AMD MI300X
Last updated on Nov. 27, 2024
Price Comparison
| Provider | GPUs | VRAM | vCPUs | RAM | Price/h | |
|---|---|---|---|---|---|---|
RunPod
|
1x MI300X | 192GB | 24 | 283GB | $3.49 | Source |
Oracle Cloud
|
8x MI300X | 1536GB | 56 | 2000GB | $6.00 | Source |
TensorWave
|
8x MI300X | 1536GB | -- | -- | On Request | |
CUDO Compute
|
1x MI300X | 192GB | -- | -- | On Request | Source |
Note: Prices are subject to change, and may vary by region and other factors not listed here. I included links to launch GPUs using Shadeform (sponsor) so you can check if they're available right now. There are no fees to use their service.
Specifications
| TF32 (TFLOPs) | 653.7 |
| FP16 (TFLOPs) | 1307.4 |
| BFLOAT16 (TFLOPs) | 1307.4 |
| INT8 (TOPS) | 2614.9 |
| FP8 (TFLOPs) | 2614.9 |
| FP64 vector | 81.7 |
| FP32 vector | 163.4 |
| FP64 matrix | 163.4 |
| FP32 matrix | 163.4 |
| Decoders | 4 groups for HEVC/H.265, AVC/H.264, V1, or AV1 |
| JPEG/MJPEG CODEC | 32 cores, 8 cores per group |
| Virtualization support | SR-IOV, up to 8 partitions |
| Form factor | OAM module |
| Lithography | 5nm FinFET, 6nm FinFET |
| GPU compute units | 304 |
| Matrix cores | 1216 |
| Stream processors | 19,456 |
| Peak engine clock | 2100 MHz |
| Memory capacity | Up to 192 GB HBM3 |
| Memory bandwidth | 5.3 TB/s max. peak theoretical |
| Memory interface | 8192 bits |
| AMD Infinity Cache™ (last level) | 256 MB |
| Memory clock | Up to 5.2 GT/s |
| Scale-up Infinity Fabric™ Links | 7x 128 GB/s |
| I/O to host CPU | 1 PCIe® Gen 5 x16 (128 GB/s) |
| Scale-out network bandwidth | PCIe Gen 5 x16 (128 GB/s) |
| RAS features | Full-chip ECC memory, page retirement, page avoidance |
| Maximum TBP | 750W |
Source: official AMD MI300X datasheet.