{"id":26675767,"url":"https://github.com/isc30/ryzen-gpu-passthrough-proxmox","last_synced_at":"2026-03-09T12:38:05.987Z","repository":{"id":199870315,"uuid":"703975440","full_name":"isc30/ryzen-gpu-passthrough-proxmox","owner":"isc30","description":"Get the Ryzen processors with AMD Radeon 680M/780M integrated graphics or RDNA2/RDNA3 GPUs running with Proxmox, GPU passthrough and UEFI included.","archived":false,"fork":false,"pushed_at":"2026-01-12T07:46:57.000Z","size":34360,"stargazers_count":1066,"open_issues_count":61,"forks_count":115,"subscribers_count":15,"default_branch":"main","last_synced_at":"2026-01-12T17:47:04.427Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Batchfile","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/isc30.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2023-10-12T09:34:02.000Z","updated_at":"2026-01-12T14:36:40.000Z","dependencies_parsed_at":null,"dependency_job_id":"79888890-30c4-479d-8e98-439ade06faac","html_url":"https://github.com/isc30/ryzen-gpu-passthrough-proxmox","commit_stats":null,"previous_names":["isc30/um773-lite-proxmox","isc30/ryzen-gpu-passthrough-proxmox","isc30/ryzen-7000-series-proxmox"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/isc30/ryzen-gpu-passthrough-proxmox","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isc30%2Fryzen-gpu-passthrough-proxmox","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isc30%2Fryzen-gpu-passthrough-proxmox/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isc30%2Fryzen-gpu-passthrough-proxmox/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isc30%2Fryzen-gpu-passthrough-proxmox/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/isc30","download_url":"https://codeload.github.com/isc30/ryzen-gpu-passthrough-proxmox/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isc30%2Fryzen-gpu-passthrough-proxmox/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":28799834,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-01-27T01:07:07.743Z","status":"online","status_checked_at":"2026-01-27T02:00:07.755Z","response_time":168,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-03-26T03:20:20.069Z","updated_at":"2026-01-27T03:36:47.816Z","avatar_url":"https://github.com/isc30.png","language":"Batchfile","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Proxmox - Ryzen 7000 series - AMD Radeon 680M/780M/RDNA2/RDNA3 GPU passthrough\n\nThis is a guide to get the Ryzen 7000 series processors with AMD Radeon 680M/780M integrated graphics or RDNA2/RDNA3 GPU running with Proxmox, GPU passthrough and UEFI included.\n\n[![project chat](https://img.shields.io/badge/zulip-join_chat-brightgreen.svg)](https://isc30.zulipchat.com/join/36je7erap524zo5hh2gjohyx/)\n\n## Confirmed list of Proxmox versions that work:\n\n- [x] Proxmox 7.4\n- [x] Proxmox 8.0 (recommended, has a newer kernel)\n\n## Confirmed list of hardware that works:\n\n- [x] Ryzen 5 5500U (Zen 2 Lucienne iGPU) \n- [x] Ryzen 5 5600G (Zen 3 Cezanne iGPU)\n- [x] Ryzen 5 6600H (RDNA2 660M Rembrandt iGPU) \n- [x] Ryzen 7 6800H (RDNA2 680M Rembrandt iGPU) \n- [x] Ryzen 7 7735HS (RDNA2 680M Rembrandt iGPU) \n- [x] Ryzen 7 7840HS (RDNA3 780M Phoenix iGPU)\n- [x] Ryzen 7 8845HS (RDNA3 780M Phoenix iGPU) - thanks @serega404 for confirming\n- [x] Ryzen 9 7940HS (RDNA3 780M Phoenix iGPU) - thanks @mmaiero for confirming\n- [x] Ryzen 9 7900X (RDNA2 Raphael iGPU) - thanks to @engels74 for confirming\n- [x] Ryzen 9 7950X3D (RDNA2 Raphael iGPU) - thanks @romner-set for confirming\n- [x] 6900XT (RDNA2 Navi 21 dGPU) - thanks @mpaulo for confirming\n- [x] Ryzen 9 6900HX (RDNA2 680M Rembrandt iGPU) - thanks @Nucs for confirming\n\n# Installing Proxmox VE\n\n1. Download [Proxmox](https://www.proxmox.com/en/downloads/proxmox-virtual-environment/iso) and create a installation usb (with [rufus](https://rufus.ie/en/) for example)\n1. Boot your PC from the USB and run the proxmox installation (in my case it's a Minis Forum UM773 Lite)\n2. If installing Proxmox 7.4, you need to fix the graphical installer when it crashes (known issue in proxmox 7.4)\n   - Wait until it says `Starting the installer GUI`, then press `CTRL + ALT + F3`\n   - Run the following commands:\n     ```\n     Xorg -configure\n     cp /root/xorg.conf.new /etc/X11/xorg.conf\n     sed -i 's/amdgpu/fbdev/g' /etc/X11/xorg.conf\n     ```\n   - Press `CTRL + ALT + F1` and run `startx`\n4. Complete the installation:\n     - For an IPv4 static IP, type `192.168.1.XXX/24`\n     - After the installation finishes, type `exit` in the terminal\n6. Make sure that the web interface is working: `https://\u003cTHE_IP_YOU_CONFIGURED\u003e:8006/`\n7. Connect to the machine via SSH:\n    ```\n    ssh root@\u003cTHE_IP_YOU_CONFIGURED\u003e\n    ```\n3. Proxmox VE comes with Enterprise repositories configured by default, we need to switch to the non-subscription ones to get proxmox updates:\n    ```\n    bash -c \"$(curl -fsSL https://raw.githubusercontent.com/community-scripts/ProxmoxVE/main/tools/pve/post-pve-install.sh)\"\n    ```\n4. Install the CPU Microcode packages:\n    ```\n    bash -c \"$(curl -fsSL https://raw.githubusercontent.com/community-scripts/ProxmoxVE/main/tools/pve/microcode.sh)\"\n    ```\n    \n# Configuring the GPU for passthrough\n\nThe process of doing a GPU passthrough isn't complicated, it's about making sure the host doesn't load the GPU drivers and that the GPU PCI connection can be sent to the VM completely.\n\n1. First, we need to discover the GPU PCI ID:\n    ```\n    lspci -nn | grep -e 'AMD/ATI'\n    ```\n    in my case, it looks like this:\n    ```\n    34:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Device [1002:1681] (rev 0a)\n    34:00.1 Audio device [0403]: Advanced Micro Devices, Inc. [AMD/ATI] Device [1002:1640]\n    ```\n    \n    From this information, we can extract some PCI IDs and device numbers. From now on, when you see these numbers in some commands, replace them with your own numbers:\n    - GPU: `1002:1681` + `0000:34:00.0`\n    - Audio Device: `1002:1640` + `0000:34:00.1`\n\n1. Now, we need to enable iommu which allows the CPU to have full control of direct memory access devices (like the GPU)\n    ```\n    sed -i 's/GRUB_CMDLINE_LINUX_DEFAULT=\"quiet\"/GRUB_CMDLINE_LINUX_DEFAULT=\"quiet iommu=pt\"/g' /etc/default/grub\n    update-grub\n    ```\n\n1. Add the kernel modules to enable vfio (virtual function io) that allows to virtualize these devices\n    ```\n    echo \"vfio\" \u003e\u003e /etc/modules\n    echo \"vfio_iommu_type1\" \u003e\u003e /etc/modules\n    echo \"vfio_pci\" \u003e\u003e /etc/modules\n    echo \"vfio_virqfd\" \u003e\u003e /etc/modules\n    ```\n\n1. Then, we need to tell `vfio` which devices to virtualize (the GPU `1002:1681` + Audio `1002:1640`)\n    ```\n    echo \"options vfio-pci ids=1002:1681,1002:1640\" \u003e\u003e /etc/modprobe.d/vfio.conf\n    ```\n\n1. Load the `vfio-pci` driver before the original one. This prevents the host from using the GPU and allows for virtualization. These are the default AMD + Sound drivers, but you can find the ones your system is using by running `lspci -nnk` and checking the \"Kernel driver in Use\" section.\n    ```\n    echo \"softdep radeon pre: vfio-pci\" \u003e\u003e /etc/modprobe.d/vfio.conf\n    echo \"softdep amdgpu pre: vfio-pci\" \u003e\u003e /etc/modprobe.d/vfio.conf\n    echo \"softdep snd_hda_intel pre: vfio-pci\" \u003e\u003e /etc/modprobe.d/vfio.conf\n    ```\n\n1. Refresh the kernel modules and restart:\n    ```\n    update-initramfs -u -k all\n    shutdown -r now\n    ```\n    \n1. After the restart, validate that the `Kernel driver in Use` for these PCI devices is `vfio-pci` now, with\n    ```\n    lspci -nnk\n    ```\n\n\u003c!--\n## Making the GPU available to the host again after stopping the VMs (solving the AMD Reset Bug)\n\nThere is a known bug with AMD graphics cards where the host crashes after it tries to use a GPU after the passthrough. This is a mitigation effort but there isn't a real solution available for this.\n\n3. Install the `vendor-reset` kernel module:\n    ```\n    apt install pve-headers-$(uname -r)\n    apt install git dkms build-essential\n    git clone https://github.com/gnif/vendor-reset.git\n    cd vendor-reset\n    dkms install .\n    echo \"vendor-reset\" \u003e\u003e /etc/modules\n    ```\n3. Create the GPU re-initialization service. **Remember to update the GPU PCI IDs to your own ones**:\n    ```\n    cat \u003c\u003c EOF \u003e\u003e  /etc/systemd/system/vreset.service\n    [Unit]\n    Description=AMD GPU reset method\n    After=multi-user.target\n    [Service]\n    ExecStart=/usr/bin/bash -c '\\\n      echo device_specific \u003e /sys/bus/pci/devices/0000:34:00.0/reset_method ;\\\n      echo device_specific \u003e /sys/bus/pci/devices/0000:34:00.1/reset_method ;\\\n      echo 1 \u003e /sys/bus/pci/devices/0000:34:00.0/remove \u0026\u0026\\\n      echo 1 \u003e /sys/bus/pci/devices/0000:34:00.1/remove \u0026\u0026\\\n      echo 1 \u003e /sys/bus/pci/rescan'\n    [Install]\n    WantedBy=multi-user.target\n    EOF\n    systemctl enable vreset.service \u0026\u0026 systemctl start vreset.service\n    ```\n3. Refresh the kernel modules and restart:\n    ```\n    update-initramfs -u -k all\n    shutdown -r now\n    ```\n2. After the restart, you can check if the `vendor-reset` module was started properly with `dmesg | grep vendor_reset`. The output should be similar to:\n    ```\n    [    2.369873] vendor_reset: loading out-of-tree module taints kernel.\n    [    2.426548] vendor_reset_hook: installed\n    ```\n--\u003e\n\n## Creating the Windows VM\n\n1. Via the web UI, upload a Windows 10 installation ISO image to the `local` storage.\n    - Click on `local` in the left menu, then `ISO Images`, then `Upload`\n1. Create the VM with the following parameters:\n    - ISO Image: the one you uploaded\n    - Type: Microsoft Windows\n    - Version: 10/2016/2019\n    - Machine: q35\n    - Bios: SeaBIOS\n    - Qemu Agent: ON\n    - Disk: at least 64gb, Discard ON, SSD emulation ON\n    - CPU Type: host\n    - RAM: at least 12gb\n1. Run the machine and install Windows (type of installation: custom)\n2. After finishing the Windows installation, stop the VM\n\n# Configuring the GPU in the Windows VM\n\nIn order to pass the GPU device properly, we need to tell the VM which GPU BIOS to use. Luckily for us, we can extract this from the host machine via SSH or download it directly from the repo:\n1. Create a `vbios.c` file in the host (proxmox) with the following contents:\n    \u003cdetails\u003e\u003csummary\u003eExpand `vbios.c`\u003c/summary\u003e\n\n    ```c\n    #include \u003cstdint.h\u003e\n    #include \u003cstdio.h\u003e\n    #include \u003cstdlib.h\u003e\n\n    typedef uint32_t ULONG;\n    typedef uint8_t UCHAR;\n    typedef uint16_t USHORT;\n\n    typedef struct {\n        ULONG Signature;\n        ULONG TableLength; // Length\n        UCHAR Revision;\n        UCHAR Checksum;\n        UCHAR OemId[6];\n        UCHAR OemTableId[8]; // UINT64  OemTableId;\n        ULONG OemRevision;\n        ULONG CreatorId;\n        ULONG CreatorRevision;\n    } AMD_ACPI_DESCRIPTION_HEADER;\n\n    typedef struct {\n        AMD_ACPI_DESCRIPTION_HEADER SHeader;\n        UCHAR TableUUID[16]; // 0x24\n        ULONG VBIOSImageOffset; // 0x34. Offset to the first GOP_VBIOS_CONTENT block from the beginning of the stucture.\n        ULONG Lib1ImageOffset; // 0x38. Offset to the first GOP_LIB1_CONTENT block from the beginning of the stucture.\n        ULONG Reserved[4]; // 0x3C\n    } UEFI_ACPI_VFCT;\n\n    typedef struct {\n        ULONG PCIBus; // 0x4C\n        ULONG PCIDevice; // 0x50\n        ULONG PCIFunction; // 0x54\n        USHORT VendorID; // 0x58\n        USHORT DeviceID; // 0x5A\n        USHORT SSVID; // 0x5C\n        USHORT SSID; // 0x5E\n        ULONG Revision; // 0x60\n        ULONG ImageLength; // 0x64\n    } VFCT_IMAGE_HEADER;\n\n    typedef struct {\n        VFCT_IMAGE_HEADER VbiosHeader;\n        UCHAR VbiosContent[1];\n    } GOP_VBIOS_CONTENT;\n\n    int main(int argc, char** argv)\n    {\n        FILE* fp_vfct;\n        FILE* fp_vbios;\n        UEFI_ACPI_VFCT* pvfct;\n        char vbios_name[0x400];\n\n        if (!(fp_vfct = fopen(\"/sys/firmware/acpi/tables/VFCT\", \"r\"))) {\n            perror(argv[0]);\n            return -1;\n        }\n\n        if (!(pvfct = malloc(sizeof(UEFI_ACPI_VFCT)))) {\n            perror(argv[0]);\n            return -1;\n        }\n\n        if (sizeof(UEFI_ACPI_VFCT) != fread(pvfct, 1, sizeof(UEFI_ACPI_VFCT), fp_vfct)) {\n            fprintf(stderr, \"%s: failed to read VFCT header!\\n\", argv[0]);\n            return -1;\n        }\n\n        ULONG offset = pvfct-\u003eVBIOSImageOffset;\n        ULONG tbl_size = pvfct-\u003eSHeader.TableLength;\n\n        if (!(pvfct = realloc(pvfct, tbl_size))) {\n            perror(argv[0]);\n            return -1;\n        }\n\n        if (tbl_size - sizeof(UEFI_ACPI_VFCT) != fread(pvfct + 1, 1, tbl_size - sizeof(UEFI_ACPI_VFCT), fp_vfct)) {\n            fprintf(stderr, \"%s: failed to read VFCT body!\\n\", argv[0]);\n            return -1;\n        }\n\n        fclose(fp_vfct);\n\n        while (offset \u003c tbl_size) {\n            GOP_VBIOS_CONTENT* vbios = (GOP_VBIOS_CONTENT*)((char*)pvfct + offset);\n            VFCT_IMAGE_HEADER* vhdr = \u0026vbios-\u003eVbiosHeader;\n\n            if (!vhdr-\u003eImageLength)\n                break;\n\n            snprintf(vbios_name, sizeof(vbios_name), \"vbios_%x_%x.bin\", vhdr-\u003eVendorID, vhdr-\u003eDeviceID);\n\n            if (!(fp_vbios = fopen(vbios_name, \"wb\"))) {\n                perror(argv[0]);\n                return -1;\n            }\n\n            if (vhdr-\u003eImageLength != fwrite(\u0026vbios-\u003eVbiosContent, 1, vhdr-\u003eImageLength, fp_vbios)) {\n                fprintf(stderr, \"%s: failed to dump vbios %x:%x\\n\", argv[0], vhdr-\u003eVendorID, vhdr-\u003eDeviceID);\n                return -1;\n            }\n\n            fclose(fp_vbios);\n\n            printf(\"dump vbios %x:%x to %s\\n\", vhdr-\u003eVendorID, vhdr-\u003eDeviceID, vbios_name);\n\n            offset += sizeof(VFCT_IMAGE_HEADER);\n            offset += vhdr-\u003eImageLength;\n        }\n\n        return 0;\n    }\n    ```\n\n    \u003c/details\u003e\n3. Get the `vbios` binary by compiling and running `vbios.c`:\n    ```\n    gcc vbios.c -o vbios\n    ./vbios\n    ```\n0. Move the `vbios_*.bin` vbios file to `/usr/share/kvm/vbios_7xxx`:\n    ```\n    mv vbios_*.bin /usr/share/kvm/vbios_7xxx.bin\n    ```\n4. In the proxmox web UI, click on the windows VM, Hardware, Add, PCI Device:\n    - Raw device: pick the PCI ID that we identified on the first steps, in my case its `0000:34:00.0`\n    - All Functions: OFF\n    - Primary GPU: OFF\n    - PCI-Express: ON\n5. Do the same for the Audio device, in my case its `0000:34:00.1`\n6. Set the correct BIOS for the GPU:\n    - Edit `/etc/pve/qemu-server/\u003cVM_ID\u003e.conf`\n    - Modify `args` to hide virtualization from the guest\n    - Modify the `hostpci` line for the GPU\n    ```diff\n    +args: -cpu 'host,-hypervisor,kvm=off'\n    agent: 1\n    balloon: 2048\n    bios: seabios\n    boot: order=ide0;ide2;net0\n    cores: 8\n    cpu: host\n    -hostpci0: 0000:34:00.0,pcie=1\n    +hostpci0: 0000:34:00.0,pcie=1,romfile=vbios_7xxx.bin\n    hostpci1: 0000:34:00.1,pcie=1\n    ide0: local-lvm:vm-100-disk-0,discard=on,size=64G,ssd=1\n    ide2: local:iso/Windows10.iso,media=cdrom,size=4697792K\n    machine: pc-q35-8.0\n    memory: 12048\n    meta: creation-qemu=8.0.2,ctime=1696067822\n    name: win10\n    net0: e1000=E2:4A:E7:86:8D:13,bridge=vmbr0,firewall=1\n    numa: 0\n    ostype: win10\n    scsihw: virtio-scsi-single\n    sockets: 1\n    ```\n7. Run the VM and install [the most recent VirtIO drivers](https://pve.proxmox.com/wiki/Windows_VirtIO_Drivers) (virtio-win-guest-tools.exe).\n7. Also install [the official AMD GPU drivers](https://www.amd.com/en/support/apu/amd-ryzen-processors/amd-ryzen-7-processors-radeon-graphics/amd-ryzen-7-7735hs). **Use the OFFLINE installer**, the online installer will complain that the computer is not an official AMD computer.\n7. Install [RadeonResetBugFix](https://oomza.cutegay.software/inga-lovinde/RadeonResetBugFix/releases) service to make sure the GPU can be transferred properly to the host after stopping the VM. If this is not done, you will suffer from the famous \"AMD Reset Bug\".\n   \n# Using the GPU as the Primary GPU\n\nNow that we have all the drivers ready, we can enable the GPU as the Primary GPU:\n   \n1. Enable Remote desktop in the VM (In Windows: Remote Desktop Settings -\u003e Enable Remote Desktop)\n8. Change the PC name to the same name as in proxmox to use it for the remote connection\n8. Shut down the windows VM\n0. Edit the VM Hardware again:\n    - Change `display` to `none`\n    - Make the GPU PCI device the `Primary GPU`\n9. Start the VM again and login in to it via Remote Desktop. Alternatively, you can also plug a monitor and you should see the VM there, passthrough the USB devices for keyboard and mouse and you have a fully working virtualized PC.\n   \n# (optional) Getting OVMF (UEFI) BIOS working: Error 43\n\nIf you tried to follow the guide but instead of SeaBIOS you selected UEFI, you have probably encountered the famous \"ERROR 43\". Luckily the solution for this is quite simple: **configuring the UEFI ROM for the audio device**.\n   \n1. Download `AMDGopDriver.rom` from this repository. Note that this AMDGopDriver.rom might not be compatible with certain hardware. If it fails to work, consider extracting it yourself. A brief guide is available [here](https://gist.github.com/matt22207/bb1ba1811a08a715e32f106450b0418a?permalink_comment_id=4955044#gistcomment-4955044).\n2. Copy the file inside the proxmox machine, in `/usr/share/kvm/AMDGopDriver.rom`\n3. Edit `/etc/pve/qemu-server/\u003cVM_ID\u003e.conf`\n    - Modify the `hostpci` line for the Audio Device and append `,romfile=AMDGopDriver.rom`\n    ```diff\n    args: -cpu 'host,-hypervisor,kvm=off'\n    agent: 1\n    balloon: 2048\n    bios: ovmf\n    boot: order=ide0;ide2;net0\n    cores: 8\n    cpu: host\n    hostpci0: 0000:34:00.0,pcie=1,romfile=vbios_7xxx.bin\n    -hostpci1: 0000:34:00.1,pcie=1\n    +hostpci1: 0000:34:00.1,pcie=1,romfile=AMDGopDriver.rom\n    ide0: local-lvm:vm-100-disk-0,discard=on,size=64G,ssd=1\n    ide2: local:iso/Windows10.iso,media=cdrom,size=4697792K\n    machine: pc-q35-8.0\n    memory: 12048\n    meta: creation-qemu=8.0.2,ctime=1696067822\n    name: win10\n    net0: e1000=E2:4A:E7:86:8D:13,bridge=vmbr0,firewall=1\n    numa: 0\n    ostype: win10\n    scsihw: virtio-scsi-single\n    sockets: 1\n    ```\n9. Start the VM again and login via Remote Desktop. Opening \"Device Manager\" should show the GPU working properly. If you still see error 43, try rebooting the host :)\n\n# Results\n\n![image](https://github.com/isc30/UM773-Lite-Proxmox/assets/10807051/cac210d8-420c-4a1b-9f3b-e9af5d8d0521)\n\n- Win 10:\n  ![image](https://github.com/isc30/UM773-Lite-Proxmox/assets/10807051/74c9423f-875c-4b73-b4d2-b4e742777812)\n\n- Win 11:\n  ![image](https://github.com/isc30/UM773-Lite-Proxmox/assets/10807051/b06415f0-fd4b-488f-9ac7-88c3ad0af6e8)\n\n# Known issues\n\n- ## When entering via Remote Desktop, the GPU is disabled\n  I'm unsure why this happens, but it seems like `Device Manager -\u003e right-click -\u003e Enable Device` makes it work again\n\n- ## In random situations, I still get \"error 43\" when trying to initialize the GPU in the VM\n  Probably related to the \"amd reset issue\", that prevents the GPU from binding to a VM after it was used once. The only **\"real\"** solution for this is to restart the proxmox host after stopping a VM that used the GPU. :sad:\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fisc30%2Fryzen-gpu-passthrough-proxmox","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fisc30%2Fryzen-gpu-passthrough-proxmox","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fisc30%2Fryzen-gpu-passthrough-proxmox/lists"}