Releases: amd/MxGPU-Virtualization
GIM 8.7.1.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
Along with regular bug fixes and improvements, this release βexpands KVM SR-IOV virtualization support for MI325X on RHEL 10β and enables ROCm profiler tool support within KVM SR-IOV virtualized (Guest) GPU environments..
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 Kernel 6.14 | RHEL 9.6, RHEL 10.0, Ubuntu 24.04 | ROCm 7.2.1 | 1 | between 8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 Kernel 6.14 | RHEL 9.6, RHEL 10.0, Ubuntu 24.04 | ROCm 7.2.1 | 1 | between 8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 Kernel 6.8 | Ubuntu 22.04 | ROCm 7.2.1 | 1 | between 8 GPUs |
| AMD Instinct MI325X | RHEL 10.1 Kernel 6.12.0-124 | RHEL 10.1 | ROCm 7.2.1 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 Kernel 6.8 | Ubuntu 22.04 | ROCm 7.2.1 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 Kernel 5.14.0-427 | RHEL 9.4, Ubuntu 24.04 | ROCm 7.2.1 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 Kernel 5.14.0-427 | RHEL 9.4, Ubuntu 22.04 | ROCm 7.2.1 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 Kernel 6.14 | Ubuntu 24.04 | ROCm 7.2.1 | 1 or 4 or 12 | N/A |
Partition Mode Support
| Number of VFs per GPU | Dynamic Compute Partitioning | NPS1 | NPS2 | NPS4 |
|---|---|---|---|---|
| 1 | SPX (Default) | MI300X, MI325X, MI35XX | ||
| 1 | DPX | MI35XX | ||
| 1 | CPX | MI35XX (Preview) | ||
| 8* | CPX | MI300X (Preview) |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
* Multi-VF (8 VF) Support requires a compatible firmware BKC bundle and ROCm which will be released in coming months.
Please refer to the ROCm Firmware Compatibility Table for the list of required firmware versions and supported configurations.
Host AMD SMI Versions
- Library: 51.0.0
- Tool: 34.1.1
Please checkout out SMI doc for more details.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
- MI300: MultiVF configuration for MI300X requires a compatible firmware BKC bundle and ROCm which will be released in coming months. Please refer to the ROCm compatibility matrix for supported version
- MI210: Unloading the Guest driver on MI210 may produce kernel call trace. Note that this is just a warning and not a kernel panic. Guest VM reboot is required to restore the operation β
- Navi32: AMDSMI CLI --gpureset command on Nv32 doesn't reset all the GPUs like the output is saying, but just GPU 0.
GIM 8.7.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
This release has regular bug fixes and improvements, including MI350/MI355 Node Power Management for multi-GPU nodes (for more info see ROCm release notes) and optimization for MI350/MI355 NPS2+DPX.
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 Kernel 6.14 | RHEL 9.6, RHEL 10.0, Ubuntu 24.04 | ROCm 7.2 | 1 | between 8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 Kernel 6.14 | RHEL 9.6, RHEL 10.0, Ubuntu 24.04 | ROCm 7.2 | 1 | between 8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 Kernel 6.8 | Ubuntu 22.04 | ROCm 7.2 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 Kernel 6.8 | Ubuntu 22.04 | ROCm 7.2* | 1 or 8* | between 8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 Kernel 5.14.0-427 | RHEL 9.4, Ubuntu 24.04 | ROCm 7.2* | 1 or 8* | between 8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 Kernel 5.14.0-427 | RHEL 9.4, Ubuntu 22.04 | ROCm 7.2 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 Kernel 6.14 | Ubuntu 24.04 | ROCm 7.2 | 1 or 4 or 12 | N/A |
Partition Mode Support
| Number of VFs per GPU | Dynamic Compute Partitioning | NPS1 | NPS2 | NPS4 |
|---|---|---|---|---|
| 1 | SPX (Default) | MI300X, MI325X, MI35XX | ||
| 1 | DPX | MI35XX | ||
| 1 | CPX | MI35XX (Preview) | ||
| 8* | CPX | MI300X (Preview) |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
* Multi-VF (8 VF) Support requires a compatible firmware BKC bundle for the GPU.
Please refer to the ROCm Firmware Compatibility Table for the list of required firmware versions and supported configurations.
Host AMD SMI Versions
- Library: 51.0.0
- Tool: 34.1.1
Please checkout out SMI doc for more details.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
- MultiVF configuration for MI300X requires specific firmware version. please refer to the ROCm compatibility matrix for supported version
- MI210: Unloading the Guest driver on MI210 may produce kernel call trace. Note that this is just a warning and not a kernel panic. Guest VM reboot is required to restore the operation β
GIM 8.6.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
This release has regular bug fixes and improvements.
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 | RHEL 9.6, RHEL 10, Ubuntu 24.04 | ROCm 7.1.1 | 1 | between 8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 | RHEL 9.6, RHEL 10, Ubuntu 24.04 | ROCm 7.1.1 | 1 | between 8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.1.1 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.1.1* | 1 or 8* | between 8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4, Ubuntu 24.04 | ROCm 7.1.1* | 1 or 8* | between 8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.4, Ubuntu 22.04 | ROCm 7.1.1 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.1.1 | 1 or 4 or 12 | N/A |
Partition Mode Support
| Number of VFs per GPU | Dynamic Compute Partitioning | NPS1 | NPS2 | NPS4 |
|---|---|---|---|---|
| 1 | SPX (Default) | MI300X, MI325X, MI35XX | ||
| 1 | DPX | MI35XX | ||
| 1 | CPX | MI35XX (Preview) | ||
| 8* | CPX | MI300X (Preview) |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
* Multi-VF (8 VF) Support requires a compatible firmware BKC bundle for the GPU.
Please refer to the ROCm Firmware Compatibility Table for the list of required firmware versions and supported configurations.
Host AMD SMI Versions
- Library: 50.2.2
- Tool: 32.4.1
Please checkout out SMI doc for more details.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
- When using the configuration of 64 ROCM devices in 1 VM (aka CPX mode with 1 VF) on a MI350 platform, we would see a hang in workloads which use SDMA engines
- MultiVF configuration for MI300X requires specific firmware version. please refer to the ROCm compatibility matrix for supported version
- MI210: Unloading the Guest driver on MI210 may produce kernel call trace. Note that this is just a warning and not a kernel panic. Guest VM reboot is required to restore the operation β
- SMI CLI topology --link_type command returns wrong value for the first BDFβ
- SMI CLI static --vram command returns 0 for the bit width valueβ
- Some BDF values returned from SMI get_link_topology_nearest APIs are wrong
GIM 8.5.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
This release adds MI350 CPX support and SMI CLI command improvement on top of the regular update and bug fixes.
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 | RHEL 9.6, RHEL 10, Ubuntu 24.04 | ROCm 7.1 | 1 | between 8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 | RHEL 9.6, RHEL 10, Ubuntu 24.04 | ROCm 7.1 | 1 | between 8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.1 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.1 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4 | ROCm 7.1 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.4, Ubuntu 22.04 | ROCm 7.1 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.1 | 1 or 4 or 12 | N/A |
Partition Mode Support
| Number of VFs per GPU | Dynamic Compute Partitioning | NPS1 | NPS2 | NPS4 |
|---|---|---|---|---|
| 1 | SPX (Default) | MI300X, MI325X, MI35XX | ||
| 1 | DPX | MI35XX | ||
| 1 | CPX | MI35XX (Preview) | ||
| 8* | CPX | MI300X (Preview) |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
* Multi-VF (8 VF) Support requires a compatible firmware BKC bundle for the GPU.
Please refer to the ROCm Firmware Compatibility Table for the list of required firmware versions and supported configurations.
Host AMD SMI Versions
- Library: 40.0.6
- Tool: 30.0.3
Please checkout out SMI doc for more details.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
With NPS1/SPX and 8VMs (1VF per VM) on MI300, triggering reset via "/sys/kernel/debug/kfd/hang_hws" causes the GPU to become unusable. Rebooting the guest should solve the problem.
GIM 8.4.1.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
This release adds VF FLR improvement on Navi32 as well as MI300 multi Vf on top of the regular update and bug fixes.
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0.2 | 1 | between 8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0.2 | 1 | between 8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.0.2 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.0.2 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4 | ROCm 7.0.2 | 1 or 8* | between 8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.4, Ubuntu 22.04 | ROCm 7.0.2 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0.2 | 1 or 4 or 12 | N/A |
Partition Mode Support
| Number of VFs per GPU | Dynamic Compute Partitioning | NPS1 | NPS2 | NPS4 |
|---|---|---|---|---|
| 1 | SPX (Default) | MI300X, MI325X, MI35XX | ||
| 1 | DPX | MI35XX (preview) | ||
| 1 | CPX | |||
| 8* | CPX | MI300X (preview) |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
* Multi-VF (8 VF) Support requires a compatible firmware BKC bundle for the GPU.
Please refer to the ROCm Firmware Compatibility Table for the list of required firmware versions and supported configurations.
Host AMD SMI Versions
- Library: 33.1.3
- Tool: 27.8.0
Since the last release, the library and tool will be automatically installed when installing the gim-dkms packageβmanual build step is no longer required.
If you manually copied AMD SMI tool/library to /usr, you may need to reset the ldconfig cache and/or remove the AMD SMI files under /usr.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
- Ubuntu kernel
6.8.0-64and6.14.0-24and possibly other kernels on other distros have a regression.
Please update to a later kernel that contains the fix. - For V710, the VF active time reported by
amd-smiis not accurate.
GIM 8.4.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
This release adds dynamic partitioning support as well as V710 validation on top of the regular update and bug fixes.
Hardware / Features Supported for this Release
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI355X | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0 | 1 | between 2/4/8 GPUs |
| AMD Instinct MI350X | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0 | 1 | between 2/4/8 GPUs |
| AMD Instinct MI325X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.0 | 1 | between 8 GPUs |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 7.0 | 1 | between 2/4/8 GPUs |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4 | ROCm 7.0 | 1 | between 2/4/8 GPUs |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.5, Ubuntu 22.04 | ROCm 7.0 | 1 | between 4/8 GPUs |
| AMD Radeon PRO V710 | Ubuntu 24.04 | Ubuntu 24.04 | ROCm 7.0 | 1 / 4 / 12 | N/A |
Partition Mode Support
| Partition Mode | SPX | DPX | QPX | CPX |
|---|---|---|---|---|
| NPS1 | MI35XX, MI325X, MI300X | |||
| NPS2 | MI35XX (preview) | |||
| NPS4 | MI325X, MI300X |
For the full list of supported partition modes and the relationship between partition mode and number of VFs per GPU, see the GPU Partitioning.
Host AMD SMI Versions
- Library: 33.1.3
- Tool: 27.8.0
Since the last release, the library and tool will be automatically installed when installing the gim-dkms packageβmanual build step is no longer required.
If you manually copied AMD SMI tool/library to /usr, you may need to reset the ldconfig cache and/or remove the AMD SMI files under /usr.
Getting Started
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known Issues
- Ubuntu kernel
6.8.0-64and6.14.0-24and possibly other kernels on other distros have a regression.
Please update to a later kernel that contains the fix. - For V710, the VF active time reported by
amd-smiis not accurate.
GIM 8.3.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver. This release adds MI325 validation on top of the regular update and bug fixes.
Hardware/Features tested for this release:
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI350X | Ubuntu 24.04 | Ubuntu 24.04 | To be announced | 1 | 1/2/4/8 VF in a single VM |
| AMD Instinct MI325X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 6.4 | 1 | 1/8 VF in a single VM |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 6.4 | 1 | 1/2/4/8 VF in a single VM |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4 | ROCm 6.4 | 1 | 1/2/4/8 VF in a single VM |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.5, Ubuntu 24.04 | ROCm 6.4 | 1 | 1/4/8 VF in a single VM |
Host AMD SMI Versions:
Library: 30.0.1
Tool: 27.4.0
Starting with this release, the library and tool will be automatically installed when installing gim-dkms package, manual build step is no longer required. If you had manually copied amd smi tool/library to /usr, you may need to reset ldconfig cache and/or remove the amd smi files under /usr.
Getting Started:
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known issues:
The -follow flag for amd-smi ras command is not working, to query CPERs run the command without the --follow flag.
GIM 8.2.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver. Besides regular update and bug fixes, this release also contains development preview for MI350X.
Hardware/Features tested for this release:
| Hardware | Host OS | Guest OS | ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|---|
| AMD Instinct MI350X | Ubuntu 24.04 | Ubuntu 24.04 | To be announced | 1 | 8 VF in a single VM |
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04 | ROCm 6.4 | 1 | 1/2/4/8 VF in a single VM |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4 | ROCm 6.4 | 1 | 8 VF in a single VM |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.5, Ubuntu 24.04 | ROCm 6.4 | 1 | 1/4/8 VF in a single VM |
Host AMD SMI Versions:
Library: 28.5.0
Tool: 25.5.0
Getting Started:
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known issues:
amdsmi_clear_vf_fb does not clear vf fb buffer.
GIM 8.1.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver. The SMI library version for this release is 28.3.2, and the smi tool version is 25.3.1
Hardware/Features supported by this release:
| Hardware | Supported Host OS | Supported Guest OS/ROCm version | Number of VFs per GPU | Infinity Fabric (XGMI) Configuration |
|---|---|---|---|---|
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04/ROCm 6.4 | 1 | 1/2/4/8 VF in a single VM |
| AMD Instinct MI300X | RHEL 9.4 | RHEL 9.4/ROCm 6.4 | 1 | 8 VF in a single VM |
| AMD Instinct MI210X | RHEL 9.4 | RHEL 9.5, Ubuntu 24.04/ROCm 6.4 | 1 | 1/4/8 VF in a single VM |
Getting Started:
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known issues:
SMI API amdsmi_clear_vf_fb does not clear vf fb buffer.
Driver reload could trigger a rare race condition that causes SMI to show wrong XGMI topology.
GIM 8.0.0.K Release Note
We are excited to announce a new open-source release of our GIM SR-IOV Driver.
Hardware/Features supported by this release:
| Hardware | Supported Host OS | Supported Guest OS/ROCm version | Number of VFs per GPU |
|---|---|---|---|
| AMD Instinct MI300X | Ubuntu 22.04 | Ubuntu 22.04/ROCm 6.4 | 1 |
Getting Started:
Please check out our User Guide for instructions on how to set up GIM and example configurations to run SR-IOV enabled VMs.
Known issues:
SMI API amdsmi_clear_vf_fb does not clear vf fb buffer