Cerebras Systems that makes the world’s largest computer chip, the CS-1 in collaboration with the federal Department of Energy’s National Energy Technology Laboratory (NETL) announced record-breaking performance of its CS-1 chip. According to the announcement, the CS-1 was 200 times faster than the Joule Supercomputer on the key workload of Computational Fluid Dynamics (CFD). The Cerebras CS-1 is the ‘world’s fastest deep learning computer system built with the wafer size engine architecture, which has 1.2 trillion transistors. On the other hand, the Joule Supercomputer is No. 82 on a list of the top 500 supercomputers in the world. In practical terms, this test means that certain AI neural networks that previously took months to train can now train in minutes on the Cerebras system.
“This work at the intersection of supercomputing and AI will extend our understanding of fundamental scientific phenomena, including combustion and how to make combustion more efficient. Together, NETL and Cerebras are using innovative new computer architectures to further the NETL Mission and benefit the United States,” said Dr. Brian J. Anderson, Lab Director at NETL.
The workload which was under test was to solve structured linear equations. Simply put, computational fluid dynamics, which simulates the movement of fluids in places such as a carburettor. In other words, certain specific combustion reactions such as coal burning and how they can be achieved with the best results can be considered an area of application of these chips.
In these tests, Cerebras CS-1 outshone its Joule competition by orders of magnitude. The CS-1 also outperformed traditional GPUs in these calculations as it was 10,000 times faster. The Joule Supercomputer costs tens of millions of dollars to build, with 84,000 CPU cores spread over dozens of racks, and it consumes 450 kilowatts of power. In contrast, a single Cerebras CS-1 is 26 inches tall, fits in one-third of a rack and consumed just 20 kilowatts of power for this specific workload. Not only does the CS-1 draw way less power, it requires a lot less space as well. Moreover, it is easier to deploy.
“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,” said Andrew Feldman, Founder and Chief Executive Officer, Cerebras Systems.
The research was led by Dirk Van Essendelft, machine learning and data science engineer at NETL, and Michael James, Cerebras cofounder and chief architect of advanced technologies. The results came after months of work.