Welcome to Incremental Social! Learn more about this project here!
Check out lemmyverse to find more communities to join from here!

realitista ,
@realitista@lemmy.world avatar

Not clear from this what the difference is between a TPU and an LPU.

AA5B ,

When Apple shows their iPhone comparison they list number of processor cores of different types. one type of core is “neural engine”. However I’ve never seen specifics on how it works or how well it works. I wonder if it is similar

lowleveldata ,

That's cool. I want cheaper GPUs.

tsonfeir ,
@tsonfeir@lemm.ee avatar

Oh lord, here comes the $4000 Nvidia LPU card grifters

Gork ,

Time for some LTX On / Off memes

tourist , (edited )
@tourist@lemmy.world avatar

LTX off: 5 fingers
LTX on: 7 fingers and a fucked up looking dog

bruhduh ,
@bruhduh@lemmy.world avatar

The more you buy the more you save

scottmeme ,

I'm just trying to get my hands on some faster hardware, https://groq.com has been able to do some crazy shit with their 500/tokens/sec on their LPUs

Lojcs ,

What kind of a website is that? Super slow and doesn't work without web assembly. Do you really need that for a simple interface

scottmeme ,

It's not about their frontend, they are running custom LPUs which can process LLM tokens at 500/sec which is insanely impressive.

For reference with a max size of 2k tokens, my dual xeon silver 4114 procs take 2-3 minutes.

Lojcs ,

No I got what you meant, but that site is weird if it's not doing anything on its own

Finadil ,

That with a fp16 model? Don't be scared to try even a 4 bit quantization, you'd be surprised at how little is lost and how much quicker it is.

Amaltheamannen ,

Isn't it those that cost $2000 per 250mb of memory?? Meaning you'd about 350 to load any half decent model.

scottmeme ,

Not sure how they are doing it, but it was actually $20k not $2k for 250mb of memory on the card. I suspect the models are probably cached in system memory.

Lmaydev ,

That is insanely fast! I figured we'd be getting "AI cards" at some point soon.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • technology@lemmy.world
  • incremental_games
  • meta
  • All magazines
    •