Extropic's Z1T models claim up to 140x energy efficiency over GPUs on probabilistic chips
beffjezos · x · 2026-09-05
Extropic founder beffjezos introduced Z1T, the company's first family of transformer-like models built for its sparse probabilistic hardware, claiming up to 140x energy efficiency gains over GPUs and revealing a new scaling law for sparse transformers.
- Z1 is not a digital computer: it exploits the natural jitter of electrons in transistors to perform computation directly, rather than running transistors in deterministic digital mode.
- A single pbit encodes continuous numbers via the average of samples — more samples yield higher effective precision.
- pbits use roughly 10,000x fewer transistors than digitally emulating them.
- Ported layers on Z1 see 1,000x efficiency gains; layers not yet ported still run on FPGA/XPU, bringing the blended figure down to 140x.
Related event: Extropic Unveils Z1T Models, Claiming 140x GPU Energy Efficiency(3 posts)→
More from Infra
- Podcast Jokes About Scouting a 20GW Datacenter Site in Tahoe — chrisalbon · 2026-09-05
- Datacenters vs Shale Oil: Investors Wrong to Dismiss Selling Compute as a Commodity — abhiadesai · 2026-09-05
- Omnigent's Four-Line Defense Against Surprise LLM Bills: Routing, Budgets, Visibility — matei_zaharia · 2026-09-05
- GPT-6 Astra reportedly pretrained on 100,000+ GPUs at Stargate in Abilene, Texas — NinaDSchick · 2026-09-05
- Neural network runs on FPGA with no CPU, OS, or software — pure Verilog logic — blaizedsouza · 2026-09-05
- Mighty Heaton takes on the reasons you hate data centers — csuwildcat · 2026-09-05