The computing devices that pervade our lives and generally keep the world running can be taught new tricks with software upgrades. But those can only change so much about a device, because the designs of the chips inside are locked in silicon from the moment they’re made.
Two startups are now trying to give chips some flexibility, too. And they say that chip and device manufacturers are interested in adding reprogrammable elements to their chip designs. That could make devices such as smartphones, networking gear, and the computers that power the Internet more powerful and efficient. Some devices might even get upgrades to their chips after they are out in the world.
One of those companies is Flex Logix Technologies, which got its first test chip back from the factory last week. “You can implement a small processor one day or an encryption algorithm another,” says Geoff Tate, CEO and cofounder of Flex Logix. Flex Logix won’t make chips itself, but instead will license the designs needed to add reprogrammable sections to chips. Tate says adding a reprogrammable area to a chip adds as little as 15 cents to the cost.
Another startup, Efinix, has its own technology aimed at making reprogrammable chips widely used. One possible use case is allowing mobile devices to constantly analyze sensor data without quickly draining the battery.
A small chip with Efinix’s technology can be programmed with very specific algorithms for a particular stream of sensor data, allowing it to be around 20 times more efficient than an off-the-shelf chip doing the same job, says Tony Ngai, Efinix’s chief technology officer and cofounder. “This could be useful for new health applications,” says Ngai.
Chips made up entirely of programmable circuits, called FPGAs, have existed for decades (see “Programmable Chips”). They are carpeted with individual elements with switchable connections that can be used to create the different fundamental components of digital circuits and implement just about any possible chip design. But they remain mostly limited to a handful of specialized applications, such as prototyping new chip designs and processing signals inside cellular base stations.
One reason is the bulk and power consumption of those FPGAs. They’re bulky because of the space taken up by the grid of switchable connections that link up blocks of reprogrammable circuit elements. The new startups say they have invented ways to make reprogrammable circuitry more compact.
Flex Logix designs use a trick developed at University of California, Los Angeles, to streamlines those switchable connections, says Tate. A paper on the technique won the Lewis Award, a major honor in electrical engineering, at the International Solid State Circuits Conference this week in San Francisco.
Tate says chips like that have multiple benefits. Some companies like the idea of being able to produce a single chip that can be tuned to fit multiple product lines with different use cases, he says. Others are attracted to being able to update chips that handle communications or encryption quickly – even after a product is in use – to quickly adopt new industry standards or useful algorithms.
Both Flex Logix and Efinix say they are talking with potential customers that want to use chips with reprogrammable parts to squeeze more performance out of the servers that power cloud services. Microsoft last year began using conventional FPGA chips to run the algorithms behind its Bing search engine more efficiently.
Jonathan Rose, a professor of electrical and computer engineering at the University of Toronto, says that the idea of adding reprogrammable patches to chips is a good one. But it is far from new. “The motivation has been around for a long time and has yet to succeed,” he says.
One reason for that is that making space for a reprogrammable section on a chip is not a decision to be made lightly, says Rose. Chips are expensive to make so every bit of real estate is precious, and programming and reprogramming reconfigurable circuits is a complex design process. If reprogrammable chips are to become the norm, Efinix and Flex Logix will need to produce strong evidence that those penalties will be more than compensated for with improved performance or flexibility, says Rose.
DeepMind’s cofounder: Generative AI is just a phase. What’s next is interactive AI.
“This is a profound moment in the history of technology,” says Mustafa Suleyman.
What to know about this autumn’s covid vaccines
New variants will pose a challenge, but early signs suggest the shots will still boost antibody responses.
Human-plus-AI solutions mitigate security threats
With the right human oversight, emerging technologies like artificial intelligence can help keep business and customer data secure
Next slide, please: A brief history of the corporate presentation
From million-dollar slide shows to Steve Jobs’s introduction of the iPhone, a bit of show business never hurt plain old business.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.