NVIDIA Adjusts Rubin GPU Specifications, Lowering HBM and NVLink Bandwidth, Boosting Network Bandwidth
Source：
www.ithome.com
According to IT Home, a comparison between NVIDIA's October 2026 documentation and a third-party archived June document reveals that NVIDIA has reduced the HBM memory bandwidth and NVLink bandwidth of the Rubin GPU system, while increasing network bandwidth. Specifically, the HBM4 memory bandwidth of the Vera Rubin superchip was lowered from 44 TB/s to 38.5 TB/s, NVLink bandwidth from 7.2 TB/s to 6 TB/s, and scale-out network bandwidth was increased from 0.8 TB/s to 0.9 TB/s. The HGX NVL8 system saw its HBM capacity drop from 2.3 TB to 2 TB, bandwidth decrease from 176 TB/s to 130 TB/s, and NVLink Switch bandwidth reduced from 28.8 TB/s to 24 TB/s.
