Architect Labs releases Frostbite, a CPU autonomously designed by AI
Understand this story
Architect Labs has released Frostbite, a general-purpose CPU in which an AI system autonomously completed everything from architecture exploration, instruction set, RTL, and verification through to the final GDS, using a new instruction set based on interaction nets. The company says Frostbite has completed synthesis on a 7 nm process, reaching 800 MHz timing at the synthesis stage and 500 MHz in the final GDS. Architect Labs also says it will publish Frostbite's RTL and ISA and provide a software testbed hosted on an FPGA for anyone to run programs. Frostbite's actual performance has not yet been disclosed. Architect Labs says that in the coming months it will use industry benchmarks such as SPEC and PARSEC to compare Frostbite's mesh configuration against a multi-core "TCPU" design featuring large caches and larger, wider out-of-order cores.
AI overview based on source reports · Updated 1 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIFeaturedFrostbite: A new CPU architecture discovered, designed and verified by AI
The post introduces Frostbite, a general-purpose CPU with a new instruction set built from the ground up around interaction nets, in which every step is local, every dependency explicit, and every value has exactly one owner. AI systems autonomously iterated over compiler, ISA, architecture, microarchitecture, RTL, verification and final GDS, achieving a 4.59x performance-per-watt geometric mean versus a modern silicon-proven superscalar CPU across 25 workloads (15 held out until final evaluation), with 8.77x performance-per-area. Synthesized on a 7 nm process, each Frostbite quad plus four L3 nodes is 0.792 mm², and the authors describe emergent work-stealing behavior and cross-stack optimizations discovered by agent swarms.
Activity for this story
- Current comparable activity
- 5
- Peak comparable activity
- 10Nov 19 16:00
- Change in the past 24 hours
- -51%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.