535
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 02 Jul 2024
535 points (99.6% liked)
TechTakes
1097 readers
127 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
there’s this type of reply guy on fedi lately who does the “well actually querying LLMs only happens in bursts and training is much more efficient than you’d think and nvidia says their gpus are energy-efficient” thing whenever the topic comes up
and meanwhile a bunch of major companies have violated their climate pledges and say it’s due to AI, they’re planning power plants specifically for data centers expanded for the push into AI, and large GPUs are notoriously the part of a computer that consumes the most power and emits a ton of heat (which notoriously has to be cooled in a way that wastes and pollutes a fuckton of clean water)
but the companies don’t publish smoking gun energy usage statistics on LLMs and generative AI specifically so who can say
I always thought data centers ran clean and dirty loops of cooling (as far as computers are concerned).
The clean loop has all the chemicals and whatnot to keep cooling blocks and tubing "safe". The dirty side is just plain old water. And a big heat exchanger transfers the heat from the clean (hot) loop to the "dirty" (cold) side.
Is there really that much pollution in that? Can't be worse than rain going through storm drains or whatever.
But AI does use a phenomenal amount of power.
And, IMO, it's a problem compared to the lack of value people are getting from AI.
The new Blackwell B200 consumes 1.2kw of power, and will produce 1.2kw of heat.
A cooling system with a COP of 5 needs to consume 240w to dissipate this.
The backplane for the B200 holds 8 of these GPUs in a 10 RU space, and with overheads will peak 14.3kw (cooling would be 3kw consumption).
So, a 42u data center rack with 3 of these, supporting hardware and UPS efficiencies (80%) is going to be 52kw (+10kw cooling). 62kw total, which is like 4 homes drawing their full load all the time.
I hope they finally find an application for AI, instead of just constantly chasing the dragon with more training, more parameters, more performance etc.