Leading Architecture Design
Based on AMD® Genoa/Turin Processor, with up to 500W TDP
Support DDR5 technology + PCIe Gen5 latest technology
PCIe passthrough configuration minimizes latency for ultimate performance
Front and rear I/O maintenance compatible design, adaptable to legacy and modern server rooms
Flexible Topology
CPU-GPU supports Passthrough and Switch configurations, with up to 10 DW 500W GPUs
CPU-Switch supports X16 or X32 lane interconnection for high-bandwidth communication
Support diverse CPU-GPU topologies (e.g., common and performance modes) for broader application scenarios
Extensive Partner Ecosystem
Compatible with single and double-wide GPUs
Support GPUs from domestic and international brands
Dedicated team for customized development support
| Name | Taihang H8330 AI Server |
| CPU | AMD Genoa/Turin Scalable Processor，TDP 500W |
| Memory | Up to 32x DDR5 Memory，12 channels per CPU |
| Storage | Up to 12x 3.5-inch SATA/SAS/NVMe drives Up to 24x 2.5-inch SATA/SAS/NVMe drives+16x NVMe drives 1x M.2 SSD |
| PCIe Extension | Rear PCIe Slot: 8x FHFL DW GPU + 5x FHHL+1x OCP Front PCIe Slot: 3x FHHL+1x OCP |
| Management | BMC: AST2600, 1x dedicated RJ45 MGMT port |
| IO | 3x USB, Type-A 1x PWR BTN with LED 1x Status LED 1x UART PORT, Type C 1x Front UID BTN with LED + 1x Rear UID LED 1x Front VGA |
| Thermal | 12x high-performance hot-swappable fans, N+1 Redundancy Optional Cold Plate Cooling |
| Temperature | 5ºC - 35ºC |
| PSU | 4x CRPS power supplies, 3+1Redundancy |
| Dimension | 900mm (D) x 447mm (W) x 175mm (H) |
