- Normal IT announces CN101, the first thermodynamic computer chip in the world
- The startup indicates that its approach supports the scale of workloads AI within the power limits of the current data center
- Future conceptions are intended to provide higher performance in the existing infrastructure
Normal IT has announced the success of the CN101 band, which he describes as the “first thermodynamic computer chip in the world”.
The startup considers this development as a natural response to the requests for EIA mounting energy and scientific workloads.
Unlike the CPU and GPUs which are based on deterministic logic, CN101 is designed to exploit natural dynamics such as fluctuations, dissipation and chance.
Normal promising roadmap
The idea is to accelerate certain reasoning tasks while lowering energy consumption by drawing on processes that existing fleas generally remove.
The company affirms that CN101 is targeted in two specific calculation categories. One implies linear algebra on a large scale, which is at the heart of the problems of optimization and scientific modeling.
The other is stochastic sampling, where the random approach to the normal network is intended to accelerate statistical methods, including Bayesian inference.
“In the past few months, we have seen that AI’s capacities are approaching a flattening curve with today’s energy budgets and architecture, even if we plan to evolve the training runs 10,000x others over the next 5 years. Thermodynamic IT has the potential to define the next decades of scaling by exploiting the physical realization of AI algorithms, including that moment-the-way architecture. Paradigme – executed by a radically small engineering team, “said Faris Sbahi, CEO of Normal Computing.
For the future, the company has established a roadmap that begins with CN101 but extends in the next decade.
“Our vision of the scale models with our stochastic equipment begins with the demonstration of key applications on CN101 this year, then to obtain advanced performance on Genai tasks at a medium scale next year with CN201, and finally to obtain several years of magnitude performance for the Genai on a large scale with CN301 from now on.” Patrick Coles, chief scientist of normal IT, explained.
Normal engineers say that the banner also represents the first step towards the characterization of the way these ideas behave in real silicon.
“CN101 represents the first silicon demonstration of our thermodynamic architecture which takes advantage of randomness, metastability and noise to perform sampling tasks. said.
Normal IT was founded in 2022 by engineers from Google Brain, Google X and Palantir.
