Updated
Updated · NVIDIA · Jul 21
NVIDIA Unveils Rubin GPU With 10x Agentic AI Efficiency, Expanding Vera Rubin Platform
Updated
Updated · NVIDIA · Jul 21

NVIDIA Unveils Rubin GPU With 10x Agentic AI Efficiency, Expanding Vera Rubin Platform

3 articles · Updated · NVIDIA · Jul 21

Summary

  • Rubin is built for sustained agentic inference rather than one-shot chat, with NVIDIA saying the GPU delivers up to 10x more throughput per unit of energy than Blackwell.
  • 50 petaflops of NVFP4 performance, 288 GB of HBM4 memory and 22 TB/s bandwidth are paired with NVLink 6 and new attention, MoE and kernel-scheduling features to cut latency and keep multi-step workloads fed.
  • At rack scale, Vera Rubin NVL72 adds liquid cooling, power steering and Intelligent Power Smoothing, which NVIDIA says trims average power use by about 10% and 50 ms peak power by about 20%.
  • DSX MaxLPS can let operators provision up to 40% more GPUs within the same power budget, extending NVIDIA's broader Vera Rubin push after it detailed the Vera CPU and said early systems were already shipping to major AI companies.

Insights

After Blackwell's overheating issues, can customers trust the new Vera Rubin system to run cool under pressure?
Will Nvidia's CPU strategy successfully bypass US export controls and capture China's massive AI market?
Is Nvidia's monolithic chip a brilliant move or a costly gamble against the industry's pivot to chiplets?

NVIDIA Vera Rubin vs. AMD Helios: The Race to Lowest Token Cost in Next-Gen Agentic AI Factories

Overview

NVIDIA's Vera Rubin platform marks a major leap in AI infrastructure, specifically designed to enable agentic AI—AI that can autonomously act, plan, and reason in complex environments. As the third generation of NVIDIA MGX rack-scale systems, Vera Rubin offers a mature, scalable, and high-performance architecture. This platform addresses the growing complexity and efficiency needs of advanced AI workloads, providing organizations with the essential technology to build sophisticated agentic AI factories at scale. By shifting AI from simple information processing to autonomous decision-making, Vera Rubin is transforming how AI systems are developed and deployed.

...