Brand: AMD | Category: GPUs
SKU: AMD-100000001236 | Part #: 100-000001236 | MPN: 100-000001236
Contact for Pricing — Request a Quote
The AMD Instinct MI300X Platform (8-GPU OAM baseboard) from AMD is enterprise-grade GPUs hardware built for data centers. AMD Instinct MI300X is a 192GB HBM3 accelerator platform engineered for large-scale generative AI inference and distributed deep learning training workloads. Available in brand new condition. AMD gpus products are trusted by enterprises and data centers worldwide. Available from Omnixon Global. Contact our sales team to request a quote.
| Brand | AMD |
| Category | GPUs |
| SKU | AMD-100000001236 |
| Part Number | 100-000001236 |
| Condition | New |
| Product Family | AMD Instinct MI300X |
| Platform Configuration | 8-GPU OAM Baseboard |
| GPU Architecture | AMD CDNA 3 |
| Accelerator Package Technology | 3D chiplet packaging (8 compute dies + 12 HBM3 stacks per OAM) |
| Compute Dies per OAM | 8 GPU compute dies (GCDs) |
| Memory per Accelerator (OAM) | 192 GB HBM3 |
| Total Platform Memory (8 OAM) | 1,536 GB HBM3 |
| Memory Bandwidth per Accelerator | 5.2 TB/s |
| Total Platform Memory Bandwidth | 41.6 TB/s |
| Peak FP8 Throughput per Accelerator | 383 TFLOPS |
| Peak BF16/FP16 Throughput per Accelerator | 383 TFLOPS |
| Peak FP32 Throughput per Accelerator | 163.4 TFLOPS |
| Peak FP64 Throughput per Accelerator | 81.7 TFLOPS |
| Supported Precisions | FP8, FP16, BF16, TF32, FP32, FP64, INT8 |
| Host Interface | PCIe Gen 5 x16 per OAM |
| GPU-to-GPU Interconnect | AMD Infinity Fabric (peer-to-peer, on-baseboard) |
| Compute Units per Accelerator | 304 Compute Units |
| Stream Processors per Accelerator | 19,456 |
| TDP per Accelerator (OAM) | 750W |
| Form Factor | OAM (Open Accelerator Module) baseboard, compatible with OCP OAM v1.0 specification |
| Supported Host Processors | AMD EPYC 7th Gen (Genoa) and compatible platforms |
| Software Ecosystem | AMD ROCm (PyTorch, TensorFlow, JAX, HIP, OpenCL, MIOpen) |
| Deployment Target | Hyperscale AI inference clusters, large-scale model training infrastructure, HPC data centers |