you love that BIOS display?

PCIe 1 is on numa 2 for the 2990WX, CPUs 8-15/40-47 PCIe 3 & 5 are on numa 0, CPUs 0-7/32-39 Numa 1 has CPUs 16-23/48-55 Numa 3 has CPUs 24-31/56-63

I fixed the stuttering sound issue by passing through a USB sound card though yesterday I saw a Spaceinvader One video on YouTube titled "The best way to install and setup a windows 10 vm Part 2 Hardware Passthrough" that shows how to correct what he calls "demon sounds". It's on my To Do list to try.

I've gotten past it in the past by manually editing the XML file to remove the passed through devices at the end, or by recreating the VM, or by plugging the USB device back in and removing it via the GUI.

That's kinda funny - the 2990WX sat there for three weeks before I swapped out the 2950X with it and then it sat for another week before I finished putting the CPU cooler on. Same reason, major remodel in progress. Only thing staying the same is one bathtub. FYI - New BIOS released last month

Might be a good idea to retry your benchmarks. The new BIOS seems to improve processing

LSTOPO PCIe assignments for the 2990WX

Here's the YouTube video mentioned earlier in the thread for identifying which PCIe slot is connected to which numa node

Couple of things to try: Check the IOMMU groups to ensure it & it's sound are in their own group. Move the card to a new slot (will change the domain ID)

Try creating a new VM using the same drive files and see if that boots.

Name of Plugin: Disable Security Mitigations

You really should identify the root cause of your slowdown before you start trying to blindly eliminate it or you may find yourself with a new server and the same performance. Start monitoring the server when you experience the slowdowns to see if you can pin it down. SSD's aren't ideal for long-term storage or files that don't change unless it has tech that shuffles data around for even wear leveling.

Have you ran more than one benchmark and are the drops repeatable?

I've put one of these in my case blowing on the WD Black NVMe drive (& other components) which runs quite hot. Unfortunately, I've forgotten it was in there and needs to be powered back on after power off and now the WD Black didn't show up after my last reboot. 😁 https://www.amazon.com/Fancii-Personal-Portable-Whisper-Technology/dp/B072DSHKCH/ I'm also really glad I re-read through this post, the tips on the slow/fast cores comes in quite handy now that I'm working on a new Ryzen build with a different MB. I'll be taking in four Brio web cams and then outputting them over NDI so I'd like to squeeze every bit of speed out of the Win VM.

I saw a 3% increase in the CPU score with 3DMark Timespy on a Ryzen 2950 (all but 4 cores assigned to the VM) after disabling the protections but several other tests/benchmarks showed no change within a small margin of error on it and a 1950x and an Intel CPU (can't think of the model off hand).

Updated three servers, no issues. New BIND method working seemingly perfectly vs stubbing.

I'm guessing bad device. I rebooted to free it up and now the NVMe drive doesn't register.

Looks like the cause for the IOWAIT is a fstrim being executed on a WD Black 256GB nvme drive, executed by the Trim plugin. The process gets stuck in an uninterruptible sleep. Not a bug in the Dashboard.