recursive_recursion@piefed.ca to Fuck AI@lemmy.worldEnglish · 6 months agoAI Deregulation & Corruption: Companies Now Have Too Many GPUsi.ytimg.comimagemessage-square43linkfedilinkarrow-up1333arrow-down14file-text
arrow-up1329arrow-down1imageAI Deregulation & Corruption: Companies Now Have Too Many GPUsi.ytimg.comrecursive_recursion@piefed.ca to Fuck AI@lemmy.worldEnglish · 6 months agomessage-square43linkfedilinkfile-text
minus-squarepulsewidth@lemmy.worldlinkfedilinkarrow-up13arrow-down1·6 months agoWhat would we do with them? Ai GPUs are not the same as consumer GPUs - they’re not PCIe x16 cards, they in fact don’t even have a socket format… They’re generally supplied onboard motherboards in a blade-format prefab server build.
minus-squaredamnthefilibuster@lemmy.worldlinkfedilinkEnglisharrow-up10arrow-down3·6 months agoI wouldn’t mind having a local LLM or GenAI
minus-squarepicnic@lemmy.worldlinkfedilinkarrow-up1·6 months agoI have had grid gpus. I’d love to have a few accelerators.
What would we do with them?
Ai GPUs are not the same as consumer GPUs - they’re not PCIe x16 cards, they in fact don’t even have a socket format… They’re generally supplied onboard motherboards in a blade-format prefab server build.
I wouldn’t mind having a local LLM or GenAI
I have had grid gpus. I’d love to have a few accelerators.