NVIDIA Reports Performance Gains for Vera Rubin NVL72 Systems

01 What happened

NVIDIA has released performance data for its Vera Rubin NVL72 system, highlighting improvements in throughput and energy efficiency during testing.

02 Key details

• The Vera Rubin NVL72 achieved up to 3.7x higher throughput compared to the GB300 NVL72 in MLPerf Inference v6.1 preview benchmarks.
• Internal testing shows the system delivers up to 30x higher throughput per megawatt on the DeepSeek V4 Pro model using AgentX.
• NVIDIA reports that its DSX MaxLPS technology enables up to 40% more GPU capacity within existing power budgets.
• Company software optimizations resulted in a 1.6x performance increase in MLPerf Inference v6.1 compared to v6.0.

03 Why it matters

The findings address the challenge of scaling inference capabilities within constrained power environments while maintaining efficiency for modern agentic AI models.

04 Who it matters to

Infrastructure engineers, IT architects and high-performance computing specialists.