Urgent.News

the world's headlines, one feed

AI

AMD acquires Toronto-based Taalas, which integrates model weights directly into silicon to boost inference performance, for an undisclosed sum (Tobias Mann/The Register)

Early tech demos show model-specific integrated circuits churning out up to 17,000 tokens a second — In AMD's latest bid to upset Nvidia's dominance …

AMD has acquired Toronto-based Taalas, a company that integrates model weights directly into silicon to boost inference performance. The terms of the deal were not disclosed. Taalas' approach is different from conventional GPUs, using model-specific integrated circuits that etch model weights into the silicon rather than relying on HBM.

Taalas' tech has shown promising results, with its first test chip, the HC1, fabricated on TSMC's 6nm process, serving Meta's Llama 3.1 8B at 16,960 tokens a second, according to The Register Science. This approach promises to boost inference performance by an order of magnitude or more.

The acquisition is seen as part of AMD's bid to challenge Nvidia's dominance in AI hardware. Nvidia has also made significant deals in the area, including a $20 billion licensing deal with Groq. AMD's deal with Taalas appears to be an actual acquisition rather than an acquihire.

Brief written by urgent.news from Techmeme, The Register Science — 2 reports on this story. Machine-written — read the original for the full account.

We haven't written up this one. Techmeme has the full story — the link below goes straight to it.

Read the original at theregister.com →

More in AI