Peter Krupa on Nostr: if i were investing in the GenAI craze, i would skip Nvidia and invest in Groq. ...
if i were investing in the GenAI craze, i would skip Nvidia and invest in Groq. Nvidia's GPUs are great for ML training, but Groq's LPUs are optimized for **running** the models ("inference"). their hardware is fucking **wild** it blows everything else away in terms of how fast it can respond to a query. like, using open source LLMs, it can carry on a phone conversation with a human with no pauses. it's nuts.
https://www.youtube.com/watch?v=WCYf2Agml-sPublished at
2024-04-22 17:55:43Event JSON
{
"id": "a6773df43b902dac907b9710ddee8897c35c88c466752c67f6026581e9d8fa83",
"pubkey": "358b313c445eccc8485ec27702158b45fa6b768ba7a1bcb99c829ebdc979f27f",
"created_at": 1713808543,
"kind": 1,
"tags": [
[
"proxy",
"https://thepit.social/users/peter/statuses/112316156679282069",
"activitypub"
]
],
"content": "if i were investing in the GenAI craze, i would skip Nvidia and invest in Groq. Nvidia's GPUs are great for ML training, but Groq's LPUs are optimized for **running** the models (\"inference\"). their hardware is fucking **wild** it blows everything else away in terms of how fast it can respond to a query. like, using open source LLMs, it can carry on a phone conversation with a human with no pauses. it's nuts. https://www.youtube.com/watch?v=WCYf2Agml-s",
"sig": "a8c09ec76eb3911956c223d4e91031f0392621f7f7084ca29777e4c96b8aad565e042f14bf0ba257a59c4dda290503a6948add28df803b326ccec793632723bf"
}
