‹ BackHN Continuity

Thread

Nvidia’s Vera Whitepaper Has a Thread Loose

193 points · 44 comments · pella

  1. KeplerBoy · · focus · HN ↗
    AMDs marketing is just as sketchy. Their new Zen6 server page claims 3.3x performance per watt over vera on "agentic workloads" for a 100 kW rack. Maybe they are comparing a CPU heavy rack to an nvidia vera rack with 50 kW of GPUs sitting idle, who knows?

    <a href="https:&#x2F;&#x2F;www.amd.com&#x2F;en&#x2F;products&#x2F;processors&#x2F;server&#x2F;epyc&#x2F;9006-series.html" rel="nofollow">https:&#x2F;&#x2F;www.amd.com&#x2F;en&#x2F;products&#x2F;processors&#x2F;server&#x2F;epyc&#x2F;9006-...

    1. wtallis · · focus · HN ↗
      &gt; who knows?

      You could try reading the footnotes, which include a link to <a href="https:&#x2F;&#x2F;www.amd.com&#x2F;content&#x2F;dam&#x2F;amd&#x2F;en&#x2F;documents&#x2F;solutions&#x2F;ai&#x2F;methodology-description.pdf" rel="nofollow">https:&#x2F;&#x2F;www.amd.com&#x2F;content&#x2F;dam&#x2F;amd&#x2F;en&#x2F;documents&#x2F;solutions&#x2F;a...

      1. KeplerBoy · · focus · HN ↗
        That doesn&#x27;t clarify it and AMD says as much &quot;Because these estimates rely on published results, internal measurements and projection-based scaling factors, they are intended to provide directional comparison rather than direct measured rack benchmarks.&quot;

        It&#x27;s marketing after all and nobody should make buying decisions based on that.

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.