inference/research
verifiedhigh · 0.95updated 2026-10-03

Atlas 650E AI Server product page and llms-content spec dump

spec-sheet

Fields

id
sources/cn-huawei-atlas-650e-spec
Publisher
Huawei HiSilicon / Ascend Community
Kind
spec-sheet
Accessed
2026-10-03
notes
PRIMARY VENDOR SOURCE, unusually specific by Chinese-vendor standards. 14U air-cooled, 8x 950DT NPU, 8x 96 GB on-chip memory = 768 GB, 4.0 TB/s peak memory bandwidth, 12.48 PFLOPS mxFP4 / 6.43 PFLOPS mxFP8-FP8-HiF8 / 3.40 PFLOPS FP16-BF16 for the server, UB Link 8 x 784 GB/s bidirectional intra-server and 2 x 8 x 1.68 TB/s for 16-NPU full-mesh, UBoE and RoCE 400 Gbps per NPU unidirectional, PCIe 5.0 x5, ~14.5 kW. Machine-readable copy also served at https://www.hiascend.com/llms-content/hardware/en_hardware_ai-server.md

Notes

PRIMARY VENDOR SOURCE, unusually specific by Chinese-vendor standards. 14U air-cooled, 8x 950DT NPU, 8x 96 GB on-chip memory = 768 GB, 4.0 TB/s peak memory bandwidth, 12.48 PFLOPS mxFP4 / 6.43 PFLOPS mxFP8-FP8-HiF8 / 3.40 PFLOPS FP16-BF16 for the server, UB Link 8 x 784 GB/s bidirectional intra-server and 2 x 8 x 1.68 TB/s for 16-NPU full-mesh, UBoE and RoCE 400 Gbps per NPU unidirectional, PCIe 5.0 x5, ~14.5 kW. Machine-readable copy also served at https://www.hiascend.com/llms-content/hardware/en_hardware_ai-server.md

Cross-references

This record references no other records.