-
Bug
-
Resolution: Done
-
Undefined
-
None
-
None
-
False
-
-
False
-
-
Expected behavior
- Update the RHEL AI 1.5 Hardware requirements section for IBM Cloud to add AMD GPU and Intel Gaudi 3 instances.
Updated content
- End-to-end workflow
1.1. Hardware requirements for end-to-end workflow of Granite models
1.1.1. Bare metal
Updated Table:
| Hardware Vendor | Supported accelerators (GPUs) | Aggregate GPU Memory | IBM Cloud instances | Recommended additional disk storage | Change Requested |
| NVIDIA | 2xA100 | 160 GB | gx3d-48x240x2a100p | 3TB | Already exists |
| NVIDIA | 8xH100 | 640 GB | gx3d-160x1792x8h100 | 3 TB | Already exists |
| NVIDIA | 8xH200 | 1128 GB | gx3d-160x1792x8h200 | 3 TB | Already exists |
| AMD | 8xMI300X | 1536 GB | gx3d-208x1792x8mi300x | 3 TB | Add |
- Inference
3.1 Hardware requirements for inference serving Granite models
3.2.2 IBM Cloud
Updated Table:
| Hardware Vendor | Supported accelerators (GPUs) | Minimum Aggregate GPU Memory | IBM Cloud instances | Recommended additional disk storage | Change Requested |
| NVIDIA | L4 | 24 GB | gx3 series | 1 TB | Already exists |
| NVIDIA | L40S | 48 GB | gx3 series | 1 TB | Already exists |
| NVIDIA | A100 | 80 GB | gx3 series | 1 TB | Already exists |
| NVIDIA | H100 | 80 GB | gx3 series | 1 TB | Already exists |
| NVIDIA | H200 | 141 GB | gx3 series | 1 TB | Already exists |
| AMD | MI300X | 192 GB | gx3 series | 1 TB | Add |
| INTEL | Gaudi3 (Technology Preview) | 128 GB | gx3 series | 1TB | Add |
