There's a problem with that argument: software caused obsolesce. Had a supercomputing data center manager tell me all about it last week. Essentially boils down to NVIDIA cranking out new CUDA versions that other apps depend on and, surprise, these new versions don't recognize older GPUs.
@vince_zinon@BullTheoryio It isn't that they necessarily stop innovating, it's that they get out innovated by new cheaper things that are 'good enough.' There are many, many examples of this throughout the history of tech.
reposting this, bad link!
Just dropped: Real TCO numbers for a production 248-GPU AI cluster. On-prem vs AWS, GCP, and OCI.
The gap is bigger than most people expect.
Full report with real pricing and transparent assumptions.
Read it here: https://t.co/a2ALXJRuJ6
@_mostwanted002_@HotAisle I would have a one word response to this: El Capitan...oh, wait, that's two words. But Frontier is one word. Pretty sure both have decent performance and scale (wish there was a 'heavy sarcasm' font)
@HotAisle Yes, but is it AI enabled? Does it learn from all of your past fan usage and then predict your future usage? Seems like that's the next step, right? 😆
@libbyemmons This isn't real. The picture is too good as is the audio. The cuts and slight angle changes show pro editing chops. Also her makeup is perfect too. As is her story line.
Awful Tech Today: how AI will scr*w me today. Delta, my carrier of choice, is going to use AI for ticket pricing. Wonder if it will be smart enough to figure out that I go to Germany for ISC every year? Or to SC, or other events, and use that to jack up my personal price? Nah, they wouldn't do that, right? https://t.co/yvhIF1rvQu
I just dropped my 3,000 rand ($160 US) into the BackaBuddy fund to get Team Wits from South Africa to China for the ASC25 cluster competition.
Challenging anyone in the HPC community to meet or beat my number.
https://t.co/XT3vQfJkvY