This article is part of our exclusive IEEE Journal Watch series in partnership with IEEE Xplore.
As society's insatiable demand for computing power continues to grow, so too does the need for more efficient processors. A group of researchers in Switzerland has devised a new processor design that may help meet our computational-intensive needs. It is physically small and computationally agile—and aptly named Snitch. (Harry Potter fans will get the reference.)
Florian Zaruba, a postdoc at the Integrated Systems Laboratory at the Swiss Federal Institute of Technology (ETH) in Zürich—and a researcher involved in the creation of Snitch—notes that there's a trend with commercial, general-purpose cores, which are relying on larger and more energy-hungry processors. "Snitch is the opposite," he says.
Typically, processors try to find an efficient instruction order on the fly, which requires additional hardware and thus uses more power. But Snitch is able to execute the majority of its basic instructions instantaneously, bypassing the need for this extra, burdensome hardware.
Because of this efficient computing approach, Snitch—built around the streamlined, RISC-V chip architecture—can perform most basic instructions within a single clock cycle. As well, it was designed to execute longer latency instructions without stalling and waiting for their completion. "This leads to a very compact and high-performance design compared to conventional processors that achieve high performance," explains Zaruba.
Zaruba and his colleagues describe their design in a study published October 7 in IEEE Transactions on Computers, where they compared it to other benchmark designs. They found that a single Snitch processor with its custom extensions was two times more energy efficient than the other processors analyzed in the study. When multiple processors were used in parallel, Snitch proved to be 3.5 times more energy efficient and up to six times faster than the others.
Notably, there are other hardware components, such as GPUs, that also outperform standard processors in computing speed—but such feats tend to be for highly specialized tasks. Snitch, on the other hand, is much more versatile, performing a variety of tasks while still executing calculations quickly and efficiently.
On the other hand, Zaruba notes, Snitch is more complicated to program. Still, he says that he strongly believes energy efficiency will be the number one priority for next generation computers, and that Snitch's energy efficiency will make it appealing despite its more complex programming requirements.
The researchers have made Snitch's hardware design freely available, and note that they have seen growing interest from industry consortia, for example from the Open Hardware Group, in supporting commercialization efforts.
Moving forward, the team plans to build larger systems based on Snitch. "While we could already demonstrate a very energy-efficient and versatile 8-core Snitch cluster configuration in silicon, there are exciting opportunities ahead in building computing platforms scalable to thousands of Snitch cores, even spreading over multiple chiplets," says Zaruba, noting that his team is currently working towards this goal.
This article appears in the January 2022 print issue as "This RISC-V Powerhouse Goes Light on the Power."