Skip to content
Trending storyMonitoring

SemiAnalysis reports NVIDIA AgentX claims on HBM chip and Rubin NVL72 inference speed

1 article1 sourcesince Nov 1Last article

Overview

AISummary of 1 article

SemiAnalysis reports that NVIDIA's AgentX submission claims HBM chips reach over 1,000 tokens per second per user on models above 700 billion parameters, using software written by an agent swarm.

The same submission claims Rubin NVL72 delivers 20.9x better performance per total cost of ownership than GB300 NVL72 at 325 tokens per second per user, and up to 9.11x better at 200 tokens per second per user. These figures come from NVIDIA's submission as relayed by SemiAnalysis, and no independent benchmark of them has been reported.

Written by AI from the articles below · updated Nov 1, 1:00 PM ET

Check the sources:

Latest developmentNov 1, 12:49 PM ET
NVIDIA AgentX submission claims 1000+ tok/s/user on 700B+ models

Article timeline

The articles in this story. Times are ET.

Nov 1
  1. SemiAnalysisX
    NVIDIA AgentX submission claims 1000+ tok/s/user on 700B+ models

    AINVIDIA's AgentX submission, according to SemiAnalysis, reports HBM chips reaching over 1,000 tokens per second per user on 700B+ parameter models using agent-swarm-written software. It also claims Rubin NVL72 delivers 20.9x better performance per TCO than GB300 NVL72 at 325 tokens per second per user, and up to 9.11x at 200 tokens per second per user.

    Image from @SemiAnalysis_'s post

Heat Trend

Current comparable heat
5
Peak comparable heat
10Nov 1, 1:00 PM ET
Change in past 24 hours
-50%

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.