Cerebras’ CS-1 computer system recently demonstrated its ability to perform some simulations 200 times faster than one of the fastest supercomputers in the world.
In order to show the possible uses of CS-1, the Cerebras researchers carried out a combustion simulation in a power plant. The complex equations of fluid dynamics take into account over a million variables, including flow velocity, pressure, and density. The probable performance of the system was 200 times faster than that of the Joule supercomputer, which is one of the 100 fastest supercomputers in the world.
The best modern PC games (summer 2020 update)
“This means that the CS-1 for this job is the fastest machine ever built, and it’s faster than any clustering combination of other processors,” ”CEO Andrew Feldman told ZDNet. If you start the simulation of the CS-1 at the same time as the reaction begins in the plant core, the simulation will end first. The results of the simulation, by Cerebras researcher Kamil Rocki and others at the National Energy Technology Laboratory, were published in a paper, which has not been peer reviewed. Cerebras presented them at the recent SC20 supercomputing conference.
The CS-1 uses the company’s chip that occupies an entire wafer of silicon. Chipmakers like Intel and Nvidia take a block of silicon and slice it into hundreds of chips, each smaller than a postage stamp. The Cerebras Slice Scale Motor (WSE) is closer to the width of a piece of paper in diameter, and it’s a big chip. While a recent Nvidia supercomputer chip has 54 billion transistors, the WSE has 1.2 trillion. Transistors are the electronic switches that control the flow of electrons, much like the tap on a faucet that determines the amount of water flowing. Computers are already in use in artificial intelligence training.
Nvidia GeForce RTX 3070
The size and design of the WSE played a key role in its performance against the Joule supercomputer. In the same test, it was also faster than Intel’s Xeon server chip. Although Xeon cores – individual computing units – have more memory than Cerebras’s, Rocki postulates that Cerebras cores are more efficient and do not compete for shared RAM like Xeons do. The WSE has approximately 380,000 cores. the next version will have 850,000.
While the Cerebras article speculated on some of the fluid dynamics issues the CS-1 could help solve – like landing a helicopter in windy weather or predicting equipment failures – the CEO of the company wants to go further. “We need to apply this to issues beyond computational fluid dynamics,” he told ZDNet.
To learn more about the technology, find out how Nvidia RTX GPUs will remain in short supply until at least April, and check out our recent breakdown of How does the RTX 3060 compare to the PS5 and Xbox Series X.
Jenny McGrath is science writer for IGN. She never tweets, but here she is @JennyMcGeez.