DVCon 2006 reason & rancor trump
February 26, 2006 It's true that Joe Sawicki brought a long blonde wig and Brett Cline leaped off the podium to mount a audacious Cooley cartoon on the overhead project but in the main, John Cooley's BigWig Panel circa 2006 at DVCon was a completely different beast than its distant cousin, Cooley's BigWig Panel circa 2005. What was the difference? After all the panelists this year: Gary Smith Senior EDA Analyst at Gartner/Dataquest were pretty much the same as last year: Gary Smith Grand EDA Poobah at Dataquest But despite the hopes of the audience, this year the conversation was markedly (surprisingly) substantive, informative, and rich in content in a contentious, combative, co-opetition-ish (is that a word?), and communal sort of way. So what was said? Well, in a time-honored, 1-year-old tradition, here's the basic script (see below), give or take a word or two. (Read: "This is a paraphrased version and not the verbatim transcript that would hold up to close scrutiny when compared side-by-side with the video.") Will there be a video? You betcha. Will it be worth watching? You betcha, unless you've got people to meet, places to go, and money to earn. Which is the underlying reason, in fact, why Cooley's annual event continues to have market traction. It's all about people you need to meet, places you need to go, and money you're going earn if the gods of automated, optimized algorithms are smiling down on you, that is. By the way, before we launch into this recital and recap it occurred to me recently that John Cooley is long overdue for some sort of recognition for the special gift, and grist mill, he brings to the industry. If John were not out there putting his bravado and bold-faced, in-your-face face out there, how would this industry communicate with itself? You may feel differently, and I'd love to hear about it, but if it were left to the glossy, happy-talk tabloids and you know who you are where first-time success stories are the mainstay (and mythology) of the copy (not to mention the bread and butter of the PR community), you'd all be under the impression that vendors love their customers, those customers love their customers, and a good time is had by all. Worse, if it were left to the EDAC CEO Forecast panel on February 2nd to let us all in on the future of the industry, even less would be known. That exercise in window dressing for Wall Street simply did not compete for a moment short of some commentary from Wally Rhines with the informational content that was in Cooley's panel at DVCon February 23rd. So, go get that coffee and read on and try to remember that half of the audience came to be amused. They were looking for a laugh-a-minute and went away disappointed. The other half came to be annoyed. They wanted prove the industry takes its role in life way too light-heartedly. They also went away disappointed. This 90 minutes was not funny and it wasn't annoying. It was downright informative and in some respects, downright tragic. Like an Italian opera, it was a lot of passion and music til the curtain came crashing down. The libretto? Rome is burning, but Nero continues to play. Last fall, I had a moment of optimism about EDA. After observing the panel, I think that optimism might have been premature. But even if Cooley's panel is the reality, perhaps it's not too late for EDA to put its house in order. They've got to address these many issues, however, if they hope to survive. ************************************* John Kathryn, there's a rumor that Jasper is dying. Kathryn No, it's not dying. We've made more progress in the last 6 months, than in the last 3 years. Personnel changes are instigating rumors you don't join a start-up and get a manual Craig Shirley [VP of Sales] left Jasper and moved on. He's going to a new position in life. Also, we've been doing a lot of sales experimentation people go and do stuff and crack the code but usually when they leave, it's a fit issue. John How many people have left? Is it 10 sales people in the last year? Kathryn It's more than 5, and less than 10. [A howl from the audience] John And in engineering? Harry Foster left. [Harry, former Jasper Chief Methodologist, was sitting up front and highly visible.] Kathryn Harry left to go into a new position at Mentor in their verification unit in Texas. He had worked for us 3 weeks out of the month here [in Silicon Valley], but his family is in Texas. I'm grateful to Harry, but can't fault him for leaving. There's no brain drain at Jasper that's just gorilla marketing from our competition. And in the past 14 months, we've had a big shift in the center of gravity. Now we have teams in Sweden and Brazil. John Sweden and Brazil? [Another howl from the audience.] Kathryn We haven't seen folks coming out of Stanford or Berkeley with the expertise we need, but formal verification is centered in Sweden and Brazil. John Antun? Brazil or Sweden? Antun There are skill sets around the globe. Kathryn It's a growing team. In engineering, ours is the largest group around. John Ted? Why did the Cadence Design Foundry fail? Ted It didn't fail. Now we're extremely profitable. John You had 2000 people, and now there are 200 or less. [Laughter] Ted We sell software and help designers get chips out. I'll be blunt we tried to bring in designers to build on demand, but it didn't work. The industry wasn't ready to accept commodity work, or specialized work. We put a huge investment in a direction that wasn't right. Now, we bring competency to help differentiate designs. You need teams to assemble leading-edge designs. There's been a transformation we're past the point of mechanical integration. John You had to dump people, and so you dumped them and now you have intelligent design? [Loud laughter] Mentor and Synopsys is that a success? Ted Define success! Joe Making money! [Howls of laughter] Ted We are successful today. I'd hold up our design services up to anyone's in the industry. John What did Synopsys and Mentor do differently? Ted They grew incrementally. Antun Your comment on transformation is accurate, Ted. We're trying now to provide complementary services, not replacement services, to our customers. Last week, we announced three 65-nanaometer design tape outs. John Joe? At Mentor? Joe Mentor and Design Services is an oxymoron. Now, our biggest product is Calibre implementation. John Joe, you paid fifty million dollars for 0-In Design. Before the acquisition, my stats say 47 percent of the market was using the tools. After the acquisition, it was 26 percent. Is 0-In being replaced by competing tools like Synopsys' Magellan and Cadence's IFV? Kathryn ... or like Jasper! John Is it competitive pressure, or is Mentor where tools go to die? [Hearty laughter] Joe I brought my wig today because I thought this was the BigWig panel. [Blonde wig and big howls] Kathryn Way to change the subject, Joe! [More howls] Joe Cooley's survey? I love it, but remember those are the impressions of people in an F-16 flying down the street at 100 feet above the ground. We flipped over every 0-In contract in play after the acquisition, plus brought in over 100 new customers. We've completely reinvigorated the technology and couldn't be happier. At least that's what my notes here say. I had this conversation with our guys before I came here, and that's what my notes say. [Big crowd guffaws] John Rajeev [John's interrupted mid-sentence and doesn't finish posing the question.] Rajeev We've exchanged a lot of letters! John But you sent the first letter. Rajeev Fundamentally, it's not about who started it. Magma's spending 25 million dollars a year on this lawsuit, and I think Synopsys must be spending a similar amount. That's 50 million dollars being spent by the industry the equivalent of 400 employees who are not working! 55 lawyers are working instead! So just jumping in and asking about a letter or a lawsuit is not the point. 50 million dollars and 500 jobs in the industry are not producing anything new! Antun We're not spending anything near to that amount. The letter you received implied that if you were taking steps But the story is that I sent you a letter just to say hello. Rajeev This industry is spending a lot of money on lawyers! When you can't do the technology, call the lawyers! [The sound of a catfight offered up from somewhere in the audience.] John [turning to the audience] And you wonder why I seated them apart. [Audience howl] Antun I agree that the legal fabric should be working in a non-exploitive fashion. You should always behave responsibility. But we will take letters seriously and protect our IP. Rajeev If you respond to every letter with litigation instead of actually sorting things out Antun Each of us has a way to sort things out. John Antun, are you actually using the patent, or is it a matter of ego? Antun We are not fighting for ego, but protecting our IP. John Just answer the question. Are you actually using the patents? Antun What we are using is a different story, the claim is. the claim is [Audience rumbles and grumbles.] John Why not just say? Rajeev? Are you using the patents? Rajeev It takes a long time to go through the legal process. We don't use the patents. Certain papers were presented in Europe. One of these guys saw our paper and came back and filed for a patent. Antun We didn't start the fear, dread, and uncertainty. John Antun? Would Nassda say that Synopsys was a litigious company? Antun The case of Nassda was completely different! The court cited Nassda for destroying evidence and source documents. We're not looking to be litigious. Rajeev And Ambit? [Pounding the table] Antun That involved a non-compete agreement! There were no John Gary? You get the final word and they can't rebut your answer. [Huge laughter] Gary Dick Cheney had the right idea! [Shouts of laughter] John Brett? Are you the poster child for SystemC because it's market challenged in the U.S.? Brett I'm the poster boy, not child, but the U.S. market is catching up quickly. Look at the OSCI website downloads and places like Intel and Starkey are using SystemC. The market's alive and well here, but it's a late bloomer. John Oh ugh! Brett 277,000 downloads of the simulator. A growth from 32 percent to 42 percent in ESL. 26 companies on SystemC committees, including some major semi's. Very strong growth. John Ted? Ted It's a regional issue based on the locale of SystemC users. It's not RTL, but the architectural community that started it. As it matures, more companies will be making trade-offs. John Is Cadence backing SystemC because they don't have lead in SystemVerilog? [Laughter] Ted No, the technology's solid, but SystemVerilog has some click factors and advantages for that same old digital guy. John Gary? Gary SystemC is being used as much in the U.S. as in Japan and Europe. There are three different ESL models, and tools at the algorithmic level is only one of them, like Forte's synthesizer. They're selling to toy designers in Japan and Europe, and there aren't a lot of toy designers here in the U.S. Here, it's mostly memory and logic and those guys are using in-house tools. Most ESL tools are in-house, so the EDA industry is not seeing the market. Rajeev The reality is that ESL is happening. There are certainly obvious signs everywhere in SystemC or C, but the tools are lacking. When you moved from schematic to RRTL base, it was about having push-button tools. ESL is very different from RTL to GDSII, but things will happen in ESL too. John Rajeev, I'm flummoxed as to why you'd dump on your own market. [Laughter] Mike? SystemC in DFM? Mike No comment on SystemC. John Ted? Are you selling Get2Chip technology? Ted The Get2Chip compiler was never taken off the market, but the technology was integrated back. It's continued to evolve with those opportunities in the market. John Yes or No. Are you selling it? Ted We are not selling tools in architectural synthesis now. John Will you be soon? Ted I believe there will be a need and we will have the tools. John Antun, is Synopsys pimping SystemVerilog? [Dark howls] Antun Between SystemC and SystemVerilog, the commonality is only the word System in the name. SystemC is in a completely different space. We support SystemC in higher level tools I don't see confusion there. John Jasper? Kathryn We see customers who are using SystemVerilog or PSL assertions, who also use SystemC. There's little overlap. We support Verilog, SystemVerilog, and PSL. John Joe? Are you happy in SystemVerilog? Joe Our simulators support both. We focus on supporting standards. John Is ANSI C an anti-Brett maneuver? Wait are you looking for your notes again? Joe [Looking over at Gary] Gary says we do both. [Huge laughter] Brett The other fact is where people are doing large designs. If you want to put it into the hardware, you have to use SystemC. SystemC is just a class library on top of C++, it's not a language unto itself. It has all of the C functionally, so you always start out in C and build on that. John Ted? Freescale? Ted I'm usually a nice guy, but this is a bad question impugning the integrity of the management at Freescale! Those guys had a challenge as a spin out, and they came here and looked at what we have. They talked about what they were spending a shit load of money on tools and they said, 'We're not getting out of all of this what we really need.' So it's not a question of Cadence or Synopsys, these guys had a problem. John But all you can eat? Ted They were coming out and finding a partner for everything under way! You're damn straight there are gaping holes in the flow, and they understand they have to fill those holes. But I'm passionate in my opinion about this I'm sorry, but the industry has to get over it! John That's a tough position, so what happens next? Joe We're filling those gaping holes! [Huge laughter] Rajeev We've had several big companies try to lock us out of the market, but none of the underlying solutions are coming from that. Sometimes there are financial requirements, but we're not a Costco or a Wal-Mart. Basically the customers have to partner and work with the niche players and big vendors. John Then why clean their clocks in DRF? Rajeev In routing and place and route, you need an OPC-absorbent product. Joe's got big shoes to fill. What size shoes do you wear, Joe? Joe Size 15 Rajeev Like I said, Joe's got big shoes to fill. John Mike, with over 40 different DFM companies how do you compete? Mike This discussion is very encouraging to me. EDA is flat at 4 billion dollars, but there's money to be made if you can help with efficiency. It's stupid for the big EDA vendors to be giving away the tools, but if they want to crater the market, they'll screw it up for everyone ultimately! Rajeev Ultimately, eat for free is nothing more than shooting yourself in the foot. The industry can't afford to do it for very long. John Antun, do you guys do all kinds of deals? Antun The large companies do not give away tools for free. The question here is the role for specialized tools. John Kathryn? Kathryn I know that Synopsys has whole product lines where if one product in the line gets dropped, theyll take an untested tool to fill in that hole. But we have numerous cases where customers say, 'Even though we can get it for free in the mix, we would rather pay for the Jasper tool.' John Antun? Antun Well, there are mixes and there are mixes. [Howls] John That's very Zen, Antun. [Bigger howls] Antun People always optimize the use of their EDA dollars. Rajeev I disagree. Companies do do mixes, even if it doesnt serve the process but some people do it anyway! Kathryn There's lots of pressure to do it. John Gary? Gary There's been a line of back-end tools from Synopsys, Cadence, and Magma and there's been a line of front-end tools form Mentor, Cadence, and Synopsys. At 65 nanometers, all 3 have taped out now. So the buyers are having a ball playing each of the vendors off against each other. The simple fact is that the CAD budgets have not been fully spent over the last several years. They've won in the IC implementation toolset wars. John That's one of those weird Gary Smith category things. [Laughter] Gary The power users have driven that by going back to build-or-buy point tools. What's driving this is OpenAccess, and the market maturing. John And Freecscale? Gary That's a start-up using the OpenAccess flow to decrease integration problems. Anything outside of one set of tools completely changes their tool purchasing. But companies no longer are cutting out start-ups now they're relying on the start-ups. Rajeev Why are there leftover CAD budgets in these companies? We used to be over 2 percent of the semiconductor market now we're only at 1.7 percent. Fundamentally, we haven't automated and offered what our customers need. Mike We've got to stop pricing to budget, and instead price to value! People will pay the price for good tools. When I worked at Cadence, I saw a re-distributor FAM [flexible access model] sell a simulator for $30! John Gary, who gives away the most tools?
Gary Up til DAC last year I've had a Daddy Gap, so I'm only current up through the last DAC Mike [Fister] and Aart [de Geus] said they would both hold prices. Rajeev We're walking away from deals right now, because as long as my mother isn't designing chips we have to put some pressure on the customers. We've walked away from deals! Kathryn It has to do with the cycle. I took a Mommy Gap myself from 1999 when I was last CEO at Verisity to 2003. In 1999, there was a demand for tools from customers who wanted to leapfrog over their competition. By 2003, the customers were saying they didn't want to jeopardize their jobs with new tools. John Are you seeing that pressure, Mike? Mike There's not as much pressure as before, but there's huge pressure upstream that's just miserable. John Mike, there are 40 startups in DFM. Which ones will still be there in 3 years? Mike Gary would sell you a report for $10,000 that would answer that question [Laughter] but of course, Aprio, and Clear Shape, KLA-Tencor, PDF Solutions and Luminescent. They'll be around in 3 years. It's the whole DFM situation the manufactures dont know what they're doing with respect to the design. And the design guys dont know what they're doing with respect to the manufacturing. The manufactures are saying, 'Can you help us?' John Cadence? Mentor? Synopsys? Antun You asked about start-ups, or new players? A DFM view and a shape simulator this is a lithography level issue. John Mike? Mentor? Cadence? Synopsys? Mike It's about a link to manufacturing. Cadence has its new litho-aware approach, which is a great approach, but it only links to the backend. How do you move above that? Mentor has great back-end, but little up stream. Magma has lots more to do, but [so far] Mojave is great. Synopsys has got it all TCAD and mask data. [For them], it's the integration that needs to happen. Rajeev Look, it's the amount of resources it's not just D or M. You have to do a design flow, and you have to have manufacturing we're learned quite a bit. All of us are going to fight it out and time to market will tell the story. Just taking shape into account isn't enough we've got timing, extraction, and placement in place. John Cadence? Synopsys? Ted? Antun? Ted For the knowledge and understanding of what you need to do down to 32 nanometers, 22 nanometers, you have to have strong relationships. You have to have visibility into the foundry, but is the industry going to accept [that design also means] you have to know how to make the thing. We're encouraged by the 40 guys in DFM. They will make a contribution, while Cadence needs to focus on what it all means on the design side. Antun Synopsys has all the pieces, and you'll be hearing much more from us. John Joe? Is Mentor kind of screwed? [Laughter] Joe We're the Switzerland of the back-end. John The Calibre checker tool? Joe We're not just a checker tool. You're a front-end guy, John. You don't know the back-end. Eventually, we're all going to talk about the back-end. Right now, Synopsys is only has player competing with us. You have my word on it the market will decide. John Mike? Mike Upstream and downstream, you need a lot of improvement in the tools. KLA-Tencor, PDF solutions [are offering tools]. Luminescent is such a disruptive technology, if only they can keep their doors open. There's been a lot of tire-kicking there, however. John Mike, what's your exit strategy? Mike There's a whole set of obvious answers about being bought by an EDA player. Or there are foundries or equipment manufacturers TSMC, UMC, Applied Material, KLA-Tencor. It's just whichever guy has the biggest checkbook. [Laughter] John You close to a sale? Mike Whoever can write the biggest check. [More laughter] John Verisity? e? Vera? Is Cadence where tools go to die? [Hearty laughs] Ted Verisity was a very profitable deal. [In fact], in this particular conference, we're talking about Vera and e. John But Cadence isn't pushing e. Ted With respect to the technology, e is the largest installed based of verification users in the world with 50 millions lines of code. It's what people run to today to get the job done. Verisity dying? Their president is running our entire verification division! John Specman? Ted There are 8 companies using it! John Who told you that? [Laughter] Ted I talked to our marketing group! [More laughter] John Antun? e? Antun SystemVerilog is our focus. John Joe, does ModelSim support e? Joe I have no idea. [Big laugh] John Kathryn? Kathryn We're not doing e. John Gary? Gary There are 5 languages and e is one of those languages, and it's doing quite well. SystemC, the test portion, is half-assed by comparison. [Howl] Brett Gary, I thought your were my friend! For those who use C, they can access IP written in C, test cases in C, and verify in C. John Meaning ANSI C? Brett I'm not just talking about hardware. Now they're doing hardware, firmware, software and test benches in SystemC. John Brett, why did Jacob Jacobsson [former CEO] leave Forte? Brett Jacob was 5 years with the company. John But he left so suddenly. Brett It's not as bad as that. The Board and Jacob decided it was time to move to the next level of management. David Sear, the new CEO and my boss, is right here in the middle of the room. [Brett leaps out of his chair, off the stage, switches on the overhead projector nearby, and puts up a cartoon of Cooley standing on a box of explosives labeled SystemC.] John [Distinctly blushing, stammers] Oh shit! You've thrown off my rhythm. [Moment of hilarity reminiscent of 2005] uh Gary? SystemC? Gary There are three different communities using the tools. If you look at Celoxica and Forte, they're completely different. Forte is for semiconductor design, and Celoxica is for the system guys. The problems [they're addressing] are completely different. Celoxica first optimizes and then slams [the design] onto an FPGA. John So you don't favor Forte? Gary No, I ask what the hell are you guys trying to do? CatapultC is more about competing with Celoxica. Brett We do lots of FPGA designs we're not being outsold! We've got 90 percent of the market in Japan, and 6 out of the top 10 semi's in the world. This technology is growing and something you're going to have to learn to use. John Rajeev? Are you cutting tools prices overseas in India and China? Rajeev That's an absolute myth. We have a deal with Texas Instruments in India, but that deal was struck with TI Dallas. There's no difference in our prices in India or China. In India, the bigger companies are doing complex designs at 90 and 65 nanometers. In China, we're worried about IP protection, but we're not dealing with bigger companies in China mostly smaller homegrown design shops. There's no ASP difference between China and India. John Ted? Ted We deal with multi-nationals. There are lots of opportunities to sell into China, but they're buying the lower-cost technology. John Antun? Joe? Antun We dont differentiate prices [in different markets]. Joe No. John Gary? Gary They're telling the truth. China there's no price dumping at all. Ted nailed it most IC design in China is at 250 nanometers. We did see some price dumping in India, but it was brought to the attention of the presidents of some of the EDA companies and it stopped. John Rajeev, can you expand on the IP problems in China. Rajeev When you see a company doing 12 chips in 9 months, you know [there's pirated software involved]. But when they're only doing 250 nanometers, they say, 'Why should I pay?' Now the multinational companies [are forcing the situation] to be policed from outside of China. John Ted? Piracy in China? Ted [Recently], somebody showed me a website. It's all in Mandarin, except two words "Cadence" and "Synopsys" and next to each word was a picture of a disk and the size of the file for downloading. There's more money to be made in software today than in drugs it's selling by the kilo [byte]. That site was brought to the attention of the Chinese government, and they took aggressive action. The Chinese government is now engaged [and addressing the problems]. John Antun? Antun We are pleased to see that the Chinese government is now taking a stand on this issue. John Okay. That's it. Everybody go have a beer.
************************************* Epilogue: Some folks attending DVCon were not happy with the BigWigs panel, wondering what it had to do with design and verification, and the collegial spirit of DVCon. Maybe they're right maybe Cooley's panel belongs at DAC where it would draw an even bigger crowd. Or maybe the real question is: Which is the bigger problem? The Messenger or The Message *************************************
Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com
|