Normal Computing Unveils World's First Thermodynamic Computing Chip

Normal Computing has announced the successful tape-out of CN101, the world's first thermodynamic computing chip, as stated in a press release. This milestone marks a significant step in validating Normal's Carnot architecture, which is designed to accelerate computational tasks by leveraging the intrinsic dynamics of physical systems. The CN101 chip aims to achieve up to 1000 times energy efficiency on targeted AI and scientific workloads.

The CN101 chip is a Physics-Based ASIC that utilizes natural dynamics such as fluctuations and stochasticity to compute more efficiently than traditional chips. This approach allows for significant acceleration in tasks like linear algebra and matrix operations, as well as stochastic sampling with Lattice Random Walk (LRW), which are critical for AI and scientific computing.

Normal Computing plans to transition into characterization and benchmarking of the CN101 chip, with findings guiding the development of future chips, CN201 and CN301, aimed at scaling AI workloads. The company envisions scaling diffusion models with their stochastic hardware, starting with key applications on CN101 this year and achieving state-of-the-art performance on medium-scale GenAI tasks next year with CN201.