Content by dilip patlolla, wenqing lan, mingfeng yang, ali vaziri, aishwarya prasad tonpe, huseyin yildiz, anshul jain and aditya bhamidipati (1)

Dilip Patlolla and coauthors compare Azure ND GB200 v6 and ND GB300 v6 virtual machines with bare-metal systems and InferenceX benchmarks. Their results show closely matched inference throughput and latency across tested models and serving stacks, while highlighting workload-dependent gains, low-concurrency variation and limits on attributing differences to virtualization.
Community

End of content

Rejoining the server...

Rejoin failed... trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please reload the page.