Skip to Content

Startup Attempts to Reinvent the CPU to Make Computers Less Power-Hungry

Rex Computing’s 19-year-old founder has a plan to dramatically cut the energy used by powerful computers.
July 21, 2015

Every fairy-tale giant has a weakness, and Thomas Sohmers thinks the same is true of Intel, which ships hundreds of millions of chips every year. To his mind, Intel’s chips guzzle too much electricity. At his startup, Rex Computing, Sohmers is working on an alternative way to architect chips that he says will use a 20th of the power that Intel’s use.

The 19-year-old Sohmers began working on high-performance computer clusters at the Institute for Soldier Nanotechnologies while still in high school. He dropped out and started Rex in 2013 after receiving a $100,000 grant awarded by investor Peter Thiel to encourage people to start companies instead of finishing their education. Sohmers founded Rex with another Thiel Fellow, Paul Sebexen, now the company’s chief technology officer. The company recently received $1.25 million in funding from Founders Fund, a venture capital firm cofounded by Thiel.

Rex is initially aiming its “neo” chips at very high-end data crunching machines like supercomputers, but Sohmers says that once established, his company will move to get its power-sipping chips used more widely. “I wouldn’t state that we’re a high-performance computing company,” he says. “In five-plus years I think we can be something that is practical in Web servers and, past that, more general-purpose systems.”

Server chips that can provide the same computational bang for fewer bucks spent on power could be a boon to cloud companies such as Google and Amazon, whose giant collections of servers run up vast power bills. If they ever make it to consumer devices, Rex’s chips might help extend their battery lives.

Sohmers says he is already talking with companies that use high-powered computers for things like processing images, running machine learning software, or processing signals such as for wireless communications. Rex is aiming to get prototype chips to its partners next year, and to start selling final versions by mid-2017.

Rex’s chips use less power because they don’t have a block of circuitry that’s standard on chips from Intel and other companies. Sohmers says that such circuitry is a wasteful remnant from an earlier age. Those circuits manage the movement of data between memory stores, or caches, built into a chip and the processor core that actually works on data. They were introduced decades ago to make life easier for programmers, but have grown large and wasteful, says Sohmers. A typical high-end Intel chip expends 40 times as much energy to move the result from a single computational operation into memory as it does to perform that operation in the first place, he says.

Rex chips use software to manage their memory instead. That makes it possible to throw out the circuitry that usually does it, and makes chips that have the same computational power but are smaller and need less power, says Sohmers.

Richard Vuduc, an associate professor at Georgia Institute of Technology who works on high-performance computing, says that although Rex’s approach can’t magically slash the power use of every type of computation, the basic idea makes sense. Power consumption is a significant problem for people running supercomputers and data centers, he says. Rex’s solution echoes some earlier ideas tinkered with in academia and industry that didn’t turn into new families of chips, says Vuduc.

However, the energy savings that Rex’s design might offer don’t come for free. Existing software would have to be modified to work on a Rex chip. Rex is working on tools that could make it easy to do that, but Sohmers acknowledges there is still much work to do. “The first customers of this are going to need to have some hand-holding by us and have their own development chops—that’s the trade-off,” he says.

Rex’s plan is to first target companies most constrained by efficiency concerns, buying time to invent ways to make it easier for less urgently motivated companies to switch later. 

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

How scientists traced a mysterious covid case back to six toilets

When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.