Reading the runes

Consider the indicators. The 2007 sales projection from the Semiconductor Industry Association is for just 1.7% growth; it started the year at 10%. The numbers for fab equipment are worse. The Information Network recently forecast that market would shrink by 2.3% this year and its president Robert Castellano told a familiar tale. “For the third time this decade, too much money was spent by chip manufacturers to meet their production. We're clearly seeing it in the memory market as ASPs fall from excess inventory,” he says. Elsewhere, though, things look rosier. The Consumer Electronics Association projects US wholesale volume reaching $155.2billion this year and believes its sector worldwide now consumes more than 50% of chips sold. CE growth is not slowing. Beyond firm US performance, there is burgeoning demand from developing markets that have added 1bn new consumers – although there is a circle to be squared. As Aart de Geus, chairman and ceo of Synopsys (pictured), notes: “Conventional wisdom and the realities of the second billion consumers say devices must get cheaper, but design complexity says devices will be more expensive. Economic reality says we have a problem.” The industry believes it can solve this – the bullishness amid the uncertainty – through innovation, more holistic design strategies and more attention to both the back end and software. The real issue is the speed with which it responds. This element has perennial components: innovation and people. For veteran analyst Gary Smith, chief analyst at Gary Smith EDA, 2007 has been marked by ‘the realisation that the software development infrastructure is broken’. “Because of the power problem we have abandoned von Neumann computing and moved to a multiprocessing architecture. Unfortunately, all our software has been developed to run on a von Neumann computer. Until we develop a concurrent software infrastructure, we cannot take advantage of multicore/multiprocessor products. General purpose computing dies after four processors.” Microprocessor developers will be first to feel the pain, but the trend towards the multiprocessor SoC, particularly for portable and high end CE devices, will widen its effect. Here, the lack so far of a common vision represents a serious threat. “The hardware and software communities suffer from a degree of myopia,” says Ian Mackintosh, president of OCP-IP. “They’re each trying to leverage and protect their legacy technologies when we need to look at things in different ways.” Mackintosh does not expect a few tools and a miracle methodology to emerge overnight, but he does believe that steps are not being taken as quickly as they should. The scale of this looming problem is breathtaking. Kazu Yamada, vice president at NEC Electronics America, says it could take a 40 engineer team 10 years to complete a 32nm design. Doing it in a more typical 90 weeks could require 250 engineers. And so, appropriately, to ‘people’. Two issues stand out. The first is simply getting the raw material. “One major threat is the state of the education system and developing more qualified engineers and scientists in the US and around the world,” says Mike Fister, president and ceo of Cadence Design Systems. “In addition to forming alliances with major universities and centres of learning, I think we have to push for government funding of scientific research and educational programs, as well as make a corporate commitment to advanced research.” The second issue, though, is then allowing young and not so young innovators to breed new concepts through start ups. “The pain of 65nm has had an impact on start ups,” says Rajeev Madhavan, chairman and ceo of Magma Design Automation. “The impression built that either design was very expensive or you had to have a spectacular architecture to be in the game. So VCs became more wary of funding new companies. We have to get past that – any slowing down in start ups is dangerous for the industry.” None of these problems is insurmountable. There is aggressive belief they can be solved. But 2008 must be a year of answers, rather than questions.