Brand: AMD | Category: GPUs
SKU: AMD-100000001077 | Part #: 100-000001077 | MPN: 100-000001077
Contact for Pricing — Request a Quote
The AMD Instinct MI300A is a monolithic APU (Accelerated Processing Unit) integrating a 12-core EPYC CPU, 104 CDNA 3 GPU cores, and 192 GB of HBM3 memory on a single chip. This unified architecture eliminates data movement penalties between CPU and GPU, delivering up to 5.3 petaFLOPS of peak FP8 performance and enabling memory bandwidth of 5.2 TB/s. The MI300A targets demanding HPC simulation codes, climate modeling, molecular dynamics, finite-element analysis, and large-scale AI model training where unified memory coherence and extreme compute density are critical competitive advantages.
| Manufacturer | AMD |
| Model | Instinct MI300A |
| GPU Architecture | CDNA 3 |
| GPU Cores | 104 |
| CPU Cores | 12 |
| CPU Architecture | EPYC (Bergamo-derived) |
| Memory Capacity | 192 GB |
| Memory Type | HBM3 |
| Memory Bandwidth | 5.2 TB/s |
| Peak FP8 Performance | 5.3 petaFLOPS |
| Peak FP32 Performance | 661 teraFLOPS |
| Peak FP64 Performance | 330.5 teraFLOPS |
| Tensor Performance (FP8) | 10.6 petaFLOPS dual-issue |
| TDP | 500 W |
| Process Node | 5 nm |
| PCIe Generation | 5.0 |
| Cooling | Passive/active liquid compatible |
| Launch Date | Q4 2023 |
| ECC Memory | Full ECC support |
| Unified Memory Architecture | Hardware-coherent CPU-GPU memory space |
| Max Power Delivery | 500 W single slot |
Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.
| Brand | AMD |
| Category | GPUs |
| SKU | AMD-100000001077 |
| Part Number | 100-000001077 |
| Condition | New |
| Model | Instinct MI300A |
| GPU Architecture | CDNA 3 |
| GPU Cores | 104 |
| CPU Cores | 12 |
| CPU Architecture | EPYC (Bergamo-derived) |
| Memory Capacity | 192 GB |
| Memory Type | HBM3 |
| Memory Bandwidth | 5.2 TB/s |
| Peak FP8 Performance | 5.3 petaFLOPS |
| Peak FP32 Performance | 661 teraFLOPS |
| Peak FP64 Performance | 330.5 teraFLOPS |
| Tensor Performance (FP8) | 10.6 petaFLOPS dual-issue |
| TDP | 500 W |
| Process Node | 5 nm |
| PCIe Generation | 5.0 |
| Cooling | Passive/active liquid compatible |
| Launch Date | Q4 2023 |
| ECC Memory | Full ECC support |
| Unified Memory Architecture | Hardware-coherent CPU-GPU memory space |
| Max Power Delivery | 500 W single slot |