Nvidia's Vera Rubin NVL72 trains models with 75% fewer GPUs than Grace Blackwell
This story is only covered by news sources that have yet to be evaluated by the independent media monitoring agencies we use to assess the quality and reliability of news outlets on our platform. Learn more here.6 Articles
6 Articles
The Greens have long been promising increasingly efficient AI accelerators. However, it seems that this time, their ambitious goals had to be scaled back. NVIDIA has quietly revised the official specifications for the Vera Rubin platform, the successor to the Blackwell chips used in AI data centers. Comparing the current information with documentation from the first half of 2026 reveals a reduction in HBM4 memory parameters […] The article "NVID…
NVIDIA Quietly Lowers Vera Rubin HBM4 Bandwidth By Up To 26%, NVLink Speeds Also Reduced
NVIDIA has quietly revised the specifications of its next-generation Vera Rubin AI computing platform, reducing HBM4 memory bandwidth and sixth-generation NVLink bandwidth across multiple configurations while increasing scale-out networking capabilities. Interestingly, the changes appear to have been made over a month ago, rather than being introduced recently. The revisions, highlighted by Guru3D and ITHome, are confirmed by comparing NVIDIA's …
NVIDIA has rolled back the speed of its Rubin accelerators - The company cannot keep the specifications it originally promised. -- server, nvidia, news,
NVIDIA Updated Vera Rubin Specs Show Lower HBM and NVLink Bandwidth
A comparison of NVIDIA documents published in October with archived June material indicates that some listed Vera Rubin system bandwidth figures have changed. The reported revisions lower the listed HBM4 and NVLink bandwidth for several configurations, while increasing some networking figures.
Coverage Details
Bias Distribution
- There is no tracked Bias information for the sources covering this story.
Factuality
To view factuality data please Upgrade to Premium
