Huawei Atlas · Supernode / Cluster
Atlas 800I A3 Supernode Server
The Atlas 800I A3 inference supernode server targets LLM inference in internet and finance, with high performance, reliability and easy deployment in standard air-cooled data centers.
Key Highlights
- Up to 384-NPU interconnect with 48TB unified on-chip memory and 784GB/s non-blocking fabric
- 10U chassis up to 4.48 PFLOPS FP16 with 8 Ascend 910 processors per node.
Technical Specifications
| Form Factor | 10U rack server (supernode node) |
|---|---|
| NPU | 8x Ascend 910 per node, multi-node up to 384 cards |
| CPU | 4x Kunpeng 920 |
| System Memory | 32x DDR5 slots up to 5200MT/s |
| On-chip Memory | 8x128GB, 3.2TB/s |
| AI Compute | up to 4.48 PFLOPS FP16 (官网原文另列8.96 POPS INT8，照录原文用字) |
| D2D Bandwidth | 784GB/s bidirectional |
| Network | 8x 400GE QSFP-DD RoCE + 56x 400GE QSFP-DD bus |
| PCIe | up to 5x PCIe 5.0 |
| Cooling | air-cooled |
Typical Use Cases
Large-scale LLM inference for internet and financemulti-node inference supernode clusters.
翻译稿；归入supernode_cluster（超节点服务器节点）；官网原文AI算力写作“4.48 PFLOFPS FP16、8.96 POPS INT8”，按原文照录未改写；img1: main（背面接口图）; 资料链接为官网《Atlas 800I A3超节点技术白皮书》落地页。
Related Products
Huawei Atlas
Atlas 200I DK A2 Developer Kit
Easy-to-use AI development kit for fast innovation View details →
Huawei Atlas
Atlas 200I A2 Acceleration Module
Edge AI accelerator module embedded in smart devices, robots and drones View details →
Huawei Atlas
Atlas 300I A2 Inference Card
High-performance AI inference card integrated into servers View details →
Huawei Atlas