I know this doesn’t compare to your collection though @HikariKnight
I know this doesn’t compare to your collection though @HikariKnight
Hcant respond any other way
LI understand, I don’t blame you.
Hyoure catching up though
KDear god man. That's a lot of machines.
KOh wait those are just a lot of displays
Hnvidia test VM (with gpu passthrough), main machine, HTPC (stuck on F39 due to polaris atm but moving to F40 soon once some patches land)
on the go hobby work laptop, legion go, deck (running steamos)
2 penguin plushies my friends sent me as secret santa gifts as jokes (the sunglasses on the small one was a dlc that got sent to me last year, didnt know my secret santa gift came with a season pass)
bunch of controllers i use for testing (plus another 40 in a drawer behind the desk)
on the go hobby work laptop, legion go, deck (running steamos)
2 penguin plushies my friends sent me as secret santa gifts as jokes (the sunglasses on the small one was a dlc that got sent to me last year, didnt know my secret santa gift came with a season pass)
bunch of controllers i use for testing (plus another 40 in a drawer behind the desk)
Tmoved my dots and iso to my ventoy usb drive, wish me luck 
Hgood luck!
K...you have like 50 controllers?
Agnome merged accent colors?
K
Aam i dreaming?
Tdo you have experience in gpu passthrough when only nvidia gpus are in a pc?
Tsince im planning to make use of a second nvidia gpu in a windows vm, so it would be unpractical for me to unload the nvidia module fully
KWe've been blessed
Hthey need to be detected as different GPUs, if you have lets say 2x 3090s you need to add custom scripts tailored for your card into initramfs, this is not something we do
Ti see, i would have 2 4090s, exactly the same manufacturer
Tso i guess thatll be a bit harder?
Hyeah since you need to make a custom script to manually bind the correct card by pci address to vfio-pci at boot before nvidia driver grabs the card as vfio-pci only understands vendorid:deviceid
which would be identical for the 2 cards
which would be identical for the 2 cards
Thmm... i see
Tguess theres a bunch of preparation i will do before the pc gets to me next week
Hnot something we support sadly, if you know how to set it up feel free to PR but i have no way to test it since i just do different cards as it is 1000x easier
Twill definitely do a PR if i figure it out
Hi use amd on host and nvidia on VM
Tits more of an idea than something i instantly prep
Hwendell has a guide on level1tech for debian based systems, no idea how to do the same in fedora tbh
Tlike, i just dont have any real usecase where i would have a good use for the second 4090 (my boss sends me the PC as a gift), so i figure something out where i can put it to good use
Tand given there are some new windows games that dont yet work (as far as i know) i wanna try out, it may be worth a shot
Hsell one of the 4090s and get an amd card? 
MAbout to join you with the multi-gpu passthrough.
Hoh you are? 
Ti mean yea i could do that, but in case i would have a usecase for it, it most probably is for AI. ROCm is.... questionable to use tbh
Tit has gotten better, but its still shit compared to what nvidia offers
Hamd on host, nvidia in VM is how i do it
Hhowever linux vm would need a physical monitor or rely on sunshine (without mesa breaking it)
Twould probably be the other way around for me, i intent to use atleast one nvidia gpu on containers
Tand WSL still has a weird design flaw where it doesnt release ram it used previously
MDummy plug
MBest thing ever
Hlinux vm will need dummy plug + sunshine or physical monitor
no looking-glass host for linux (no maintainer for that portion and its currently broken)
no looking-glass host for linux (no maintainer for that portion and its currently broken)
MOh. Yeah for Linux I always used a monitor
Him glad we made passthrough pretty simple to setup here since we have set clear guidelines
2 different cards (must have different
2 different cards (must have different
vendor:device ids and have good iommu groups) and rest should in most cases just work Hif you want anything else, youre on your own, but we enabled what is required for you
MSo I'm going to do it with zero layering.... Hopefully.....
MThe only gotcha I have is trying to do the vfio stub for the second gpu
MBut the goal is only kargs and containers
Hyou will have to have dracut regeneration enabled unless you do all that required stuff in build time like you do the initramfs stuff in bluefin (which we now use in bazzite)
DWhy is my reboot and logoff buttons gone?
D
