reexe

If you use HDMI audio and your audio device wont work after a reboot it is most likely because the card did not have enough time to start, to fix this you can change the enable script to look something like this: devcon enable *VEN_1002* timeout 20 devcon enable *VEN_1002* This means that whatever device that is affected by *VEN_1002* will get the enable command twice, the second command will be sent 20 seconds after the first one.

This is a guide on how to fix blackscreen and or screen tearing issues regarding to graphics cards when using gpu passthrough after a reboot of the guest. This fix relies on disabling the graphics driver before a shutdown and then enabling it again after a startup. Download devcon.exe and place it in System32 (I found a guide on google on how to extract it from a file downloaded from Microsoft). Now find out the id of the graphics card for example "PCI\VEN_1002&DEV_67B0&SUBSYS_20151462&REV_80\3&13C0B0C5&0&28: AMD Radeon R9 390 Series" by typing "devcon find *" in a cmd window. We will only be using the "VEN1002" part (from my example, yours may be different). Make a text file in my documents named "gpufix enable.bat" and type: devcon enable *VEN_1002* Then make another one named "gpufix disable.bat" and type: devcon disable *VEN_1002* Start gpedit.msc and move to "User Configuration/Windows Settings/Scripts (Logon/Logoff). Side note: I know it makes more sense to have the script run as Startup/Shutdown but i could not get that to work. Double click Logon on the right side and press "Add..." Press browse and select the enable file Now we will do the same thing with Logoff except we use disable instead of enable. Now you are done! This is my first guide and i have had this in my notes for a long time, but after having to redo it so many times after fiddling around with unRaid i finally thought that i should share this in case someone else was in need of it (and for my own sake so i can find it on my profile ). This bug is briefly talked about on one of Linus tech tips videos and Linus made it sound like it had something to do with the graphics card not being shut down correctly.

Have you tested to change the switch on your GPU to the other BIOS? I am not sure if your card has one, but some do, if your have it and your lucky it may solve your problem

I used the latest drivers and i have also tried the ones that windows 10 downloads by default when first booting. However i have put unRAID on ice until this is fixed or i get another gpu. Edit: The latest at the time ofc since i have not tried unRAID since then.

Disabled cool "n" quiet in bios, made sure i had the latest virtio drivers with no unknown/other devices in device manager (everything updated) and downloaded the tool "MSI_util" my GPU was already checked but i made sure to check my hdmi audio and integrated audio. Restarted computer and tested GW2 again, still getting the same problems, it does seem to run slightly better then b4 all the tweaks i have been doing the past few days, but its is not nearly close to bare metal.

Sounds promising, I will try that, have to look into what MSI is though, dont know what that is rly. What cpu are u running in your AMD system?

Since you asked this, did you follow the directions from here when doing the install (the order supposedly matters, but I doubt the order would cause your issues!)? http://lime-technology.com/wiki/index.php/UnRAID_Manual_6#Installing_a_Windows_VM I loaded only viostor driver on install and then installed the rest when windows was installed, i used version 112.

I cant seem to get gpu passtrough to work in OVMF. I get black screen both on my main gpu and secondary gpu when trying them one at a time. When using vnc as gpu it works fine.

It worked! ty! Off to trying out gpu passtrough and cpu benchmarks.

Lol sry for being stupid. Going to test now, "holding my thumbs"

Managed to extract it, was just me not being used to 7zip. However when i try to move the file from a shared unraid folder to the place described i get this sudo mv OVMF-pure-efi.fd /boot/config/OVMF-pure-efi.fd mv: failed to preserve ownership for ‘/boot/config/OVMF-pure-efi.fd’: Operation not permitted

When i extract the second file i find nothing inside it. Also where is /boot/config? is it in the vdisk for my vm? If so how do i get access to the files inside whitout having a vm running? do i need to mount that vdisk to another working vm? Edit: Oh, I guess u mean /boot/config in the unraid os?

Currently trying to get OVMF to work again, but after install and first reboot i just cant get it to work again. It installs just fine and in the middle of the install it reboots 2 times, those times i need to use fs0: cd efi/boot bootx64.efi to get it to start again, after sucessfull install i am left at the desktop and everything seems fine, after i then shutdown or reboot the vm i can not get it to start at all again. After running those commands again from the shell the vm just restarts. So i mounted a secondary vdisk that have rEFInd on it and i can boot into that using the shell commands, and rEFInd can see the windows boot manager but the vm just restart again if i select it.

Im going to give OVMF a shot again and hope it will work untill 6.2 is relesed and that it will automaticly fix everything when i use OVMF... "6.2 is so close"

I feel lag when i watch youtube or play games or have multiple chrome windows open at once. I have tried the other gpu and got the same problem. Also tried OVMF but i did not get far as it seems the vm got corrupted after 3 boots or so, could not even start it using the commands fs0: cd efs/boot bootx64.efs. Even tried using rEFInd boot manager.

Since i am not entirely sure how emulatorpin is supposed to be used i removed it, but i still got the same results when trying to use 4 physical cores assigned to 2 virtual cores. <vcpu placement='static'>2</vcpu> <cputune> <vcpupin vcpu='0' cpuset='4-5'/> <vcpupin vcpu='1' cpuset='6-7'/> </cputune> I still did not get the expected scores.

I dont know what version of qemu unraid uses, have tried to figure it out but i dont rly know how. i think i read somewhere that cpuset 0-2 should be supported in qemu 0.9.0.

I thought emulatorpin only told the host to run the emulator process on those specific cores? Edit: what i mean is that i thought that the cores specified in emulatorpin was real cores that i could tell the host to offload some processes to.

The 4k videos on youtube lag even if i dont run them fullscreen. So i think you can still try them as reference, LinusTech have 4k if y are willing to try it out. My cpu goes to 100% on 4k and both video and sound stutters.

I installed ethernet from a virtio iso and i used it when i added a driver at install to get my win10 to recognize the vdisk, is there something else you can do with it?

I did this and i got way less scores when trying to use 2x cores for each vm core, so i would say unraid dont support this, unless there is some other way of doing it that i am not aware of. When looking at the forum i posted about earlier it seems that vfio supports this outside of unraid. Edit: To clarify the picture, upper test is <vcpu placement='static'>2</vcpu> <cputune> <vcpupin vcpu='0' cpuset='4-5'/> <vcpupin vcpu='1' cpuset='6-7'/> <emulatorpin cpuset='2-3'/> </cputune> and then the bottom test is <vcpu placement='static'>4</vcpu> <cputune> <vcpupin vcpu='0' cpuset='4'/> <vcpupin vcpu='1' cpuset='5'/> <vcpupin vcpu='2' cpuset='6'/> <vcpupin vcpu='3' cpuset='7'/> <emulatorpin cpuset='2-3'/> </cputune>

I am now trying this <cputune> <vcpupin vcpu='0' cpuset='4-5'/> <vcpupin vcpu='1' cpuset='6-7'/> <emulatorpin cpuset='2-3'/> </cputune> And this should make my vm believe i have 2 cores at around 8Ghz per core since 1 physical core is around 4Ghz, but my vm only reports 2 cores at 4Ghz. Or is there any other way i can see if it works or not? System feels slow so i dont think it works.

I understand what u mean, but it does not seem that i get the improvements of the 2 cores, it seems like the vm only uses one core even if i type 0-1, i dont know how to confirm if its working or not, but i dont see any speed improvements

Changes that i make stay, but when i set 0-1 the VM only recognizes the 0, and when i open the gui edit in the webgui i only see the nr 0 selected (the xml file still shows 0-1). I tried <emulatorpin cpuset => and when i selected the same cores as the vm was using i got rly horrible results, the DPC Latency program was all red, but when i changed to different cores then the vm was using i got better results, still not nearly good enough tho. When i have multiple chrome windows open and for example a youtube video at the same time my vm cpu freaks out completely and stays at 100% all the time. So i am beginning to think its CPU related for sure...

Here is a DPC Latency result using the same settings as i have in my XML in the first post. Sometimes i get huge spikes, and it does not seem very stable, this test ran while i was running Guild Wars 2.