AMD Instinct MI300X 192GB HBM3 OAM
AMD Instinct MI300X offers 192GB HBM3 and 5.3TB/s bandwidth. It can be useful for memory-heavy inference when the software stack supports ROCm. Quoted per project with availability confirmed in writing.
Price Available Upon Request
We confirm the configuration, price, delivery estimate, and warranty terms before you order.
What it's for
The AMD Instinct MI300X has 192GB of HBM3 memory and uses the OAM form factor in supported server platforms. It is an option for memory-heavy training and inference when the workload runs well on ROCm. We quote the exact system, software requirements, and delivery terms for each project.
Key Features
- ✓192GB HBM3 memory per accelerator
- ✓5.3 TB/s memory bandwidth
- ✓304 Compute Units with 19,456 Stream Processors
- ✓ROCm software ecosystem
- ✓AMD Infinity Fabric for multi-GPU scaling
- ✓Eight-GPU platform options through system partners
- ✓OCP Accelerator Module (OAM) form factor
Use Cases
- →Large Language Model training (Llama, Mistral, GPT)
- →Generative AI inference at scale
- →Multi-modal AI models (text, image, video)
- →Scientific computing and research
- →Drug discovery and molecular dynamics
- →Financial modeling and risk analysis
Technical Specifications
| Architecture | CDNA 3 |
| GPU Memory | 192 GB HBM3 |
| Memory Bandwidth | 5.3 TB/s |
| Memory Interface | 8192-bit |
| Compute Units | 304 CUs |
| Stream Processors | 19,456 |
| FP64 Performance | 163 TFLOPS |
| FP32 Performance | 653 TFLOPS |
| FP16 Performance | ~1,307 TFLOPS |
| BF16 Performance | ~1,307 TFLOPS |
| INT8 Performance | ~2,614 TOPS |
| Max TDP | 750W |
| Thermal Solution | Air or liquid, depending on OEM system |
| Form Factor | OAM (OCP Accelerator Module) |
| Interconnect | Infinity Fabric |
Manufacturer references
Related Products
NVIDIA H100 80GB PCIe Gen5
NVIDIA H100 PCIe accelerator with 80GB of HBM2e. Fits supported PCIe server platforms for training and inference. Request a current configuration and quote.
Current quote on request
NVIDIA H100 80GB SXM5
NVIDIA H100 SXM GPU with 80GB HBM3, used in HGX server platforms. Request a complete OEM system quote with current delivery terms.
Current quote on request
NVIDIA H200 141GB HBM3e SXM5
Hopper GPU with 141GB HBM3e and 4.8TB/s bandwidth. A useful option for memory-heavy inference. Contact us for current availability and lead time.
Current quote on request
NVIDIA B200 180GB SXM
Blackwell GPU with 180GB HBM3e per GPU in 8-GPU HGX systems. Ask us for a complete OEM server quote with current availability and lead time.
Current quote on request
Recommended Reading
Related buying guides
Background on memory, power, cooling, and deployment planning.
AMD MI350X vs NVIDIA B200: What a Buyer Should Compare
Compare MI350X and B200 memory, software, OEM systems, cooling, and the details needed for a useful quote.
H100 vs H200: Choosing a Hopper Server
Compare H100 and H200 memory, bandwidth, server fit, and the details needed for current quotes.
H100 vs MI300X: Hardware and Software Buying Guide
Compare H100 and MI300X memory, CUDA and ROCm requirements, OEM systems, and quote terms.