Re:$2 H100s: How the GPU Rental Bubble Burst
by swyx on Friday October 11, 2024 @02:01AM
(#41806640)
so why not buy up all the little h100s and enough together for a cluster? seems like a decent rollup strategy?
ofcourse it woudl still cost a lot to do... but if the difference is $2/hr vs $4.49/hr then there's some size where it makes sense
Re:$2 H100s: How the GPU Rental Bubble Burst
by ipsum2 on Friday October 11, 2024 @03:05AM
(#41807024)
Only if they're networked with Infiniband.
Re:$2 H100s: How the GPU Rental Bubble Burst
by pico_creator on Friday October 11, 2024 @02:17AM
(#41806736)
Makes sense, though only folks like runpod / sfcompute / etc, have enough visibility to maybe pull this off?
Its a risker move - then just taxing the excess compute now, and print money on the margins from bag holders
Re:$2 H100s: How the GPU Rental Bubble Burst
by latchkey on Friday October 11, 2024 @03:08AM
(#41807045)
Correct me if I'm wrong, but if I recall, neither of those two companies own their own compute. They are marketplaces.
Re:$2 H100s: How the GPU Rental Bubble Burst
by qeternity on Friday October 11, 2024 @04:11AM
(#41807365)
I think sfcompute does own a lot or most of the current compute on their platform? Not entirely sure though.
Re:$2 H100s: How the GPU Rental Bubble Burst
by pico_creator on Friday October 11, 2024 @03:16AM
(#41807087)
Yup, but they at-least know where all these "small unused clusters" are.
Bag holders, do not want to be shouting to the world they are bag holders.
Is it big enough for foundation model training from scratch = ~$3+ Otherwise it drops hard
Problem is "big enough" is a moving goal post now, what was big, becomes small