DAC 2008 – Trepidation to Triumph
Q – What’s the difference between MultiCore and ManyCore?
MultiCore is anything from 2 cores up to 8 cores on-chip, while ManyCore ranges from 16 to 64 cores on-chip.
(Not to take sides, but the effort going into refining these definitions seem likes a lot of who-ya over nada.)
At the panel that followed Keutzer’s earlier afternoon panel – an event moderated by Georgia Tech’s Wayne Wolf, that included Renesas’ Toshihiro Hattori, STMicro’s Pierre Paulin, ARM’s Mike Muller, CoWare’s Achim Nohl, and Sonics’ Drew Wingard – Paulin said that beyond 64 cores, you’ve potentially got a CoreEmporium, then a Core-zilla, and finally a Holy Core!
Paulin’s humor struck a cord with a room filled to the overflow with several hundred technology folks, and the place erupted in laughter. Meanwhile, all you Core-zilla fans out there, get on with it! Find a way to take advantage of all of this hardware by introducing better concurrency into your software. And hurry!
Q – So was Management Day strictly business and did it, therefore, belong Downstairs?
Are you kidding? Again, a giant Argggg!
Although I was running around elsewhere in the Anaheim Convention Center for the bulk of the day on Tuesday, Management Day, I eventually found my way to Room 204 at 3 PM in time to catch the last hour of the early afternoon session.
Thanks to Virage Logic’s Yervant Zorian, I was given access to all of the slides from all 7 presenters that had comprised the day’s program. The content was overwhelming!
Each of the speakers over the course of Management Day at DAC had given a highly detailed, 30-minute tutorial on some particular aspect of the design and manufacturing of an advanced-node SoC.
From studying those slides (which will be available online shortly, plus the talks will be featured in an upcoming issue of IEEE’s Design and Test), it was patently clear that if you want to manage the business of making high-volume SoCs at small geometries, you better have your technical skills and your business skills equally up to snuff. This is definitely an Upstairs and a Downstairs kind of a thing.
Qualcomm’s Charles Matar spoke about wireless SoC designs. TI’s Bob Pitts spoke about 45-nanometer power optimized SoCs. Microsoft’s Srinivas Nori addressed issues related to the XBOX 360 processor. SanDisk’s Manuel D’Abreu emphasized the need for complex, accurate specifications. Intel’s Elinora Yoeli talked about the 45-nanometer Atom Processor. MediaTek’s Andrew Chang presented a case study on a 65-nanometer 5-million transistor chip. STMicro’s Philippe Magarshack narrated the development of a 45/40-nanometer low power wireless multi-media SoC.
Obviously, these were dense, highly technical presentations that had consumed the entire day. I only took the podium at the end of the afternoon to try to bring some comic relief to the gravitas of the day’s hard work, and managed to get each presenter to tell me why I should hire them for my new semiconductor start-up.
Most of the job candidates did a good job of articulating their unique skillsets. Magarshack, however, got the nod because he said he not only brought great technical management skills to the table, but also a wide knowledge of French wines, as well. Appropriately, Yervant Zorian waved a wand at that point and the bar in the back opened early.
The crowd relaxed into a lively Q&A with the presenters, now arrayed across the stage at a table. One particular question lobbed at the panelists got pretty much the same answer. With few qualifications, they would each choose cost reduction as a goal for a project over time-to-market. They also all offered that no more than 10 percent of their current toolsets are internally grown – and we’re talking here about players as diverse as STMicro, Intel, Qualcomm, TI, Microsoft, and SanDisk. These companies say upwards of 90 percent of their CAD tools come from 3rd-party tool vendors.
Before you crack open the champagne, however, all of the speakers also said they need “extremely heavy scripting” to knit the flow together and those scripts are definitely their secret sauce.
Hmmm. Not such good news there, in my opinion, for the EDA vendor community.
Drawing some quick conclusions …
Q – What was the main focus of DAC’08?
As a journalist, I’m often asked what I think is the “main focus” of something like DAC. I usually chuckle and ask the same question in response. After 5 days spent in Anaheim, attending a host of different sessions on a plethora of topics, it’s clear that nothing is clear about what the “main focus” is for DAC. Even given that it was a wireless theme this year.
Q – Instead, what were the Best Sessions at DAC?
I thoroughly enjoyed the Wild and Crazy Ideas Session, new to DAC last year and full of equally compelling content this year. A special stand-out presenter at the WACI session was IMEC’s Min Li. It’s still surprising to me when someone can be an accomplished technologist and and an entertaining public speaker, as well. In fact, I heard presentations from EPFL, IBM, and U.C. San Diego at the WACI session, which were all great.
I was also particularly partial to the all-day Biochips Workshop that took place on Sunday, organized by University of Pittsburgh’s Steve Levitan. The morning keynote was an excellent hour-long talk given by U.C. Berkeley’s Jan Rabaey – an overview of the various and diverse research initiatives into invasive and non-invasive devices, gadgets and sensors that are being developed to mitigate physical conditions and improve quality of life. Rabaey ended with a suggestion that our brain waves will soon be accessible to the Internet, and hence will facilitate human-to-human interconnect. A question from the floor got a big laugh. The speaker asked Rabaey why we don’t just interconnect brain-to-brain. Why do it through the Internet?
There were other workshops on Sunday I would like to have attended, as well, but ran out of time. In particular, the 5th Annual UML for SoC Design Workshop and the High-level Synthesis Workshop. I also didn’t get to the IP-XACT User Group meeting on Tuesday, but Peter Flake did and told me it was “dynamic and well attended.” Flake said the fact that the 30+ people in the room argued so vociferously among themselves about the standard means that it’s being used actively, and is well on its way to becoming an established tool for the IP industry.
Although I missed that session on Tuesday, I did not miss the SIGDA PhD Forum on Tuesday evening. Wow. That was very cool. I’m guessing there were well over 200 people there, looking at dozens of posters from promising young grad students, discussing the future of the technology, and networking in a way that bodes great things for the future of the EDA industry. It was without a doubt among the most dynamic of sessions at DAC, to be among all of those bright young researchers and their advisors as they shared details of their research and technical visions for their future careers in the industry. Congratulations to Carnegie Mellon’s Diana Marculescu for her committee’s effort that has brought such distinction to this annual Poster Session.
Best of Show …
What fun is coming to DAC if you don’t look back over the week and single out a couple of absolute favorite events and/or moments?
* Number 1
Of course for me, the first would be the 2008 Workshop for Women in Design Automation. As chair of the event, I would have to feel that way!
Mar Hershenson’s speech was spot-on superb, as was the panel that followed moderated by IBM/AMI veteran CAD tool manager Karla Reynolds . Reynold’s panel included Virage Logic’s Sabina Burns , Carbon Design’s Elizabeth Abraham , Intel’s Gila Kamhi , and On Semiconductor’s Ann Rincon . For a candid, freewheeling discussion of the trepidations and triumphs of a career in technology, you couldn’t have found a better group or a more compelling conversation.