Dueling DACs

1888. One thousand eight hundred eighty eight. That’s the number of conference attendees reported in the “Preliminary Attendance Figures” for the 46th Design Automation Conference (DAC) held last week in San Francisco. The press release proclaims attendance was up “more than 12% over last year.”

Four years ago, in 2005, in an article titled “Ditchin’ DAC” I gave a brief history of the Design Automation Conference and wove a controversial tiny-tome telling why I thought the conference and the industry it represents were walking hand-in-hand down a daisy-lined path to doom. At that time, I had attended 21 of the 42 DACs. I skipped that one and wrote about it instead. Reactions ranged from wild enthusiasm and agreement to polite but pointed suggestions from media professionals who represented the conference at the time. It was even suggested that I join a committee to help devise the strategic plan for future DACs. Convenient. Hard to write an article criticizing something if you helped plan it.

In preparing to write this article, I went back and reviewed that one. Surely, I thought, I would wince a bit – surprised at my naivety of the time, or embarrassed by subsequent developments in a future I failed to foresee. Because of my role as an industry analyst and commentator, I often have to throw my opinion out there – complete with predictions for the future. Unfortunately, those forecasts are not left in some dusty paper archive to die a dignified death. Thanks to the tremendous efficacy of Google and the elephantine memory of the world-wide information repository we all helped to create, my words regularly re-surface to haunt me years later – with astute readers all-too-helpful in pointing out inconsistencies between past speculations and current reality.

This time, I have to say that four years have proven my prognostications to be painfully on track. I won’t repeat that article – or the points it made here because – hey, you can still read it – and you should. Go do that now. I’ll wait…

Oh, hey, welcome back. OK – so now to the present.

DAC doesn’t want to die a slow death. They hire marketing and PR professionals to do their magic. These people are not evil. In fact, they’re very nice and quite professional. They do their job, and their job is to make it seem like DAC isn’t headed downhill. They’re pretty good at it. Let’s take a look:

In this press release they report that 2009/DAC46 attendance is “Up More Than 12% Over Last Year” (if you capitalize all those words, they seem more truthful, right?) The press release says that there were 3247 exhibit-only attendees plus (our magical) 1888 technical conference attendees.

Our calculator says that’s a total of 5135.

Last year’s press release says there were 2405 exhibit-only attendees and 2423 technical conference attendees.

Our calculator speaks up and says that means we didn’t have a 12% increase. It’s more like 6%. See how this marketing stuff works? You can usually double the boost and nobody will catch on. That’s spin-doctoring 101. However, there is more advanced stuff going on here. The “real” attendees – full-technical conference registrants, dropped from 2423 to 1888 from DAC 45 to DAC 46. That’s a 28% year-to-year decrease. Ouch!

So, let’s review – what do you do when technical conference attendance drops 28% in one year? You issue a press release saying that attendance is up 12%. Problem solved. DAC Rocks! Our marketing folks earned their keep.

“OK, Kevin – you’re just being mean now. You probably found one silly math error, and you’re turning it into this big conspiracy-theory marketing thing. DAC is awesome – you know it is – you’ve been 25 times!”

Yes, that’s right, voice-in-my-head-person-representing-the-audience. I have been to DAC 25 times. I go for the Denali party. This year, I didn’t pre-register and Denali ran out of tickets. I’m angry. There – full disclosure. Read on.

There is actually more magic to getting the numbers up than we see above. The 2008 DAC (45th) was in Anaheim. The 2009 DAC (46th) was in San Francisco. Pop Quiz – which one is easy day-trip driving distance to Silicon Valley? Bing – that’s correct, San Francisco – come up and get your prize. What else is different from 2008 to 2009? The re-emergence of “Free Monday”

Now, anybody in Silicon Valley with a little time on their hands can drive up to the Moscone Center, pay outrageous parking fees, and get into DAC for free. Take that, DAC 45 (2008) in Anaheim! Now we’ll get our attendance boost! Think that’s all we marketing folks have got? Oh no – check this:

That’s right, it’s not just a technical conference and trade show. It’s a “Daycation” – you can bring the whole family (if they’re old enough) and then “DAC could easily become a family event.” Every member of your family can be one of those 2405 “exhibit-only” attendees. Now we’re talkin!

This probably seems like one of those out-of-context rants where I selectively pick statistics and try to make it look like a conspiracy. You know – like the photo of US President Obama that makes it look like he’s checking out the cute girl, when he’s really helping a different lady down some steps.

So, let’s get back in context and look at the bigger picture:

DAC folks don’t always give us all the statistics, and they don’t always break them down the same way. However, we can get some pretty good estimates form the data they do give:

Technical Conference and Exhibit-only Attendees:

DAC 46 (2009 – San Francisco): 5135 (1888+3247)
DAC 45 (2008 – Anaheim): 4828 (2423+2405)
DAC 44 (2007 – San Diego): 4701
DAC 43 (2006 – San Francisco): 6877
DAC 42 (2005 – Anaheim): 5506
DAC 41 (2004 – San Diego): 5574

Obviously, San Francisco is a big boost factor. With an average attendance of 6005 for the last 2 DACs there, but an average of 4764 for the last 2 non-San Fransisco DACs, we put the San Francisco bonus at about 26%. Specifically, Free Monday pulls about double the number of guests in San Francisco as it has in Anaheim and San Diego.

Let’s do a little city-adjusting, then, and boost ALL these DACs to a San-Fransisco Equivalent number. (Our audience should be comfortable with this kind of adjustment – it’s like “equivalent LUTs” in an FPGA.)

That gives us numbers like:

2009 – 5135
2008 – 6083
2007 – 5923
2006 – 6877
2005 – 6937
2004 – 7023

If you see a 27% decline over a 5-year period, you’re just like us.

See any trend there?

If we look at actual conference attendees – you know, the folks that go to a conference and actually attend the sessions, the trend is even worse: (and since most of these people pay to attend the conference, they also can afford to travel, so we’re not giving a San Francisco boost here)

2009 – 1888
2008 – 2423
2007 – can’t find it, can you?
2006 – 3173
2005 – 3074
2004 – 3246

That looks like a real, non-adjusted, 42% decline in technical conference attendees over the same 5 year period.

How about “participants” – the people that come to DAC to sell stuff to the others?

2009 – not reported yet.
2008 – 3543
2007 – 3550
2006 – 5077
2005 – 4689
2004 – 4944

Looks like another near 30% decline.

What about companies exhibiting?

We count 201 on this years list – matching the “over 200” on the press release. However, if you look at that list, 14 of those companies are TSMC, and 3 are Synopsys.

At the previous San Francisco DAC (2006), there were 251 companies exhibiting, so we have another 20% decline, this time over only 3 years, and we don’t factor in the nod-and-wink of 14 TSMCs (who had a wonderful, large booth with lots of mini-booths inside dedicated to specific partners – most of whom were also exhibiting elsewhere on the floor.)

So let’s wrap this math stuff up – over the past 6 DACs or so, everything is in a steady decline. Full-conference attendees off by 42%, total attendees and exhibitors both off by 20-something percent, and the number of companies exhibiting down significantly. Makes you pretty proud of that “Up More Than 12%” doesn’t it?

Why is DAC declining? From most of the speeches at the conference, we might be distracted into thinking that it’s the global economic downturn, and that the EDA industry is just struggling along with the rest of the industrialized world trying to weather the storm. As you can see from DAC, however, this industry has been in a tacit decline since long before the econo-quake hit last year. EDA has kept most of its focus on the ASIC/SoC business – and with design starts steadily declining and process nodes growing more complex, they end up making more and more complex tools for smaller and smaller audiences.

EDA has ignored the majority of electronic designers in the world – who are not doing ASIC/SoC design. Probably 95% or more of the people designing electronic systems in the world today are not designing ASICs, but are instead using FPGAs and commodity processors on custom circuit boards. With a few exceptions – EDA has not invested heavily in serving this 95%, but has instead put the bulk of their development resources into supporting the high-flying 5% doing nanometer IC design.

Today, out of the big four EDA companies – Magma is in serious trouble, Cadence is on the ropes, Synopsys is experiencing a period of growth, but largely at the expense of the two ailing competitors, and Mentor has diversified their product portfolio well enough into non-ASIC areas like FPGA, PCB, and embedded system design that they are less exposed to the dwindling ASIC design community. Altium, the not-appearing-at-DAC, alternative EDA company has put 100% of their focus on the non-ASIC community. The next few years will let us see how these various strategies compare. Meanwhile, DAC has to keep finding better, and more optimistic strategies for marketing an event whose utility is dwindling rapidly and whose cost-effectiveness as a sales and lead-generation vehicle for the EDA industry is indisputably terrible.

The marketing doesn’t stop with DAC, however. While we’ve got the calculator out, and as a side note – I noticed a booth from a publishing company (full disclosure again – this is one of the competitors of our company, and I won’t say which one) with a sign that said their website was used by “over 30,000 EDA industry professionals.”

You know what I’m thinking here…

Synopsys – 5700 employees
Mentor Graphics – 4425 employees
Cadence Design Systems – 4400 employees
Magma Design Automation – 732 employees

Total for the “big 4” – 15,257 (yes, we know – that guy that mows the grass at Cadence isn’t really what you’d call an “EDA Industry Professional,” we’re being generous…)

Let’s do a very liberal addition of 30 EDA Industry professionals each for the 200 companies exhibiting at DAC – that’s another 6,000 (yes, we know again, we’re double-counting some) That brings us to 21,000.

Now, let’s throw in 300 for Altium – the biggest EDA company you won’t find at DAC. Let’s add another 1,000 each for Xilinx and Altera who, although not strictly EDA companies, have large design tool development teams.

We’re still at no more than 24,000 EDA Industry Employees – and that’s if everybody from engineering to human resources reads this company’s publications. Perhaps there is a bit of marketing in all of us. Our own IC Design and Verification Journal has about 10,000 very real e-mail newsletter subscribers. We’re proud of that!

Why does all this matter? Let’s go back to where we started – 1888 – the number of technical conference attendees. These are the hard-core customers of the EDA industry. To reach these 1888 individuals, over 3500 EDA Industry professionals traveled to San Fransisco for a week. Over 200 companies rented booths, sponsored parties, hired magicians, flew on planes, stayed in hotels, splurged on meals… We’re guessing (and our calculator didn’t work hard on this one) that roughly $20M of industry money annually goes to wooing those 1888 individuals at DAC. If so, that’s over $10,000 per attendee.

It’s kinda surprising they have to pay to register.

“The coldest winter I saw was the summer I spent in San Francisco,” quipped Mark Twain. It’s good to know some things never change. The City by the Bay (or simply “The City” to those within driving distance) didn’t disappoint, as the comforting midsummer fog enshrouded nerds from near and far. Grab an Irish coffee and hop aboard the cable car; it must be time for the 46th annual Design Automation Conference.

San Francisco is practically an island, surrounded on three sides by water, so when it’s baking hot elsewhere in California, the low atmospheric pressure draws moisture over the top of San Francisco and, like Brigadoon, it disappears from view. Only the tips of the Golden Gate Bridge peek through the fog, like orange traffic cones guiding visitors across the bay.

Beneath the gray blanket, DAC offered a veritable smorgasbord of snacks and larger meals. The biggest deal, from my perspective, was Mentor’s acquisition of Embedded Alley. The former, you’ll recall, is one of the Big Three in the EDA industry but also a major player in embedded development, with its Nucleus operating system and Microtec and EPI tools. The latter is an embedded-Linux startup based in Silicon Valley. By acquiring Embedded Alley, Mentor moves itself squarely into the suddenly hot embedded-Linux camp, alongside players like MontaVista and, as of last month, Intel.

Embedded Alley is more than just another pretty embedded-Linux face. The company specializes in Android development – or at least, as much as any company can, given Android’s short time on this planet. Since Android is new to almost everyone, yet is intended for fast-moving product categories, hired help is important. Everyone wants Android now, but nobody has any experience working with it. That puts Embedded Alley, and now Mentor, in a good position.

So what does this do to Mentor’s other embedded operating system, Nucleus? Mentor spokesmodels were careful to point out that Nucleus and Android can quite happily live side by side, and I tend to believe them. Most cell phones already have two or more processors, and plenty of them also run two operating systems. Larger devices (mobile Internet devices, for example) can also tolerate twinned operating systems, either because of their two-headed nature or for reasons of compatibility. Thus, it’s no stretch to think that future devices could use Android for the main OS while Nucleus lurks inside doing the real-time dirty work. It’s worth pointing put that Nucleus is royalty-free (as is Android), so it’s appealing to high-volume manufacturers who hate paying tribute to their software suppliers.

Now that Intel has scooped up Wind River and Mentor has acquired Embedded Alley, we’re left to wonder, “Who’s next?” Embedded Linux development has gone from being a curiosity, to being fashionable, to perhaps being profitable. Even Intel doesn’t shell out $800 million just to acquire a few programmers and obnoxious marketing dweebs. There are cheaper ways to hire talent and acquire customers. Potential suitors may be knocking on the doors at MontaVista, PTR Group, Secret Lab, and others. Now is a good time to be an embedded Linux developer.

How Do They Do That?

Reverse engineering has always had the faint whiff of cheating about it. Real engineers don’t pry open others’ products, the thinking goes. Conversely, “Those who do not learn from history are doomed to repeat it.” There’s a lot to be learned from others’ mistakes, and possibly even more from their successes. Your views on this topic will likely influence your opinion of Canadian company Chipworks.

Chipworks is, in a nutshell, in the business of reverse engineering. They help chip-level engineers understand how other chips are designed, through a number of surprisingly elaborate, ingenious, and expensive techniques. For starters, Chipworks will decapitate any chip you care to name and produce a startlingly detailed floorplan, right down to the individual transistors, resistors, and capacitors. You can have the whole chip denuded, or just a portion that catches your fancy. Prices range from five figures on up, depending on the specific chip and how much you want to see.

But wait — there’s more! Seeing the physical layout is one thing, but understanding the circuit is another. And here’s where the magic comes in. After Chipworks has laid bare the physical design of a chip, it can then hand you a working schematic. By carefully examining the silicon-level components and their connections, the company can back out the entire schematic, complete with annotation of each component’s physical location on the chip. You get both the logical and the physical design of the part.

It gets better. Using Chipworks’s own schematic-capture software, you can redraw the schematic, rearranging, reorganizing, and renaming components and netlists to make the whole thing more readable or to suit your personal design style. The software, however, prevents you from breaking the actual circuit design. For example, it won’t let you add components to a net that aren’t there in the real chip, nor can you remove components or change their logical relationship. You can beautify the schematic all you want, but the result is guaranteed to mirror the actual device.

So who would want to do this? Quite a few people, judging from Chipworks’s headcount of 120 and its apparent financial health. Lots of engineers are curious how their competitors managed to achieve a certain clock frequency, or how they keep power consumption so low. A detailed analysis of someone else’s chip can sometimes lead to a “Eureka!” moment. “So that’s how they did that!” Nothing wrong with that; we all learn from our peers.

Other times, the motivations are more legal than technical. Patent lawyers often call upon Chipworks to prove (or disprove) infringement. Since semiconductor patents rarely cover physical-level circuit design, it’s often tough to tease out the legally relevant detail merely by decapitating the chip. Some circuit analysis is required, and that’s where Chipworks comes in.

Then there’s the clean room paradox. To avoid accidental patent infringement, some engineering teams deliberately avoid looking at competitors’ parts, keeping their engineers’ minds unsullied and their lab a “clean room.” On the contrary, says Chipworks president Julia Elvidge, you’re better off examining other chips as closely as possible so that you know you’re not infringing. Why leave infringement to chance when you can be sure – and presumably prove in court – that you’ve designed around any similarities between your chip and theirs. Given the capriciousness of the patent courts, she’s got a point. Disassembling a handful of relevant chips gives you a fighting chance of avoiding legal landmines. It’s either that, or tiptoe through the minefield blindfolded, hoping your pure intentions will protect you. Good luck with that.

Dueling DACs

Related

Leave a Reply Cancel reply

DAC Cetera

Related

Leave a Reply Cancel reply

featured chalk talk