upvote
You'd probably end up spending more on pcie h200s. They're not that much cheaper than 8x sxm nodes and they only have 4 way nvlink. At the scale these guys were training at you'd probably want at a minimum an 8x nvlinked node for the trainer, even though you could technically fit the full training and optimizer states on a 4x h200 node.
reply
Savings would likely be larger with alternative accelerators such as trainium, no?
reply
Savings would likely be larger with magic beans such as floobarbaz, no?

But seriously. If there were a cheaper and reliable way, I'm sure the authors would have mentioned it instead of H200 SXM modules. They're trying to brag about cheap training.

Maybe you wanted to use an opportunity to shill tranium.

reply
> Maybe you wanted to use an opportunity to shill tranium.

This was unnecessary and against HN guidelines to engage in good faith.

reply
No, and we get way too many low-effort "how does this compare to $FOO" comments that serve only to improve brand recognition of $FOO.

Once you realize people are doing it you see it everywhere.

reply