NVIDIA Updates Vera Rubin System Specifications

NVIDIA revises Vera Rubin Superchip specs, adjusting HBM and NVLink bandwidth while increasing scale-out networking throughput.

updated its system specifications in October 2026, revealing significant changes from June 2026 archives. The Vera Rubin Superchip now lists 38.5 TB/s of HBM bandwidth and 6 TB/s of bandwidth, down from 44 TB/s and 7.2 TB/s respectively.

Additional adjustments include HGX NVL8 HBM capacity dropping from 2.3 TB to 2 TB, with bandwidth falling to 130 TB/s. However, scale-out networking rose to 0.9 TB/s, while NVL4 bandwidth increased to 1 TB. Vera CPU memory is now described as up to 1.5 TB.

These revisions clarify the performance profile of NVIDIA's next-generation architecture. While core HBM and NVLink speeds saw downward adjustments, networking throughput for scale-out configurations improved. The now targets 1,400 TB/s aggregate bandwidth.


Related news

SK Hynix Hits Record 80% Margin Amid Memory Crisis

RTX 5080 Prices Surge Past $2,300 Amid Shortage

NVIDIA Releases Hotfix Driver for RTX 50 Series

NVIDIA and Cisco Launch USFP Networking Module

NVIDIA Contributes Vera Rubin Platform to OCP