Taalas HC1: 17,000 tokens/sec on Llama 3.1 8B vs Nvidia H200’s 233 tokens/sec. 73x faster at one-tenth the power. Each chip runs ONE model, hardwired into the transistors.

  • morto@piefed.social
    link
    fedilink
    English
    arrow-up
    1
    ·
    2 hours ago

    I know very little about fpgas, so I can’t answer your question, but let’s hope someone else can