
PCI Passthrough - System Configuration - Part 1 - Chris Titus Tech
video description
Date: 2022-03-20
Related videos
Comments and reviews: 10
h3llh0und00
A couple of notes since I did this manually (had an Arch linux guide and performed all this on a Debian). First of all some GPU's have more than video and audio (I have a 2070 and it has 4 devices video, audio, usb c port and something else...). So make sure that within the group all devices are using vfio drivers (a good hint can be checking the first 4 characters of the device id which is the manufacturer id that is within the same group). Also AMD GPU's are not necessarily better. NVIDIA's code 43 error (no passthrough for you as I call it) can be fixed by changing 3 things and you can find MANY MANY tutorials. GPU's like the 5500XT that I have (8 GB version by MSI) has a ghosting issue. What happens is that it tries to load the vortial; machine. It fails and then when you forcefully shut it down, even though the virtual machine is shut down your AMD GPU holds the last screen and is stuck there. I have tried a patch but that didn't work. Finally a question. What if I have two cards of the same model? (two 1050 ti's to be exact) different vendors but the device ids are identical. How can I differentiate them in order to load vfio drivers on only one of them. Is it possible?
reply
A couple of notes since I did this manually (had an Arch linux guide and performed all this on a Debian). First of all some GPU's have more than video and audio (I have a 2070 and it has 4 devices video, audio, usb c port and something else...). So make sure that within the group all devices are using vfio drivers (a good hint can be checking the first 4 characters of the device id which is the manufacturer id that is within the same group). Also AMD GPU's are not necessarily better. NVIDIA's code 43 error (no passthrough for you as I call it) can be fixed by changing 3 things and you can find MANY MANY tutorials. GPU's like the 5500XT that I have (8 GB version by MSI) has a ghosting issue. What happens is that it tries to load the vortial; machine. It fails and then when you forcefully shut it down, even though the virtual machine is shut down your AMD GPU holds the last screen and is stuck there. I have tried a patch but that didn't work. Finally a question. What if I have two cards of the same model? (two 1050 ti's to be exact) different vendors but the device ids are identical. How can I differentiate them in order to load vfio drivers on only one of them. Is it possible?
reply
Chris
This Video is about Playing Windows Games in a Virtual Machine using a second graphics card dedicated for doing this on Linux!
Notes:
-This is geared for an advanced user as this is a complex subject
-Passing through a second video card is system dependent. Steps change based on the system you are using. I try to outline what modifications you need to adjust based on your system.
-I am using a Debian based system on Kernel 4.20, AMD CPU, and AMD Cards
-This will -NOT- work if you have two identical cards
reply
This Video is about Playing Windows Games in a Virtual Machine using a second graphics card dedicated for doing this on Linux!
Notes:
-This is geared for an advanced user as this is a complex subject
-Passing through a second video card is system dependent. Steps change based on the system you are using. I try to outline what modifications you need to adjust based on your system.
-I am using a Debian based system on Kernel 4.20, AMD CPU, and AMD Cards
-This will -NOT- work if you have two identical cards
reply
WizardNumberNext
I doubt anybody is using so old Linux kernel
Manual Enabling IOMMU was needed many years ago.
Now it is done by default
I am trying to get working AMD RX 570 (sapphire pulse) on Dell PowerEdge R715.
I got to stage it actually gets firmware loaded and it is working... till xorg starts to load... then it crashes... any trial to fix issue will hard lock whole server, which will be subsequently reset by watchdog.
Windows works, but I want to run Linux inside VM
reply
I doubt anybody is using so old Linux kernel
Manual Enabling IOMMU was needed many years ago.
Now it is done by default
I am trying to get working AMD RX 570 (sapphire pulse) on Dell PowerEdge R715.
I got to stage it actually gets firmware loaded and it is working... till xorg starts to load... then it crashes... any trial to fix issue will hard lock whole server, which will be subsequently reset by watchdog.
Windows works, but I want to run Linux inside VM
reply
--.------
Hi, it's just amazing video. Thanks for sharing it.
But i have small doubts regarding this as i working on mini supercomputer with GPU for deep learning based malware analysis..
The doubts are:
1) Do GPU passthrough for VM affects the host OS access to GPU for all users
2) I have two user login in the system in which it is been used by my friend. Will the passthrough will also affect his user id also...
will be thankful if you share your suggestions
reply
Hi, it's just amazing video. Thanks for sharing it.
But i have small doubts regarding this as i working on mini supercomputer with GPU for deep learning based malware analysis..
The doubts are:
1) Do GPU passthrough for VM affects the host OS access to GPU for all users
2) I have two user login in the system in which it is been used by my friend. Will the passthrough will also affect his user id also...
will be thankful if you share your suggestions
reply
David
Any idea's?
System specs
Ubuntu 18.04/kernel version, 4.15.0.65-generic
AMD FX 8230
2 RX570s 1 4gb, 1 8gb
32gb of ram
SSD 240gb
Running apps on dedicated gpu----2nd rx570 8gb
I am using the DRI_PRIME=1 inkscape
I want to know how I can tell the 2nd gpu is being used.
I know there is a command to check the gpu memory usage I can't find anymore. By any chance do you know?
reply
Any idea's?
System specs
Ubuntu 18.04/kernel version, 4.15.0.65-generic
AMD FX 8230
2 RX570s 1 4gb, 1 8gb
32gb of ram
SSD 240gb
Running apps on dedicated gpu----2nd rx570 8gb
I am using the DRI_PRIME=1 inkscape
I want to know how I can tell the 2nd gpu is being used.
I know there is a command to check the gpu memory usage I can't find anymore. By any chance do you know?
reply
Kyle
For anyone with 2 NVIDIA cards looking to do this make sure you are looking into another solution if you use a black list for the Neavou the other card will not be able to use this or the proprietary drivers to power other displays. You can look into a softdep solution to get around this
reply
For anyone with 2 NVIDIA cards looking to do this make sure you are looking into another solution if you use a black list for the Neavou the other card will not be able to use this or the proprietary drivers to power other displays. You can look into a softdep solution to get around this
reply
Nicholas
Thanks for the great guide and explanations. This worked for me on Ubuntu 18.04, but now, on 20, VFIO isn't part of the kernel and doesn't load at all. Do you know the best way to get VFIO working on the latest kernel? I'm sure some others have this same question
reply
Thanks for the great guide and explanations. This worked for me on Ubuntu 18.04, but now, on 20, VFIO isn't part of the kernel and doesn't load at all. Do you know the best way to get VFIO working on the latest kernel? I'm sure some others have this same question
reply
pazu_
I have an Intel CPU. Have enabled Intel virtualization from bios. I get -DMAR: IOMMU enabled- and -AMD IOMMUv2 functionality not available on this system- with grep IOMMU. Am I supposed to have amd iommu or is dmar the one for intel or...?
reply
I have an Intel CPU. Have enabled Intel virtualization from bios. I get -DMAR: IOMMU enabled- and -AMD IOMMUv2 functionality not available on this system- with grep IOMMU. Am I supposed to have amd iommu or is dmar the one for intel or...?
reply
Patrick
you skipped the Blacklisting commands and this crap doesn't work as you stated to do things and you move between commands so fast this is a Shitty tutorial you forgot to mention how to save the changes you make in those folders also!
reply
you skipped the Blacklisting commands and this crap doesn't work as you stated to do things and you move between commands so fast this is a Shitty tutorial you forgot to mention how to save the changes you make in those folders also!
reply
Frederic-xavier
Hello, gret tuto indeed
what if you have only One GPU (like I do) ?btw could not find equiv of /etc/initramfs-tools/modules in manjaro
And found that /etc/modules is indeed /etc/modules-load.d/modules.conf on manjaro
reply
Hello, gret tuto indeed
what if you have only One GPU (like I do) ?btw could not find equiv of /etc/initramfs-tools/modules in manjaro
And found that /etc/modules is indeed /etc/modules-load.d/modules.conf on manjaro
reply
Add a review, comment
Other channel videos















