跳到主要内容Skip to main content

← 返回硬件算力← Back to Compute

云算力Cloud compute

AWS EC2 P6-B200AWS EC2 P6-B200

单机 8 卡 Blackwell 的云实例,比 P6e 小一档;1,432 GB 是实例内 8 卡 HBM 合计。3.2 Tb/s 是节点网络上限,跨节点训练仍要看集群网络。页面同时给出 1,440 GB 与表格里的 1,432 GB 两种口径,本站按表格逐实例数字记录并注明差异。An eight-GPU Blackwell cloud instance, one tier below P6e; 1,432 GB is aggregate HBM across the eight GPUs. 3.2 Tb/s is the node network ceiling and cross-node training still depends on cluster fabric. The page gives both 1,440 GB and 1,432 GB in its table; this card records the per-instance table figure and notes the discrepancy.

关键参数Key specs

GPU 数 / GPUs
8x NVIDIA Blackwell
显存 / HBM
1,432 GB HBM3e (instance table)
vCPU
192
系统内存 / system memory
2,048 GiB
本地存储 / local NVMe
8 x 3.84 TB
网络 / network
3.2 Tb/s
EBS 带宽 / EBS
100 Gbps

官方原文怎么说What the source says

P6-B200 instances provide 8x NVIDIA Blackwell GPUs with 1440 GB of high-bandwidth GPU memory
1 https://aws.amazon.com/ec2/instance-types/p6/
p6-b200.48xlarge | 8 | 1,432 HBM3e | 192 | 2,048 | 8 x 3.84 | 3.2 | 100
2 https://aws.amazon.com/ec2/instance-types/p6/

资料与许可Source and licence

资料时间Source checked
2026-11-17
许可 / 类型Licence / kind
official-spec-page
许可说明Licence evidence
https://aws.amazon.com/ec2/instance-types/p6/
类目Category
云算力Cloud compute

去官网看原页Open the official page →

← 返回硬件算力← Back to Compute