To the Editor:

I know that self-timed circuits are the wave of the future, because I have been reading articles that say so for the last 25 years ("Kicking Out the Clock," by Steve Furber. ISD Magazine, May 2000 p. 30). However, it seems to me that if they were the future to anyone but university professors, CAE vendors would be facing enormous pressure from their customer base to produce tools to support such designs. Unless it's being kept a big secret, there's no pressure to speak of. Read Cooley's ESNUG stuff for a few months and count the number of references to self-timed designs - it'll be somewhere around zero. The fact is, most designers and tool vendors are having a tough enough time dealing with the synchronous design paradigm, which is a lot simpler.

Is the slow (practically nonexistent) rise of self-timed logic the classic chicken-and-egg problem? Are tool vendors waiting for customers, and customers for tool vendors? If so, isn't 25 years long enough for something to hatch?

There are people who do self-timed design, and successfully, but they're highly experienced and very detail-oriented, more so than most of the folks who do synchronous design, myself included. The next article on self-timed design should be from one of those people - someone in the industry who's done a large, high-speed, commercially-viable design. That I'd pay attention to.

Bob Perlman
Digital design consultant
Cambrian Design Works
San Jose, CA

Steve Furber responds: Bob Perlman's skepticism of self-timed technology is understandable and only time will tell which of us is right. It's true that self-timed technology isn't new and has had its proponents for (more than) 25 years - yet, it has little presence in current commercial electronic design. However, it is the role of "university professors" to look outside and beyond current mainstream practice, to anticipate problems and to seek better ways. There is no doubt that major problems are looming for globally synchronous design, and the more aggressive companies such as Intel, Sun, and Philips have active in-house asynchronous design teams looking at possible solutions in this space. Keeping a whole chip in-sync is getting harder and the costs in wasted power and excessive radio interference are growing out of control. Ignore these problems at your peril! Self-timed technology has a lot to offer here, but maybe it isn't the only solution. I also remember, clearly, the skepticism RISC processors attracted from mainstream industry while I was helping develop the first ARM chip at Acorn Computers in 1984. I'm not clairvoyant and I am not always right, but nor is mainstream industry always right in its judgement of alternative technologies. The ARM has taken 15 years to develop into the world's highest volume 32-bit embedded processor; so some ideas do take time to yield their full potential. I can't say what market share self-timed technology will hold in ten years' time. But I have spent the last ten years working in the area and trying to help hatch the asynchronous chicken (or is it to lay the golden self-timed egg?). There is a strong sense in the community that the time for this technology is upon us. Industry is showing more interest. Start-up companies like Theseus Logic, Inc., are emerging and securing mainstream investment; some products are making it to commercial production, and so on. Things are moving under the water even though there's not much of a wave on the surface - yet.

ISD Magazine had a chance to chat briefly with David Ditzel, CEO of Transmeta Corp., of Crusoe-chip fame, about self-timed circuit strategies in current chip design. Ditzel acknowledged that self-timed circuit concepts are not bad in themselves. However, they are a bit impractical and not a mainstream strategy at this point in time. He said that self-timed circuits were "probably not used" in developing the Crusoe because classical synchronous design techniques were adequate to handle the challenges of the design.

To the Editor:

In response to Tets Maniwa's editorial (May 2000), "EDA versus Dot.com Enterprise," the EDA development cost has another, sinister side-effect: As software costs increase, less start-ups can afford them. Historically, start-ups have lived on and pushed the leading edge of innovation. Will the future belong to giants who, entrenched, will dictate and slow down the development to maximize their return on the investment? Present software is being developed as integrated systems - each EDA company has to redevelop identical modules, resulting in a huge multiplication of labor and cost. Consequently, small companies can't afford it. This reduces the market and slows progress.

Nicholas J. Bucska
Sr. Technologist
PC Peripherals Inc.
Broomfield, CO.

Tets Maniwa replies: Although the dot.coms have a higher level of interest and funding from the venture capital community right now, this will change in the near future. Web companies who don't have a business model that actually makes money will sink without positive income and will go out of favor with the investment community. This change in direction will free up more funds for EDA start-ups.

Send electronic versions of press releases to news@isdmag.com
For more information about isdmag.com e-mail webmaster@isdmag.com
Comments on our editorial are welcome.
Copyright © 2000 Integrated System Design Magazine