09 February 2006
Beat the clock
Why academics and companies alike are pursuing asynchronous circuit designs. By Graham Pitcher.
Time is the fundamental measure of digital electronics. The generation and distribution of clocks is something of an exact science and, as clock rates increase, the science becomes even more precise as designers deal with nanosecond time intervals.
So there is more than a degree of culture shock when the idea of clockless processors – or even clockless systems – is raised. Yet asynchronous design dates back to the early days of digital electronics. Simon Moore, senior lecturer at the University of Cambridge’s Computer Laboratory, includes self timed circuits amongst his research activities. “The idea was tried initially in the 1950s, when people were dabbling with digital circuit design. There was a revival of interest in the 1990s, when a happy band of researchers started looking at the fundamentals.”
But why should there be any interest in asynchronous designs? One answer is power consumption – a problem that starts to get worse as process technology continues to shrink. Moore noted: “High end processors burn around a third of their power consumption distributing clocks and around a third is accounted for by leakage current. That means only a third of the power consumed is available for computation and data movement. So the logical extreme is to remove the clock.”
However, this logical extreme is not easy to achieve and doesn’t always bring the benefits that may be imagined. Moore noted: “It’s not clear what the advantages of pure asynchronous design are, but there are two immediate benefits. One is lower emi, because computing activity is not correlated to the clock. The other is composition. People talk about the pain of getting timing closure. If you don’t meet timing requirements, then your design won’t be correct. In the asynchronous world, the design will always be correct, but will you meet your performance goals?”
This apparent contradiction has resulted in people looking at how to mix synchronous and asynchronous circuits to the best advantage. And Moore is interested in this area, developing ideas that subscribe to the GALS – Globally Asynchronous, but Locally Synchronous – concept. “Clocked systems continue to work, so why not keep doing it?” he asked. The GALS concept creates systems with ‘synchronous islands’ joined by asynchronous interconnects. “This allows you to put asynchronous technology where it’s needed. There’s quite a lot going on here,” he added.
Author
Graham Pitcher
Supporting Information
Downloads
5679\beat-the-clock.pdf
Websites
http://www.arm.com
http://www.cl.cam.ac.uk/
Companies
ARM Ltd
Cambridge University
This material is protected by Findlay Media copyright
See Terms
and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the
sales team.