amd-amd-instinct-mi350x-1254
Source details
Source revision: 01ff41b8fbbd848af24e0e6926c89d424d7a051f9014db0fc96997a566b5c8b8
AMD · Data center · GPU
Exact AMD Instinct MI350X reference specifications and AI deployment guidance, separated from FusionPointAI commercial inventory.
Product images
| Field | Value |
|---|---|
| Manufacturer | AMD |
| Exact SKU | AMD Instinct MI350X |
| Product family | Instinct |
| Architecture | CDNA4 |
| Generation | Not listed |
| Entity type | GPU |
| Lifecycle | Unknown |
| Launch date | Not listed |
| VRAM | 288 GB |
| Memory technology | HBM3E |
| Memory bandwidth | 8,000 GB/s |
| Power | Not listed |
| Interface | Not listed |
| Form factor | OAM · Module |
| Interconnect | Not listed |
| MIG / partitioning | Not supported |
| Supported numeric formats | Not listed |
| Software stack | Not listed |
High memory capacity supports large checkpoints and multi-GPU inference when runtime and topology are certified.
Designed for sustained accelerator workloads; training suitability depends on interconnect, precision and cluster topology.
Confirm framework and precision support before deployment.
The manufacturer does not publish a directly comparable board-power value for this entity.
No dedicated accelerator interconnect is asserted from the available manufacturer evidence.
Use only officially documented drivers and deployment runtimes.
Source revision: 01ff41b8fbbd848af24e0e6926c89d424d7a051f9014db0fc96997a566b5c8b8
This is not a product photograph.
https://www.amd.com/en/products/accelerators/instinct/mi350/mi350x.html
Last verified: 2026-11-04