By Hassan Mujtaba
Publication Date: 2026-10-29 07:10:00
NVIDIA has clarified why it updated specifications for Vera Rubin, and says that they better reflect what the systems sustain at volume production.
NVIDIA Vera Rubin Is Now In Volume Production, Delivering “Outstanding Performance”, Says That Individual Component Specs “Do Not Determine Application Performance”
A few days ago, reports emerged that NVIDIA had quietly updated the specifications of its Vera Rubin platform. The spec change mainly impacted the HBM4 memory bandwidth and the sixth-generation NVLink bandwidth. We reached out to NVIDIA, and a company representative told us that the spec change is a better reflection of the systems that are in volume production now and doesn’t affect performance.
The specifications published during Vera Rubin’s development were preliminary architectural targets. Final specifications on nvidia.com and the Vera Rubin datasheet reflect what the system sustains at qualified, volume production.
NVIDIA Rep to Wccftech
When NVIDIA initially unveiled its Vera Rubin platform during GTC 2026 earlier this year, each Rubin GPU was advertised with 288 GB of HBM4 memory and 22 TB/s of bandwidth, alongside 3.6 TB/s of NVLink 6 bandwidth.
In the latest spec sheet, NVIDIA now shows the HBM4 bandwidth reduced to 19.2 TB/s while the NVLink bandwidth is reduced to 3 TB/s. This is a 12.7% lower HBM4 and 16.7% lower NVLINK-6 bandwidth figure. The spec change applies to every single Vera Rubin configuration,…
