|
My musings from ESC on the irrelevance of silicon (see previous post) sparked quite the flood of emails. Responses ranged from flipping the argument to how software isn't keeping up with silicon and is slow to adapt, to the need for more efficient architectures such as ASIPs and reconfigurable silicon, to the complete failure of the EDA and chip-design communities to innovate.
All were thoughtful and strongly worded, and I've included as many as I am allowed below. Take the time to peruse them, they're worth it. However, it was an email from Tien Pham with Synopsys that set me on the right track and took me out of my 'irrelevant silicon' funk. In it he said: "I think it's the idea embedded within the product that makes the difference. Hardware and software merely enable an idea to be implemented into a product." He went on to say (and here's the kicker): "...I think engineers should be more creative in using the currently available hardware and software to develop winning products. This is where the art comes in and stands side by side with the science. Similar to an artist that uses a few simple brushes and canvas to create a master piece, engineers can and should be doing the same thing." Bravo, Tien! Well said! First and foremost, design is an art. As much as an artist works with the interaction of canvas and paint, designers need to consider the silicon and software. So too for the respective silicon and software developers, each needs to consider the other and work together or they're going to lessen the designers' creation. When all is in harmony, the designer can truly be free to create. So, what are you going to paint today?
PS: I'm off to the United Arab Emirates today to see what's happening in the Dubai Silicon Oasis and catch the 17th. Annual International Electronics Forum. I'll be posting from there as I go along, and probably filing now and again for our sister publication EETimes, but if there's anything you think I should be looking at, or anything you'd like me to find out about, let me know. I'd be happy to oblige.
As always, I can be reached at pmannion@techinsights.com.
Member feedback (see more in "Discuss this article" section at bottom):
I think it's the idea embedded within the product that makes the difference. Hardware and software merely enable an idea to be implemented into a product. Let's be clear that the more powerful the hardware is, the more interesting a product can be built upon it. But it's not the factor that makes the product a winner. It's the idea, the creativity, the vision behind the product. Iphone is a good example.
Neither hardware or software is top-end but the system that embodies the elegant, innovative and user friendly interface is. My belief is that the current state of hardware can enable a much wider range of innovative products than currently available.
Instead of focusing on pushing the megahertz, or the number of cores you can put on a chip, I think engineers should be more creative in using the currently available hardware and software to develop winning products. This is where the art comes in and stands side by side with the science. Similar to an artist that uses a few simple brushes and canvas to create a master piece, engineers can and should be doing the same thing.
Tien Pham
Staff Engineer
Synopsys, Inc.
The question of silicon outpacing the needs of SW could be turned around, of course. Instead, perhaps the framing of the issue should be about SW capability not keeping up with silicon. I think Bill Schweber’s recent commentary (EE Times 4/7/08: Opinion: Time for another AI reality check) was spot on. If we had the equivalent hardware processing power of a human brain in hand, it’s not clear that anyone could program it. Perhaps it’s time to think about a paradigm shift in system architectures.
Ron Cline
Xilinx
Since I was around when that "who needs more than 640k" comment was made, I've seen software trail silicon for decades. Every once in a while someone stitches together an interesting interface based upon some standard parts. I seem to recall that the original Mac wasn't based on anything earth shattering either. Comments like Tredennick's seem to forget that its those little building blocks that allow software designers to do what they do - years after they were leading edge. Guess that just means the software world is slow to adapt. Good article, regards,
Bob Faulkner
Cranbury Capital LLC
I think the main thing now is to reduce power dissipation (energy usage), rather than cranking up the horsepower. Particularly in industrial, automotive and military applications. Not just from a battery life standpoint, but from a heat dissipation perspective. Fans and heatsinks are not an option for everyone.
Robert Stock
Senior Principal Engineer
Austin, TX
Hello Patrick, here’s a quote from your recent EETimes article: “Fast forward to the main ESC keynote, where Nick Tredennick, technology analyst for Gilder Publishing, casually remarked that no additional transistors need be embedded in chips. ICs already have enough horsepower.”
That’s not exactly what I meant to say. In fairness to all, I was given five minutes, with the objective: “…look back twenty years and look forward twenty years…” That’s a tall objective for a five-minute talk. What I intended to say (and thought I did say) was that transistors are good enough, not that applications don’t need more of them (though I might agree with this point too). I’ve attached the notes pages from a couple of pages in a PowerPoint presentation. The slides, I hope, illustrate the concept of the value PC and of the value transistor. (The rest of the presentation surrounding the attached slides can be found here: http://www.tredennick.com/ftp/pub/documents/Presentations/. I have also written a couple of newsletters on the topic of value transistors and the value PC and would be happy to send them to you.)
The introduction of the microprocessor stalled innovation in logic design; the industry came to depend on smaller, faster transistors for its progress. Today’s transistors are small enough and fast enough for most applications, so it’s time to bring back logic innovation (perhaps in the form of reconfigurable systems) as a means to further progress.
Nick Tredennick
Gilder Publishing
Hi Patrick, interesting article and observations about the embedded industry going forward. I think company's investments are becoming higher and higher in software because of the fundamental difference creating at the user interface level as done by iPhone or any Apple device. Company's also wants to use the chips they have on their devices to their fullest potential to reduce the overall cost. If additional chips are ready to provide better usability such as 3D graphics used by iPhone (nobody in mobile phone might have tried previously) companies are ready to use the extra silicon. Everything on the earth can't be performed by under stuffed silicon with great software. A better software can improve things similarly a better silicon can also improve things.
Finally, Silicon and Software needs to go hand in hand to fulfill each others capabilities at fullest potential and provide a critical differentiation to end customers. For example having Pentium IV in might have not been useful if people are still working with MS-DOS operating system. Similarly 80286 might have not sufficient to run WindowsXP. Regards,
Suresh
Director
www.squidsystems.com
Hi Patrick, an interesting and a brave viewpoint. I see a few crucial aspects blocking the growth in value of software on silicon. One of them is the predictability and reliability of complex software, both in innovation and in run-time. Second aspect is that with scale software only gets costlier to create and maintain, unlike silicon. Thirdly, the lack of adequate abstraction levels in, and design automation tools for, embedded software ensures that large and complex software will remain a "necessary evil". All these make it necessary to minimise embedded software. Given these, silicon will grow in the direction of minimising software. I see the trend towards reconfigurable silicon an evidence in this direction. With warm regards,
S Sundararajan
NXP, Bangalore
Patrick: The issue may not be so much the raw horsepower in silicon as what kind of horsepower is needed. In this, I agree with your comment that what we need is a shift from horse power to horse sense. Portable devices in particular need energy-efficient computation and communication, so that every bit of power consumed is going towards the application and very little is wasted. This is exactly where Application-Specific Instruction set Processors (ASIPs) make most sense because of their ability to maximise the intelligent use of the energy resources available in portable and indeed plugged-in devices, while delivering the performance the application requires. As this is the field I am working in, I am of course a little biased, but the importance of minimising energy consumption seems clear no matter where you are coming from. Very interesting opinion piece. Best regards,
Grant Martin
Chief Scientist
Tensilica
Patrick: I found your article on the irrelevance of silicon interesting. I've been saying for a long time now that we spend too much time chasing performance metrics that no longer make sense for what computers actually do. The basic assumptions of both EDA and Computer Architecture are no longer valid. Emerging computers are "scenario oriented" -- no longer general purpose or application specific. Mobile computers that access the internet are perfect examples. Set top boxes and gaming consoles are others. See the article, "A New Era of Performance Evaluation" in the September, 2007 issue of IEEE Computer for more on this.
JMPaul
Professor
Posted by Patrick Mannion on May 6, 2008 03:59 AM in DesignScape
|
