Are you tired of using slow and outdated technology for your artificial intelligence and high-performance computing needs? Look no further than the AMD Instinct MI300X Platform, a powerful tool designed to help you take your work to the next level. With its 8-GPU AI and HPC baseboard, this system is perfect for anyone who needs to process large amounts of data quickly and efficiently. The AMD Instinct MI300X Platform is designed to solve real-world problems, such as slow processing times and limited memory capacity, making it an essential tool for anyone working in the fields of AI, HPC, and data science.
AMD Instinct MI300X Platform, 8-GPU AI & HPC Baseboard, 1.5TB HBM3 Server System
The AMD Instinct MI300X Platform solves the problem of slow processing times and limited memory capacity, which can be a major obstacle for anyone working in the fields of AI, HPC, and data science. This problem matters to users because it can slow down their work and limit their ability to process large amounts of data. The AMD Instinct MI300X Platform provides a clear solution to this problem by offering a powerful and efficient system that can handle even the most demanding workloads. With its advanced architecture and high-performance capabilities, this system is designed to help users get their work done quickly and efficiently.
Key Features
- 8-GPU AI and HPC baseboard
- 1.5TB HBM3 memory capacity
- 5nm FinFET lithography
- 6nm FinFET active interposer dies (AIDs)
- 2432 GPU compute units
- 9728 matrix cores
- 155,648 stream processors
- Peak engine clock of 2100 MHz
- Memory bandwidth of 5.3 TB/s max. peak theoretical
- Memory interface of 8192 bits per GPU
- AMD Infinity Cache (last level) of 256 MB per GPU
Specifications
| AI Peak Theoretical Performance with sparsity |
|
| TF32 |
5.2 PFLOPs, 10.5 PFLOPs |
| FP16 |
10.5 PFLOPs, 20.9 PFLOPs |
| BFLOAT16 |
10.5 PFLOPs, 20.9 PFLOPs |
| INT8 |
20.9 POPs, 41.8 POPs |
| FP8 |
20.9 PFLOPs, 41.8 PFLOPs |
| HPC Peak Theoretical Performance |
|
| FP64 vector |
653.7 TFLOPs |
| FP32 vector |
1307.4 TFLOPs |
| FP64 matrix |
1307.4 TFLOPs |
| FP32 matrix |
1307.4 TFLOPs |
| Decoders and Virtualization |
|
| Decoders* |
32 groups for HEVC/H.265, AVC/H.264, V1, or AV1 |
| JPEG/MJPEG CODEC |
256 cores, 8 cores per group |
| Virtualization support |
SR-IOV, up to 64 partitions |
| Specifications |
|
| Form factor |
Universal baseboard (UBB) module with 8 Instinct MI300X OAM GPUs |
| Lithography |
5nm FinFET |
| Active interposer dies (AIDs) |
6nm FinFET |
| GPU compute units |
2432 |
| Matrix cores |
9728 |
| Stream processors |
155,648 |
| Peak engine clock |
2100 MHz |
| Memory capacity |
1.5 TB HBM3 |
| Memory bandwidth |
5.3 TB/s max. peak theoretical |
| Memory interface |
8192 bits per GPU |
| AMD Infinity Cache™ (last level) |
256 MB per GPU |
| Memory clock |
Up to 5.2 GT/s |
| Scale-up Infinity Fabric™ Links |
7x 128 GB/s per GPU |
| Ring of 8 aggregate bandwidth |
896 GB/s |
| Scale-out network bandwidth |
8 PCIe® Gen 5 x16 (128 GB/s) per GPU |
| RAS features |
Full-chip ECC memory, page retirement, page avoidance |
| Maximum TBP |
750W per GPU |
