NVIDIA lowers HBM and NVLink bandwidth on Rubin GPU systems, network bandwidth increases
English text is being prepared.
In its October documentation, NVIDIA lowered the HBM and NVLink bandwidth of its Rubin GPU systems, while scale-out network bandwidth increased. The HBM4 memory bandwidth of the Vera Rubin superchip dropped from 44TB/s to 38.5TB/s, NVLink bandwidth fell from 7.2TB/s to 6TB/s, and scale-out network bandwidth rose from 0.8TB/s to 0.9TB/s. For the HGX NVL8 system, HBM memory dropped from 2.3TB at 176TB/s to 2TB at 130TB/s, and NVFP4 and FP8/FP6 training compute was also lowered, while bidirectional network bandwidth increased from 1.6TB/s to 1.8TB/s. Lowering the bandwidth threshold is expected to allow more HBM4 memory to meet Rubin GPU requirements, thereby increasing final yields.
Why read this
AI score 55
Comparing the two official documents shows the bandwidth trade-offs in the Rubin system, allowing readers to understand the magnitude of the spec adjustments and capacity considerations.
Topics
Body
英伟达下调 Rubin GPU 系统 HBM 与 NVLink 带宽，网络带宽提升
2026-10-26 11:26 IT之家 - 溯波（实习）
IT之家 10 月 26 日消息，根据 NVIDIA 第一方今年 10 月文档和第三方保存的今年六月文档间的对比，NVIDIA 下调了 Rubin GPU 系统的 HBM 与 NVLink 带宽，但网络带宽有所提升。
Vera Rubin 超级芯片的 HBM4 GPU 内存带宽此前是 44TB/s，现在是 38.5TB/s；NVLink 带宽此前是 7.2TB/s，现在是 6TB/s；横向扩展网络带宽此前是 0.8TB/s，现在是 0.9TB/s。这些数据基本上等比例地反映在 Vera Rubin NVL4 和 Vera Rubin NVL72 上，不过 Vera Rubin NVL4 的横向扩展网络带宽是从 0.4TB/s 提升至 1TB/s。
而在非超级芯片的 Rubin GPU SMX 模块解决方案方面，两种 HGX NVL8 系统的 HBM 显存从 2.3TB 176TB/s 降至 2TB 130TB/s，NVFP4、FP8/FP6 训练算力均有下调，NVLink Switch 带宽从 28.8TB/s 降至 24TB/s，双向网络带宽从 1.6TB/s 提升至 1.8TB/s。
降低带宽门槛有望让更多的 HBM4 内存符合 Rubin GPU 要求，从而提升最终产量。
此外，所有 Vera CPU 的内存规格均从单颗对应 1.5TB LPDDR5X 调整为“至多”1.5TB LPDDR5X。