Jump to content

belliash

Members
  • Posts

    23
  • Joined

  • Last visited

Posts posted by belliash

  1. 10 hours ago, belliash said:

    I didn't.

    Anyway the PCI quirks patch should be no longer necessary since AGESA 1.0.0.3ABB, as AMD claims they fixed this problem. Yet, this is something I still have not checked. The 2 other patches gives you only hardware monitoring and you can safely omit them.

    I confirm, AMD fixes PCI quirks problem with AGESA 1.0.0.3ABB! No patching required any longer!

  2. 7 hours ago, helin0x said:

    How’be you applied those fixes to unraid?

    I didn't.

    Anyway the PCI quirks patch should be no longer necessary since AGESA 1.0.0.3ABB, as AMD claims they fixed this problem. Yet, this is something I still have not checked. The 2 other patches gives you only hardware monitoring and you can safely omit them.

  3. There are some issues, but they do not influent on virtualization and guests. In my case, I cannot achieve the advertised CPU boost speed and I needed to apply some patches on top of Linux kernel in order to make k10temp recognizing CPU and reporting its temperature. Also I have applied a PCI quirks patch, that allows to start VM with GPU passthrough. However I have contacted AMD support regarding this issue and seems they are aware of the problem and the upcoming AGESA 1.0.0.3ABB update should fix it, so this patch will be no longer required.

  4. Yep, read that too, that's why asking about legacy mode.

    I think I would get rid of UEFI and secure boot.

    Shame unraid does not provide some experimental kernel with above patch applied.

     

    BTW: What is the performance of your GPU in VM?

    And one more question: Could you post your VM XML config file or command line?

  5. Yeah, that's how I passed optimus-based nvidia card in my laptop.

    I wonder if this problem is really related to BIOS, not some combination of BIOS, BIOS settings and Linux kernel. I found several threads over the net, some contained patches for kernel, as well as mostly UEFI was used. I wonder if problem would still exist when booted in legacy mode.

     

    I finally came across this patch: https://clbin.com/VCiYJ that most probably fixes problems introduced by AGESA 0.0.7.2 (BIOS version F30). According to the information found over the net it works at least with Linux Kernel 5.1.3. Maybe you could try it?

     

    There is also a chance, AGESA 1.0.0.3AB will bring some bugfixes.

  6. Are you sure, you don't miss some settings in new BIOS? CSM? IOMMU?

    Aso, did you try this patch: 

    lI know its related to Threadripper, but maybe will also help you for this reset problem?

     

    There is also a parameter for vfio-pci module - disable_idle_d3:Disable using the PCI D3 low power state for idle, unused devices (bool)

  7. 3 hours ago, Caduceus said:

    I recently built an unRaid server with almost this exact setup.

    Same motherboard, same cpu, I have the 1660ti (Gigabyte 1660ti mini-itx) setup in the primary pcix16 slot) but I have an old Radon 5770 in the second slot.

    The only caveat is, I need to run the F5 version of the bios. I had upgraded to F32 and F40. Both gave me serious issues that would not allow Windows VMs to start with those bios revisions.  I did a full post in the VM Troubleshooting section.  I hope this helps.

     

    You can see my IOMMU groups here:Post with IOMMU groups

    And were you able to use Radeon 5770 as primary display for host, and to pass 1660Ti to VM?

     

    Your problem with BIOS doesn't look good, especially I am considering a different CPU to buy - Ryzen7 3700X and it needs F40.

  8. Hello,

     

    I plan to buy z Ryzen7 2700 CPU and some B450-based mATX mainboard. I woud like to passthrough GPU to Windows 10 running in VM, thus I will need to buy 2 graphics cards.

     

    Actually I am thinking of the following configuration:

    Motherboard: Gigabyte B450 Aorus M

    CPU: Ryzen7 2700

    GPU1: GeForce GT 710

    GPU2: GeForce 1660Ti

     

    I read about problems with IOMMU groups, but I found information in manual that this mainboard allows to choose which GPU to use as primary card. I am not sure if this is going to work, but according to motherboard's manual, it should be possible to use GT710 in bottom PCI-E slot in Linux and to pass GTX1660Ti from top PCI-E slot to VM. This seems to be best scenario, because:

    * first of all, top PCI-E slot seems to come from CPU PCI-E lanes and seems to be in separate IOMMU group. It is also full 16x slot.

    * the bottom PCI-E slot is slower and share IOMMU group I think. At least users reports that in other B450 based motherboards. There would be no need to use ACS if this card could be used as primary card.

     

    Can any owner of B450 based mainboard post IOMMU groups here including mainboard model and BIOS version?

    Maybe someone own the B450 Aorus mainboard from Gigabyte? From what I checked, all models (not only M) should be capable of useing bottom GPU as primary card. Can anyone confirm this is really working?

     

    Thanks in advance!

    Regards!

×
×
  • Create New...