Cerebras Systems is a company dedicated to improving AI and machine learning around the globe. They have recently unveiled a new computer called the Cerebras CS-1 which is their fastest system ever.
The Cerebras CS-1 may be only 26-inches tall but it houses a whopping 400,000-core processor with 1.2 trillion transistors. It’s not meant to run any games and its sole purpose is to process Artificial Intelligence (AI) workloads faster.
This massive processor is called the Wafer Scale Engine, or WSE (pronounced “wise”) for short. The chip has over a trillion transistors on-board that make it significantly bigger than any AMD or Intel consumer chips found on the market. The whole chip measures 215mm x 215mm with over 400,000 processor cores that have 18GB of on-chip memory and an interconnect speed in Petabits per second.
Despite the chip’s enormous size, the whole machine is actually smaller than a standard tower PC case. It replaces the need for hundreds of thousands of GPUs that not only consume enormous space but also ask for significant amounts of power and heat-dissipating solutions.
Andrew Feldman, Founder, and CEO of Cerebras Systems said:
The CS-1 is the industry’s fastest AI computer, and because it is easy to install, quick to bring up and integrates with existing AI models in TensorFlow and PyTorch, it delivers value the day it is deployed. Depending on workload, the CS-1 delivers hundreds or thousands of times the performance of legacy alternatives at one-tenth the power draw and one-tenth the space per unit compute.
The Cerebras CS-1 has already been deployed at the Argonne National Laboratory where it is being used to help accelerate studies in understanding brain injury and black holes.