Archive for June, 2009

Lots Of Work To Be Done

Thursday, June 25th, 2009

Many design engineers have had a long time to think about what’s changed in the design world. Some of them are still looking for jobs, frustrated by the remarkably slow pickup in the job market—particularly in the United States and Europe.

But as Magma CEO Rajeev Madhavan noted last year, coming out of the downturn things may not look the same as when we went into the downturn. It’s more than just the availability of jobs. It’s what the new jobs will be.

The convergence of all things digital and analog onto a single piece of silicon should give some hints about the shape of things to come and where the focus will be. Motion sensors, an explosion in MEMS, the combination of voice, video and data and more will change the notion of what constitutes an SoC. The previous distinction was Small A, Big D, or Big A, Small D. We are entering the phase of Big A, Big D, with the interconnections between the two becoming a critical area for development.

With enough real estate to pack hundreds of millions of transistors onto a single piece of silicon—along with vertical stacking of some die—there’s plenty of room for function and forms. What there isn’t, at least so far, is any structure for how to build that chip, how to integrate the various pieces and how to debug it.

That will provide plenty of work for years to come—not to mention lots of opportunity for new sales—for tools companies. It also will provide momentum for startups that can bridge some of these tools.

Most economists predict the financial world will return to health over the next six months, with some growth in 2010 and steady growth in 2011. It’s time for this industry to figure out what needs to get done and get everyone back to work.

Ed Sperling

Where The Jobs Are

Friday, June 19th, 2009

The job market for design and verification engineers seems to be exploding. In the past week, listings have been flooding onto jobs boards for LinkedIn semiconductor design groups. The only trouble is engineers may have to move to get the jobs—sometimes halfway around the globe.

There have been a bunch of job postings for semiconductor expertise in India, the United Kingdom, as well as places like South Korea (home to Samsung and LG), Switzerland (home to STMicro), and a smattering of offerings in Texas and parts of California outside of Silicon Valley. That either means companies are looking to cut costs by hiring workers in less expensive areas, or it means there are a bunch of new companies that have been financed coming out of the downturn to challenge the giants of tech.

My guess is both. That doesn’t mean Silicon Valley will disappear, of course. It’s still the core of innovation in tech, and it likely will remain so for many years to come. But it’s also an expensive place to live and do business, which is why there is a growing demand for expertise in other places.

Over the next six months, as the global economy begins its climb out of a giant economic crater excavated by ruthless bankers, financial traders and insurance companies, demand for jobs will pick up again everywhere. But what’s changed this time is that they’re being back-filled in places with the lowest overhead first, rather than stepping up operations in more expensive regions and then figuring out a way to offshore those jobs later.

The infrastructure for offshoring is already in place. Companies like Intel and IBM already have a major presence in India and China. And in a cyclical industry like semiconductors, you have to wonder what’s going to happen during the next downturn.

–Ed Sperling

Fear Of Failure

Friday, June 5th, 2009

For the past five decades, failure in the semiconductor world was nothing to be ashamed of. Mistakes were a sign of experience, and even some of the most spectacular errors in judgment didn’t cost design teams their jobs or deter more experimentation.

 

As we move down into the realm of 45nm and beyond, that’s no longer the case. Mistakes can cause an entire project to be shut down forever, and the team that made the mistake to be tossed out and replaced. Finding qualified replacements at the moment is not particularly challenging. But depending upon the size of the company and the investment in a project, mistakes also can inflict mortal wounds on the company itself.

 

At the leading edge of design, a single mistake and missed market window can cost in excess of $100 million. It can dry up a lucrative and longstanding customer relationship in a market where the potential customers can be counted on one hand. And it can eat into the profitability and market perception of even the largest semiconductor companies.

 

This is a fundamental exercise in risk management where all the variables are not known, and it could have some interesting ramifications for chip development in the future. At the very least, we are entering a stage where companies need to be able to gauge effectively when to pull the plug on development projects and move on. That will require much better communication up and down the design flow than has ever existed in the past, bringing together more parts of the flow at once.

 

If the software developers can’t write the software, then the hardware engineers need to know that. If there are changes in the hardware that could affect the software, that has to be communicated early, as well. And if the verification team is running into massive problems dealing with multiple power islands or stacked die models, then the project managers need to know that in a realistic time frame.

 

Perhaps even scarier than failure, however, is the fear of failure and what that will mean to innovation. Hollywood is a classic example of this, which is why we’re about to get a remake of The Taking of Pelham 1-2-3 and why we’re seeing yet another Terminator movie being released. With movies costing more than $100 million to make, no one wants to make a mistake.

 

In the chip world, this is a new phenomenon, and the big concern is that we’re going to ratchet down innovation and the rush to new nodes in an effort to play it safe–or simply hang back a couple nodes until the process is worked out. Ultimately, however, the cost may prove much higher than the cost of failure. Rather than losing a design team or even a company, it could mean the loss of an entire market’s potential, which is incalculable.

 

The good news is the risk is spread across multiple companies. The bad news is some of those companies won’t be around to gloat because the differences between one and another won’t be significant enough to warrant more than a couple competitors.

 

–Ed Sperling