ZICQ
中 Log in / Sign up
Newsroom LLMs & Foundation Models #NVIDIA #AI Factory #DSX Platform #Vera Rubin #Energy Efficiency

NVIDIA Unveils AI Factory Production Optimization Platform: Vera Rubin and DSX Lead Energy Efficiency Revolution

Avatar of Mr.Xu

By Mr.Xu Compiled & Reviewed by Editorial

Published: · 7 views

中文阅读 (Chinese) English Version
Co-reading Select any sentence in the article to highlight, annotate, or copy a quote card. Join discussion ↓

Summary:At the AI Infra Summit, NVIDIA Vice President Ian Buck unveiled groundbreaking advancements in AI factory efficiency, highlighting the Vera Rubin systems and DSX platform's transformative impact on optimizing AI infrastructure energy efficiency. The Vera Rubin systems, coupled with DSX MaxLPS software, can achieve up to a 40% increase in GPU capacity and a 35% boost in token throughput while reducing energy consumption. NVIDIA also showcased collaborations with partners like Emerald AI and Lambd


Revolutionizing AI Factory Efficiency: Vera Rubin and DSX Platform Lead the Way

At the AI Infra Summit, NVIDIA Vice President Ian Buck shared the latest advancements in AI factory efficiency, showcasing NVIDIA's significant breakthroughs in AI infrastructure. Here are the key highlights of the announcement:

Vera Rubin Systems and DSX MaxLPS

  • GPU Capacity Increase: With DSX MaxLPS's dynamic power allocation, AI factories can run up to 40% more GPUs within the same power envelope.
  • Token Throughput Boost: DSX MaxLPS optimizes power allocation across mixed-workload AI factories, increasing token throughput by 35% without requiring new power lines.
  • Intelligent Power Management: Intelligent Power Smoothing software and expanded energy buffering absorb short spikes, allowing systems to run closer to sustained demand, thus improving computational efficiency.

AI Factories and Grid Flexibility

  • Emerald AI Collaboration: Emerald AI partnered with NVIDIA to develop a flexible-load program, using NVIDIA DSX Flex to enable dynamic load balancing in AI factories, thereby supporting grid flexibility.
  • Real-time Grid Signal Response: DSX Flex can receive grid signals (such as load-shedding requests, demand-response events, and pricing signals) and automatically act within a predefined workload hierarchy, ensuring critical jobs continue while others pause and then resume.

Performance and Economics

  • AgentX Benchmarking: In the SemiAnalysis AgentX benchmark, NVIDIA Vera Rubin NVL72 delivered up to 30x higher throughput per megawatt on the DeepSeek V4 Pro model, with a cost per million tokens reduced by 45x.
  • Real-world Applications: Lambda, using DSX MaxLPS on NVIDIA Blackwell servers, saw a 24% increase in cluster-wide token throughput and a 23% improvement in performance per watt.

Technical Highlights

  1. Full-stack AI Factory Platform: NVIDIA has built a complete AI factory platform from silicon to grid, including Vera Rubin systems, Dynamo inference software, NeMo libraries, and NVIDIA networking technologies.
  2. NVLink and Groq 3 LPX: NVLink 6 ensures AI factory reliability through a multilayer resiliency architecture, while Groq 3 LPX provides deterministic ultralow-latency inference, complementing DSX MaxLPS and enhancing overall AI factory performance.
  3. AI Agent Support: The combination of Vera Rubin NVL72 and Groq 3 LPX enables AI agents to achieve higher inference efficiency when handling long-context tasks.

Industry Impact and Developer Recommendations

  • AI Infrastructure Transformation: The evaluation criteria for AI infrastructure are shifting from peak performance to tokens per megawatt efficiency, requiring AI factories to be codesigned from chip to grid.
  • Developer Tool Optimization: NVIDIA DSX MaxLPS and Groq 3 LPX provide powerful tools for developers to achieve higher AI performance in resource-constrained environments.
  • Enhanced Grid Interactivity: The real-time interactivity between AI factories and the grid will become an important feature of future AI infrastructure, and developers should focus on the potential of AI systems in grid flexibility applications.

Conclusion

NVIDIA's Vera Rubin systems and DSX platform demonstrate a revolutionary leap in AI infrastructure efficiency and performance. This release not only provides new solutions for AI factory energy optimization but also opens up new possibilities for AI agents in complex task applications.


Source: NVIDIA AI News (2026-09-15)

— END —

Tags: #NVIDIA #AI Factory #DSX Platform #Vera Rubin #Energy Efficiency

Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.

What's your take on this story?
🚀 Explore & Submit Works

Community Comments

Loading live comments and annotations…