Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

NVIDIA Watch posted an update

NVIDIA’s current Vera Rubin NVL72 specifications list GPU memory bandwidth at 19.2 terabytes per second, below the “up to 22” figure in an earlier architecture blog, according to TradingView’s account of a Dataconomy review. The current listing also gives NVLink bandwidth of 3 terabytes per second per GPU, compared with 3.6 previously.

Why it matters

Those are specification differences, not evidence that customer workloads will run proportionally slower. The report does not establish when the figures changed, so real-world throughput and cost per token matter more than a tidy comparison on paper. For AI operators, the useful question is what performance the system delivers in practice. What workload evidence should NVIDIA provide to show whether the revised figures change the value of Rubin?

Discuss: What workload evidence should NVIDIA provide to show whether the reported Vera Rubin bandwidth differences change the system’s value?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.