Devices supported by HAMi
Applicable version: HAMi v2.9.0.
Last verified: 2026-11-06, against the HAMi source, default device configuration, and this version’s device guides.
The table below summarizes device integration, allocation modes, and multi-card partitioning capabilities for HAMi v2.9.0. See the linked guides for plugin, driver, and runtime requirements.
| Device | Mode | Models | Integration status | Memory control | Compute control | Multi-card partitioning |
|---|
| NVIDIA GPU | HAMi-core | All | Stable | Memory quota | Time-slice throttling | Yes |
| NVIDIA GPU | MIG (dynamic partitioning) | MIG-capable models | Stable | Hardware profile | Hardware profile | Yes (subject to CUDA and driver requirements) |
| AMD GPU | Whole device | MI300X | Stable | Whole-device allocation | Whole-device allocation | N/A |
| AWS Neuron | Whole device / NeuronCore allocation | Inferentia / Trainium | Stable | No separate memory quota | NeuronCore allocation | No (multi-device requests allocate all cores of each device) |
| Cambricon MLU | Dynamic SMLU | 370 / 590 | Stable | Memory quota | Core quota | No |
| Enflame GCU | GCUshare (percentage slicing) | S60 | Stable | Shared memory/compute percentage | Shared memory/compute percentage | No |
| Ascend NPU | template | 910A, 910B2, 910B3, 910B4-1, 910B4, 310P | Stable | Template allocation | Template allocation | No |
| Ascend NPU | hami-core | 910A, 910B, 310P, 910C (ARM) | Stable | Memory quota | Compute quota | Yes |
| Ascend NPU | Whole device | 910C | Stable | Whole-device allocation | Whole-device allocation | N/A |
| Hygon DCU | Dynamic vDCU slicing | All | Stable | Memory quota | Compute share | No |
| Iluvatar GPU | vGPU (gpu-manager) | MR-V100, MR-V50, BI-V150, BI-V100 | Stable | Memory quota | Core quota | No |
| Kunlunxin XPU | VXPU | P800-OAM | Stable | Fixed partition sizes | Determined by partition size | Yes |
| MetaX GPU | sGPU | MXC500 | Stable | Memory quota | Depends on QoS policy | Yes (not with sgpu-app-class) |
| Mthreads GPU | sGPU | MTT S4000 | Stable | Discrete memory sizes | Core quota | No |
| Vastai GPU | Whole device / Die slicing | VA16 | Stable | Whole-device or Die allocation | Whole-device or Die allocation | Yes (Die allocation) |
Stable means the integration is included in this release.
Capability definitions:
- Memory control: The memory quota or partition size a container can request. Templates and fixed partitions round requests up to a supported size.
- Compute control: Devices control compute resources through time slicing, core allocation, hardware partitions, or QoS policies. The linked guides describe the corresponding limits.
- Multi-card partitioning: A single container can allocate resources across multiple physical devices while using only part of at least one device. Allocating multiple whole devices does not count as multi-card partitioning. See the linked references for allocation granularity and limits.
N/A means the mode allocates whole devices only.