Brand: AMD | Category: GPUs
SKU: AMD-100300000080 | Part #: 100-300000080 | MPN: 100-300000080
Contact for Pricing — Request a Quote
The AMD Instinct MI325X (8-GPU platform) from AMD is enterprise-grade GPUs hardware built for data centers. Eight-way AMD Instinct MI325X baseboard delivering 2TB aggregate HBM3E memory for trillion-parameter AI model deployment at scale. 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-100300000080 |
| Part Number | 100-300000080 |
| Condition | New |
| Product Family | AMD Instinct MI325X |
| Platform Configuration | 8-GPU OAM Baseboard |
| GPU Architecture | AMD CDNA3 |
| GPUs per Platform | 8x AMD Instinct MI325X |
| Memory per GPU | 288 GB HBM3E |
| Aggregate Platform Memory | 2.304 TB HBM3E |
| Memory Bandwidth per GPU | 6.0 TB/s (peak) |
| Aggregate Memory Bandwidth | 48 TB/s (peak) |
| Peak AI Compute (FP8) per GPU | ~655 TOPS |
| Aggregate Peak AI Compute (FP8) | >5,200 TOPS |
| Peak FP16 Compute per GPU | ~327.7 TFLOPS |
| Peak FP64 Compute per GPU | ~81.9 TFLOPS |
| Interconnect Technology | AMD Infinity Fabric (GPU-to-GPU, coherent) |
| Inter-GPU Bandwidth | Up to 896 GB/s bidirectional per GPU link |
| Form Factor | OCP Accelerator Module (OAM) Baseboard |
| Host Interface | PCIe Gen 5 x16 per GPU (host uplink) |
| Supported Precisions | FP8, BF16, FP16, FP32, FP64, INT8, INT4 |
| Software Stack | AMD ROCm (HIP, MIOpen, rocBLAS, rocFFT), PyTorch, TensorFlow, JAX, vLLM |
| Thermal Design | Liquid cooling required; OAM chassis direct liquid cooling (DLC) compatible |
| Launch | Q1 2025 |