Why is EDA so buggy?

Posted on January 27, 2009 by paulmcl

BuggyI have sat through numerous keynote speeches by CTOs of semiconductor companies berating the EDA industry for shipping tools that are full of bugs and that are late, not ready enough in advance of the appropriate process node. Of course this is true, and nothing I am going to say is to imply that improvement is impossible. But it is an intrinsic problem, not just laziness or incompetence on the part of EDA vendors.

In an informal setting, that is to say over a beer rather than in front of a large audience, I tell such CTOs that if they want more reliable software then they can simply use an old version of the tools that has been shipping for years. Tools like that get pretty solid. Of course that is simply a glib response since I know that they can’t design 65nm designs with 130nm tools.

So my next suggestion is that the EDA industry could delay shipping new tools for an extra year or so to allow extensive testing. Another glib response since we all know that the tools are already late and even if it were possible to test them extensively without shipping them (what would we use for test data?) the delay would be unacceptable.

There are two reasons for this state of affairs, one technical and one economic.

The technical reason is that it simply isn’t possible to know everything necessary about a new process node far enough ahead of time to allow for a robust development cycle. For example, even with a huge team of people ready to go it is impossible to develop a full suite of technology for 22nm design. We just don’t know everything we need to know to get started and neither do the semiconductor companies and their equipment suppliers. Inevitably the tools will be later than desired and customers would rather have buggy tools now than better tools in six months. It is the EDA version of MacArthur’s dictum that a good plan violently executed now is far better than a pefect plan executed next week. In fact, whatever schedule is chosen, there are always customers lining up to be beta sites, so that they can get their hands on technology earlier, and pressure to ship even earlier even though the tools will be buggier still.

There is also the fact that it is not really possible to develop an EDA tool in a vacuum. There need to be libraries and designs in the process node to be used to test and wring out the code. A new tool is often rock solid on old designs, it is the new bigger more complex designs that break the tools in new ways.

The economic argument is that EDA has to support several process nodes at once and recoup its investment in a timely manner. Those same CTOs who berate EDA companies for not being aggressive enough, work at the same companies that have CAD managers who insist that resources be diverted to back-patching bugs. They want fixes that are already available added back into obsolete (and no longer officially maintained) versions of the software because their design groups haven’t got round to switching. And those CTOs have finance managers who don’t want to increase their budget for leading edge tools only used by the small number of advanced groups.

The effect of all of this is that EDA companies make their money and recoup their investment on processes only when they become mainstream. They cannot afford to make that investment too far ahead of the mainstream for economic reasons as well as technical ones. The fact that the mainstreaming of the most advanced processes is slowing is already starting to strain this model somewhat since it delays the time to payback.

One way to improve the quality of EDA software would be open source. But if undertaken by the major suppliers, this would also destroy their business model and probably thus result in no software at all. However, EDA moves too fast for open source to simply clone the successful products in parallel. The Econtalk podcast with Eric Raymond pointed out another industry that moves fast and where open source has little impact: games. By the time a game is clearly a hit it is too late to start an open source project to clone it. The gaming community will have got bored with that game and moved onto something else by the time the open source free version is complete and widely available.

Other industries don’t have this problem because they don’t move so fast. Technologies in automotive are adopted in decades not a year or two. CAD for automotive has a lot of time to adapt. But EDA is stuck with a very short reaction cycle and even if the ROI was richer, it is not clear that much would change.

Posted in eda industry, engineering | Comments Off

Gong xi fa cai

Posted on January 26, 2009 by paulmcl

gong xi fa caiIt’s Chinese New Year, or the spring festival as the Chinese call it, the start of a year of the Ox. It is also, by coincidence, Australia day, the equivalent of July 4th down there. But there is not much potential in comparing China to Australia so let’s look at China and India instead.

They are both enormous, with over a billion people each. Both are making huge strides towards modernity having opened up to the outside world in the last twenty years or so after having very protectionist planned economies with the usual unimpressive results.

Both China and India have huge disparities in income and standards of living, with a strong growing middle class but a large population of subsistence farmers and people living on the margins in cities. And there are lots of cities, over 100 cities of over 1 million people in China so most of them you’ve never heard of. You are much more conscious of the poverty in India because everything is mixed up there. You can be in, say, the Cadence buildings in Noida and you could be in California except that some of the women are in jeans and some are in saris. But walk outside and there are cattle wandering around, people drying dung and beggars everywhere. In China, the eastern cities are prosperous and the marginalized mainly live in the rural west so if you visit Beijing and Shanghai you won’t see them.

The other interesting major difference is the system of government. India is a bit like California with a thriving private sector and dysfunctional public administration. For example, every company has to have its own generators since the power companies are micromanaged by the politicians, can’t force anyone to pay their bills, have controlled prices and so can’t make any money and thus can’t invest. As the largest democracy in the world, those rural and illiterate poor are able to force bad policies in many areas.

FallowsBy contrast, China has a non-democratic government. But as a result it is perhaps easier for it to pursue sensible policies that would not necessarily be popular. So they have managed, originally in the south far from prying eyes in Beijing (which is in the north for those of you with no Chinese geography: bei is the Chinese for north) to have business friendly policies that have ignited a boom that has lifted more people out of poverty faster than anything in history. Deng Xiaoping, who is credited with orchestrating the change in policy, is a hero in that sense, despite the unacceptable blot of Tiananmen Square. To read a current viewpoint on various aspects of China, I recommend James Fallows’s articles in Atlantic Monthly, collected in the book Postcards From Tomorrow Square.

Because of their populations and a strong cultural emphasis on education as a means to advancement, both China and India produce a lot of well-trained people in any area such as engineering and computer science. India, of course, has the legendary Indian Institutes of Technology (IITs). Almost every successful Indian you meet in Silicon Valley is from one of the IIT (even Asok, the intern in Dilbert). China also graduates a huge number of engineers, There is some debate about the actual number and whether many of what Asia calls an engineer are really what we would call technicians. But it is pretty clear that both countries are graduating more real engineers than the US. This is not necessarily bad for the US. It is the dirty secret of Silicon Valley that so much of the engineering is done by people born outside of the US. When I ran a 200 person engineering group in the 1990s I estimated that over half were immigrants, starting with me. So it is all the more important that, as I said last week, the US have sensible immigration policies to make it easy for such people to come here (or stay here when they finish their advanced degrees).

Of course the current downturn will have a major impact on both countries. China, in particular, has to keep growing fast enough to make the country rich before it becomes old due to its inverted demographic that will be created by the one-child policy. It will be interesting to compare China and India and whether Lee Kuan Yew’s view that you need to liberate economically before liberating politically is proven right or wrong. China (and Singapore, of course) are exhibit A; India is exhibit B.

Posted in culture | Comments Off

Recutting the semiconductor pie

Posted on January 23, 2009 by paulmcl

Qimonda, the memory spinout from Infineon, itself the semiconductor spinout from Siemens filed for bankruptcy today. This is not all that significant for EDA since memory companies have few designs and enormous volumes. Larry Grant, the corporate counsel when I was at VLSI, had come from Micron and was amazed at the difference with an ASIC business: “At Micron we put around one design into production per year, here we put a couple into production most days.”

One problem I understand Qimonda has is that their buried wordline technology is different from the technology used by all the other DRAM manufacturers and requires some special semiconductor equipment to manufacture which makes them even less price competitive. Their fab is in Dresden in Germany (in the old East Germany) which is also probably a handicap both in terms of labor cost, and flexibility in adjusting the size of the workforce.

More significant for EDA is that I think this is another symptom of the consolidation that is going to occur in semiconductor through this downturn. Some of this consolidation will be the usual efficiencies of scale but there is also a new wrinkle in play this time.

The large semiconductor companies have historically owned their own fabs for at least a large portion of their manufacturing capacity. They then needed to have enough product lines to fill the fabs, which in turn meant a certain scale. They were called IDMs, integrated device manufacturers, to distinguish them from fabless semiconductor companies who did design, marketing, sales but left the manufacturing to TSMC and UMC in Taiwan. Even if they were wanted to own their own fabs for some reason, they didn’t have enough products to keep them full. However, as fabs have got more expensive and the technology development necessary to have a state-of-the-art process has risen, the size necessary to justify owning a fab has increased beyond the volume of almost any semiconductor company outside of memory and Intel. So companies like AMD and Texas Instruments are going completely fabless.

Once semiconductor companies are fabless, the motivation for having the existing set of divisions in the same company doesn’t necessarily make that much sense. NXP (née Philips Semiconductors) has sold its wireless division to ST. That transaction was probably driven by the need to raise some cash by NXP’s private equity owners, but I think it is typical of the sort of transaction we are likely to see. Freescale (née Motorola semiconductor division) has its wireless division up for sale too.

It is interesting (well, to some people like me) as to why companies exist at all. Why don’t we all just be independent contractors and take orders for our services. The first person to think much about this was Ronald Coase who realized that the transaction costs involved in us all operating that way would swamp us, and so it is much more efficient to organize into companies with a more directive style. His work is known as (surprise) the Coase Theorem and dates from 1937. For a semiconductor company with a fab, it was more efficient to organize into a company built around manufacturing and have enough design and selling capacity to feed that beast.

I expect that many more changes in how the semiconductor industry is carved up into different companies will come about in the next year or two. Not just mergers, although there will probably be some of those, but divisions being sold so that when the dust settles there will be just a handful of serious competitors in any market space. For example, it looks like wireless is going to come down to TI, ST, Qualcomm and perhaps Samsung.

Posted in semiconductor | Comments Off

What color is a green card?

Posted on January 23, 2009 by paulmcl

Most American don’t know the answer, which is today off-white. Mine is alongside. Presumably it was green once.

I’m an immigrant. I was born and brought up in the UK and came to the US in 1982 on an H-1 visa. After four years I got a green card (which I think was pink in those days) and became a permanent resident. I can live here indefinitely (provided I don’t commit major felonies). Actually, the real timeline is that after 2 years I applied for a green card but it took another couple of years for the government to process the paperwork. There is no upside in good immigration processing since immigrants don’t vote. Imagine the outcry if it took two years to get a passport.

It is one of the strengths of the US that people like me could easily come here and contribute, along with other more significant immigrants such as Sergey Brin, Jerry Yang, Albert Einstein, Bob Hope, John Muir, Carlos Santana and many others. That doesn’t happen much in China or Mexico. The mayor of Vienna is not an American immigrant; Arnold Schwarzenegger came in the other direction.

On a personal note, I’m very grateful for the opportunity that the US gave me.

Most discussion of immigration centers on illegal immigration of poorly educated Mexicans, but all the evidence seems to be that while poor Americans may lose slightly through increased competition for low-paid jobs, they gain even more from things like lower cost food. But as a strategic issue for the US I don’t think this is all that big a deal. The US economy doesn’t stand or fall on the basis of how many Mexicans work here.

Much more important is the idiotic legal immigration policy we have for educated people. The most insane part is allowing students to come here for PhDs (55% of engineering PhDs are foreign-born) and expelling them when they are done, since there is no automatic route to stay here. Plus we make it harder than necessary to come here to study in the first place. First loss, these are just the kind of people that we need here in the US to drive technology businesses. Second loss, even if students go back to their home countries, they go back with a positive image of the US to counter the negative views of people who know little about the country.

The H-1 visa quota for each year opens up on 1st of April and closes immediately since twice as many applications are received that day as are available for the entire year. But those are for visas starting October 1st. When I came to the US either there was no quota or it was much higher than the number of applicants. If a company wanted to hire a qualified candidate from overseas (me) then it applied for a visa, waited about 6 weeks and got it, then the person could start. Today it is impossible to hire someone on that basis since the delay is about 9 months on average until the next October 1st after the next April 1st, and then there is only a 50:50 chance of getting a visa anyway. Companies can’t be bothered with such a lengthy uncertain process.

The result is that H-1 visas have become a way for overseas consulting companies, especially Indian, to apply for large numbers of visas knowing some will get through and their employees can then come here months later. This is not necessarily bad but it also squeezes out everyone else, every talented person that an American company wants to hire from overseas, every student who wants to stay on once they have their doctorate and so on. The best solution if it is politically unacceptable to do the sensible thing and remove the cap, would be to ‘auction off’ the visas. But I don’t mean by paying bids to the government but by using the salary that the employee would receive. The higher the salary paid the easier to get a visa for that employee. The Indian job shops would be ‘outbid’ by PhDs.

I can do no better than to quote James Fallows, an editor at Atlantic Monthly who currently lives in China (and used to live in Japan during its heydey in the late 80s). Here he is talking about an Irishman who lived in southern California but had to move to China because he couldn’t get a visa to remain here:

“I might as well say this in every article I write from overseas: The easier America makes it for talented foreigners to work and study there, the richer, more powerful, and more respected America will be. America’s ability to absorb the world’s talent is the crucial advantage no other culture can match—as long as America doesn’t forfeit this advantage with visa rules written mainly out of fear.”

Posted in silicon valley | Comments Off

Venture capital for EDA is dead

Posted on January 22, 2009 by paulmcl

Money down the drainThere have been a couple of recent articles about venture capital in Business Week, ABC News and elsewhere. They have focused on venture capital being broken since many funds are losing money and even that venture capital, in aggregate, is investing more than it is taking out in exits. There are short term problems with the current downturn too. For example, I’ve heard of limited partners (the actual investors) refusing capital calls when a fund wants to make an investment. As a result, VCs have essentially suspended new investments for Q4 of 2008 and Q1 of 2009. Of course one of the biggest problem of all is that the market for IPOs is closed completely.

In the EDA world venture capital has been broken for some time. VC funds are simply too large and too risk averse. In effect they have become private equity banks. The reason size matters is that a $500M or $1B fund simply can’t make $5M or $10M investments. A partner can only serve on a limited number of boards and funds of this size would need to make hundreds of investments. Instead, they need to focus on making fewer big investments. However, early stage investments simply don’t require that much money, and the amount of money that they do require is continuing to decrease in the software world. A cynic might look at the enormous VC investment in cleantech, especially solar, as being attractive simply because they require a lot of money to get to manufacturing. On the other hand, Web 2.0 and EDA startups require little money, often less than $10M, to get to revenue.

The herd dynamic means that VCs all either want to invest in a sector or don’t. A VC told me once that “venture capitalists make sheep look like independent thinkers”. So if the big guys like Sequoia or Kleiner-Perkins are avoiding a sector (perhaps just because their funds are too simply big for the sector or the limited partners have directed them) then everyone else seems to avoid it too. EDA has been out of favor for some time since each startup does not address a “billion dollar market” nor is it not going to be a phenomenon (although without much revenue) like Facebook or mySQL.

In the software world, the cost of hardware is basically zero (we all have a computer already and even buying a new one is something we can afford without a VC) and software productivity is improving all the time. In the web world this is driven by languages like Python and Ruby, along with environments like Django or Rails. And for web infrastructure there is Amazon S3 and Google App Engine. All the costs are variable so big upfront investment isn’t needed even to scale to millions of users. In the EDA world this is driven by the same languages, along with infrastructure like Open Access that mean that a startup doesn’t need to spend its first year or so building its underlying scaffolding, it can focus immediately on code that adds real value to users.

Paul Graham of Y-combinator thinks that VCs have become redundant in the internet space. In EDA the amount of money required is low enough that personal investment and private investors are sometimes enough to get to positive cash-flow without any venture capital at all. Altos and Apache are both profitable and were funded entirely privately. Denali has been private and reportedly very profitable for a long time; they certainly throw a good party.

How long before a venture capital fund decides to buy a company, removing any semblance of being genuinely different from private equity banks?

Posted in investment | Comments Off

Why did EDA have a hardware business model?

Posted on January 21, 2009 by paulmcl

Calma systemEDA really started back in the 1970s (late 60s in fact) with companies like Calma and Applicon. They drove the first EDA transition from cutting rubilith, red sticky plastic that was physically cut with X-acto knives, to digitizing the input and generating pattern-generation tapes for automated mask-making equipment. One legacy of this era remains with us today: Calma’s system was called the Graphical Design System or GDS, and the second generation (32-bit!) was called GDS-II. Normally it stored its data on disk but it had a format, called “stream format” for writing the data out onto magnetic tape. The disks were too small to keep the designs there permanently. This format, GDS-II stream format (or often just GDS-II or GDS) was the standard for decades for moving layout data between systems and from design environment to mask shop. It is still not dead although it is definitely coughing up blood. The transition to Oasis or other formats has gone much slower than expected. Everyone supports GDS-II and so it is the least-common denominator format.

Calma and Applicon initially thought of themselves as hardware companies. They sold computers. Calma was actually a Data-General minicomputer (it’s nothing directly to do with EDA but if you’ve never read it, you must read Soul of a New Machine about development of the 32-bit version). The business model was the same business model as most hardware was sold: you bought the hardware, digitizers, screens and so on. See the picture. And you paid an annual maintenance contract for them to keep it all running which was about 15-20% of the hardware cost per year. The software was simply bundled into the price. This was before the days of a separate pure software industry; almost all software only ran on one brand of computer the way cell-phone software or digital camera software is sold today: you can’t buy it separately, it’s bundled with the hardware.

The next generation of EDA was also hardware-based. Gate-level design was dominated by the DMV: Daisy, Mentor, Valid. Daisy built the Daisy Logician, Valid built the Scald-station (I think that was the name) and Mentor OEMed Apollo workstations instead of building their own. The business model remained the same: buy the hardware and pay an annual maintenance. I don’t know if the software was even a separate line item.

It is hard to believe, but back in that period there was a worry in EDA that as the hardware costs came down (Moore’s law) then software costs would have to come down too. It was hard to believe that someone might pay more for software than the hardware on which it ran. After all, they never had before. Today, when you can run millions of dollars of software on a box costing a few thousand dollars this seems comical. But go back to the cell-phone software I mentioned earlier. Maybe one day we’ll be paying $500 for good cell-phone software from an independent market, and then buying a cheap phone for $10 on which to run it. After all, that’s where the value increasingly is.

The next generation of EDA software, VLSI Technology (where I worked), SDA, ECAD (that together became Cadence), Silicon Compilers, SDL and other companies that I’m sure I’ve forgotten, wrote software that was more hardware independent. They would run on Vax (always) and one or more of those new-fangled workstation thingies from Apollo (or Sun once they made it to production). They would usually sell you the hardware if you wanted, but you could just buy the software and run it on your own hardware. The business model was the same old hardware business model though: pay an upfront license and annual maintenance of 15-20%. This was how we ended up at first with a hardware business model for a software business.

The final change was the switch to the time-based license, initially 3 year, that we largely have today. Essentially this is a software lease. Gerry Hsu1 is usually credited with this. He told me that he noticed that people like to lease cars so that they get a new car every 3 years or so, and decided to see if you could sell software the same way. It turned out to be a good idea and the financial side of the business liked it since it gave very predictable revenue. In any given quarter, most of the revenue was business booked in the three years before, and only a small amount from the new business booked that quarter.

Business is still done as a mixture of time-based licenses (recognized over the period) and term business (recognized up-front). But, as I said earlier, the mix is open to abuse and only the savviest followers of the industry realize how critical the percentage of ratable business is in trying to decide if an EDA company is doing well or not. It remains tempting for an EDA company to persuade a company to do that $10M deal as a term license instead of a time-based license. The $10M is recognized immediately and can fill an embarrassing $10M hole caused by the lack of $100M of time-based bookings.

1 By the way, Gerry Hsu is sometimes portrayed as a bit of a buffoon. But he was certainly extremely smart, and very perceptive. Just somewhat ethically challenged. I worked for him for 8 hours!

Posted in marketing | Comments Off

R.I.P.

Posted on January 20, 2009 by paulmcl

Small EDA companies seem to be falling by the wayside. Over the Christmas break Blaze DFM (insert cremation joke here) and Liga Systems both reached the end of the line. Blaze will certainly get picked up; it has customers and a valuable royalty-bearing relationship with TSMC. Liga not so much. Agility (née Catalytic) has also run out of money and its technology is apparently up for auction.

Even before the current downturn it was very difficult for EDA startups to raise money. VCs don’t like the EDA business model of selling licenses. There isn’t any opportunity for a big surprise upside. With only Synopsys rich enough to pick up technology at a premium if it wanted to, and with the IPO market already closed to small companies due to Sarbanes-Oxley, there is just no route to a liquidity event. The current downturn is making things worse. So expect more startups to shutter in the months to come as their current money runs out and even their existing investors prove unwilling to put in any more.

I heard of one company, that I’ll leave nameless since I don’t want to make things any worse for them, that is unable to raise money despite having marquee customers, revenue and competent management.

Posted in eda industry | Comments Off

Episode XLIV: a new hope

Posted on January 20, 2009 by paulmcl

ObamaIt’s inauguration day. Never mind how many of his policies I will disagree with, I still think that it is incredible that the US has gone in 40 years from having segregated water fountains and buses to having a black president. I find it hard to imagine an Indian prime minister of Britain any time soon, even without the segregation and legacy of slavery (at least in Britain; the British had plenty of slaves in the Caribbean sugar plantations).

So what does it mean for hi-tech, EDA, semiconductor and our ecosystem generally. I think that the dangers are two-fold: the market being superseded by picking of winners and losers in Washington, and entrepreneurial activity being killed off by laws written in Washington by people who have no experience of running a business at all, let alone a startup or tech business.

So the first risk is that success will be decided in Washington through lobbying and rent-seeking rather than in the marketplace. This is already evident today in the “green” industries which are largely subsidy seeking since most of the technologies are nowhere close to economical on their own yet. Did you know that in Texas that some wind farms actually pay the utilities to take electricity since they lose the subsidy if the electricity is not used? The reality is that the value of the electricity doesn’t just fail to cover the costs of the wind-farm, it is completely worthless at the time and place is is generated.

It is five years old now but I think that everyone should read TJ Rogers’s piece on “Why Silicon Valley should not normalize relations with Washington D.C.” We may, as smart guys, learn to lobby and do well in an environment like that, but it will not be good for the US overall nor for mankind in general if our entrepreneurial talents are used like the CEO of ADM to persuade the government to pass stupid policies like ethanol distillation that are personally enriching us at the expense of everyone else. They also slow down needed changes since old established companies (like General Motors) that are inefficient can delay necessary transformation by lobbying and penalize the companies that should be taking their place. Never mind new sunrise industries, but even Toyota or BMW suffer. For example GMAC recently were told that they could get bailout money if they raised $30B. They came nowhere close so, guess what, they got the bailout money anyway. And immediately they did they launched 0% interest financing and lower qualification standards.Without the cushion of federal money they would not be able to socialize their losses and keep their gains..

If successful companies are chosen in Washington rather than the market then the American success story will gradually grind to a halt. I will be just as effective as it was in Russia during communism or in many African countries today.

The second problem is that the democrats have been talking a lot of anti-free-trade rhetoric and increased regulation rhetoric. Sarbanes-Oxley seems to have largely killed off the IPO market by making the size at which it is possible to go public too large. And our regulation and planning processes are already uncompetitive. A friend of mine was on the board of Logitech. They built a new factory in China in, I think, 18 months from deciding to do it to mice and keyboards coming off the line. Never mind labor costs, which in semiconductor are small due to automation, it just takes too long to get things done. The US is not the worst. Logitech have been trying for 3 years to build a new office building near their headquarters in Lausanne and apparently were seriously considering moving the headquarters out of Switzerland since it was too hard to get anything done.

On the free-trade side the reality is that almost all semiconductors are made overseas, largely in Taiwan. With a fab costing $6B and up that is not something that is going to change just because tariffs change; things will just cost more and so sell less. I love the way Apple products say “Designed in California, manufactured in China.” Which of us gets the better deal? Would Apple be stronger if it had to build iPhones in Cupertino?

Silicon Valley does best when the discipline of the marketplace picks the winners and losers. I saw Vinod Khosla talk about green investment a few months ago. He berated Californians for their feel-good environmentalism: driving Priuses, putting solar panels on the roof of the Moscone Center in a notoriously foggy city, banning plastic bags which are more environmentally friendly than paper anyway. He pointed out that the only important technologies are ones that are economically compelling enough to be adopted in China and India. Everything else is window dressing and subsidy farming. China and India, with 2.5 billion people, are the market.

Today is the start of a new era. Let’s hope that for Silicon Valley it is the start of a new beginning and not the beginning of its sunset.

Posted in silicon valley | Comments Off

Open source again

Posted on January 19, 2009 by paulmcl

The blog entry on open source seems to have generated more comments than anything else. Maybe it’s because all the EDA users want software to be free, and all the EDA producers are worried that it might head in that direction. Everyone has an opinion.

In a seemingly off-topic thought, let me recommend Econtalk which is a weekly interview by Russ Roberts (a prof at GMU and Stanford) with someone knowledgeable on some aspect of economics interpreted in a wide sense. Last week it was Keynesian economics and the week before it was building schools in Africa. I typically burn it to CD each week and listen while commuting.

This week these two disconnected items tie together since Russ’s guest is Eric Raymond, author of The Cathedral and the Bazaar, talking about open source and the nature of the open source process. Recommended.

One thing about open source that I think people misunderstood is that I was not predicting that there would or should be open source EDA tools, or that the market was not big enough. I think open source is successful when the programmer and the user are the same person so there is no need to try and reduce the requirements to a specification. Or where the project is an open source copy such as creating open source Flash, or even Linux (open source Unix), so that the original serves as the specification. I’ve even seen people claim that if you need a specification the project is already off the rails. It is really hard to write good software for an application that you don’t understand well yourself, where you are not going to be your own user. EDA software is largely like that. Designers are not (good) programmers and programmers know scarily little about chip design

There seems to be a similar dynamic about many websites: Facebook, eBay, Yahoo, mySpace and many others were created to serve a need that the founders felt they needed filled for themselves, and then were smart enough to seize the moment. On the other hand I think there is lots of opportunity on the net for sites serving older people. The people who found web companies are young and as a result older people are underserved. But old people are on the web, they have money, they have time and they are a fast growing demographic. What’s not to like? Like in the open source case, the people who create such companies and write the code are unlikely themselves to be in their 60s and 70s so creating something successful is much harder.

Posted in eda industry, engineering | Comments Off

The innovator’s dilemma

Posted on January 19, 2009 by paulmcl

Clayton ChristensenThe Innovator’s Dilemma is a book by Harvard business school professor Clayton Christensen. I highly recommend the book both as one of the most stimulating and best-written business books (I know that is an oxymoron, but this is a book you will really enjoy reading). The basic thesis is that there are two kinds of innovation, sustaining (giving high-end customers what they want) and disruptive (giving a new set of less demanding customers something less that state-of-the-art). Sustaining innovation eventually gives people more than they want at a premium price point, but disruptive innovation often improves faster and eventually steals the main market from below when its basic capability addresses the mainstream at a lower price.

Here’s an example: Digital Equipment Corporation (DEC) built Vax computers in the 80s. Customers wanted more and more powerful Vaxen and had no use for the IBM PC when it came out, a low-powered machine that didn’t even have a floppy disk as standard, let alone a hard disk, when it came out. But eventually the PC destroyed DEC’s business (and it will almost certainly destroy Sun’s) as it got more powerful. The dilemma is that it is unclear what a company like DEC (or Sun today) should have done. They were not stupid, they could see the PC coming, and they even made several attempts to enter the PC market themselves. But it was of no use initially to their primary customers and they didn’t really have the capability to sell to the people who could make use of early PCs. By the time the PC was powerful enough to be of interest to the scientific and business computing segments, where DEC sold most of its kit, it was too late. Other companies (Compaq, Dell etc) were already established as the leaders and DEC was eventually dismembered with part going to Intel and part going to Compaq and so ending up inside HP. It is not that it was or is impossible to built a computer more powerful than a high-end PC, it is that the cost-differential is so large that very few applications justify paying such an enormous premium.

Clayton’s book has some other lovely examples: cable driven earth-moving equipment being driven out by hydraulic; steel mini-mills making rebar and gradually working up to high-grade sheet steel and so on.

InnovatorWhen I was at Cadence we had an annual engineering conference, a mixture of presentations of papers that could not be presented externally due to confidentiality, social getting together of engineers from dispersed sites and an opportunity to address a lot of engineering in one place (I think about a third of all Cadence engineers attended). Professor Christensen was one of the keynote speakers at one meeting and he was also a fascinating speaker.

One thing he discussed a bit was the end of Moore’s law. He predicted that Moore’s law would end because it would deliver more capability than the mainstream required at a price that was higher than the mainstream wanted to pay. This was already happening in the PC marketplace where for some time microprocessors have been “fast enough” for almost all applications (whereas through most of the 1980s and early 1990s people would upgrade their PC regularly simply because the old ones lacked oomp).

I think it is clear now that the mainstream PC market in its own turn is going to be disrupted from below by iPhone like devices. iPhones will get more powerful until most of what a PC is used for can be done on an iPhone (or a Google Android-based phone or a Nokia one; I’m just using iPhone as shorthand). Of course they don’t have big screens or keyboards but if my office and home had them, then my powerful future iPhone would simply work from my pocket when I was nearby. Or maybe it will project onto my retina or sense the muscles in my fingers or something. Who knows?

For many systems, FPGAs are disrupting ASIC from below in traditional innovator’s dilemma style. Nobody does an ASIC unless they absolutely have to, which either means an enormous amount of integration, enormous volumes, or low-power requirements (which is the Achille’s heel of FPGAs). If you can use an FPGA then you will.

Moore’s law has been driven for decades by semiconductor economics. It was always 30% or more cheaper to use the new process generation than stick with the old one. But it is not clear whether 28nm (and 22nm or whatever comes next) will have such a cost reduction. Maybe 22nm is going to be the mainframe of semiconductor processing, very expensive and delivering more capability than the mainstream market can take advantage of. The mainstream will hold back in older processes and use clever software to get what they want; after all, most chips these days are just substrates for running the software of the system.

Posted in book review | Comments Off