The Friday Interview: Tom Sonderman, AMD


22 Apr 2005

Tom Sonderman (pictured) is the most relaxed of interviewees. He is due to speak at a European manufacturing conference at Jury’s Hotel in a couple of hours but here he is, cross town in The Westbury, sitting snugly in a sofa, holding forth on AMD’s view of the world in the typical laid-back style of the seasoned US executive.

Though not quite a household name, AMD is the world’s second largest producer of microprocessors, the chips that drive PCs, servers and a range of other electronic devices. The largest is of course Intel and the battle between the two — the David and Goliath of the semiconductor world — makes for an intriguing contest. Intriguing because although Intel still enjoys a dominant position in the market, AMD is constantly snapping at its heels — a feisty terrier to Intel’s stately wolfhound.

Innovation has been a major weapon in AMD’s armoury. It has come out with a flurry of hot new chips that have made the market sit up and take notice. Prime amongst these is the now two-year-old Opteron, which stole a march on Intel, because unlike Itanium, Intel’s own chip based on 64-bit architecture, the Opteron is 32-bit compatible allowing customers to upgrade to this faster chip without having to make major software upgrades. The firm has also been a leader in Flash technology — the memory chips found in a range of mobile phones and PDAs as well as those ubiquitous USB sticks that are making home CD burners obsolete.

Now the two chipmakers are battling to be the first to introduce a new dual-core chip. As the name suggests, these chips will have two cores, which can operate independently of each other and so perform different tasks. Dual core is been heralded as the biggest advance in PC technology for years and Sonderman is inclined to agree.

“Dual core will allow us to add more and more performance without pushing raw clock [GHz] speed. Device physics dictates that clock speed alone will not get you where you want to go because of thermal effects and all that. So the idea of putting in multiple cores allows you to get more functionality out of the same processor without necessarily having to make it run faster and faster.

“It allows you to multi-task,” he continues. “Today, if you’re writing emails, running a movie and a complex spreadsheet all at the same time, you might start to notice a reduction in speed. Dual core these things can be going on in concert and you should not notice any degradation in overall performance.”

While Intel plans to introduce dual-core chips on PCs, Sonderman says AMD will be deploying them on its high-end Opteron server “because we think that’s where the real value of dual core initially is”.

While Sonderman speaks knowledgeably about upcoming product launches, his real area of expertise is the previous step in the supply chain — manufacturing. As head of automated precision manufacturing technology, he is responsible for the design, development and implementation of the software AMD uses to run its manufacturing plants, or fabs as they are known in the semiconductor industry. Sonderman holds 40 patents in the area of scheduling and control systems and his team as a whole holds 400.

“We use automation, control and software technology to extract a higher level of return on investment out of a facility,” he explains. “It allows us to differentiate how we run our fabs, make them much more flexible. Automation technology allows us to link the whole fab together and make it a lot more intelligent. You spend US$2.5bn building a fab; for a US$50m-100m investment [in software] you can get a lot more value out of that.”

Unlike other parts of the IT industry where manufacturing is often outsourced to the lowest bidder (usually in the Far East), the semiconductor industry as represented by Intel and AMD has stuck by its strategy of largely owning and running its own plants. Sonderman explains AMD’s rationale: “Fab-less may work for certain IC [integrated circuit] products but not for microprocessors; you need your own fab. Introducing technologies at a very rapid rate, attuning these technologies to the business and extracting all the efficiencies you can from manufacturing – these are best done by running the fab yourself. With microprocessors and even Flash if you don’t have your own fab, you have less flexibility, and it’s much harder to react to the changes in the market. Our old CEO used to say ‘Real men have their own fabs’ [laughs] and that still holds true.”

That is not to say that AMD hasn’t contemplated outsourcing production. Its newest fab, Fab 36, which will be its first to use 300mm wafer technology, was originally going to be built in Singapore in partnership with Taiwanese contract manufacturer United Microelectronics (UMC). However, AMD went cold on the idea when it realised its limitations and decided to build on its existing site at Dresden, Germany instead.

“We learned our lesson by thinking we could go and build a fab cheaply with the Taiwanese,” says Sonderman ruefully. “We got very concerned about intellectual property (IP) leakage and the fact they don’t protect IP like we do. In our business – a manufacturing-centric business – you really have to control your own destiny. You can design these things but in the end if you can’t manufacture them in high volume, in the right quantities, for the right market segments, you will not be successful. And I think that’s why a lot of the Transmetas and Cyrixes – what I call third-tier microprocessor suppliers – have ultimately not succeeded.”

But rather than go it alone completely, AMD is steering something of a middle course by contracting a company called Chartered Semiconductor to act as a second source of chips to supplement its production in Dresden. “Unlike the UMC deal, we have licensed our technology to teach them to do what we do,” Sonderman explains. “That was different to UMC that felt it could do it themselves. That’s what we call surge capacity or operational flexibility…You could say we have a hybrid model: we have our own fabs but we have a second source.”

Interestingly, Sonderman does not rule out Ireland as a possible location for a future fab. He concludes that it must have the right skill base and infrastructure, otherwise Intel would not be here but he does worry about AMD having to compete for talent with its larger rival and doubts whether the skill base would be sufficient to support both players. “What always a challenge is that can’t have too many fabs in one place because you can’t find good workers to fill them,” he observes.

Looking to future trends semiconductor manufacturing, Sonderman sees nanotechnology playing an increasingly important role, but it will not all be about doing it smaller, it will also be about doing it smarter.

“For a long time the industry just focused on making things smaller. But the new thrust is adding more and more exotic materials to extend the life of semiconductors. When the industry started about 20 elements from the Periodic Table were used to make an integrated circuit; today about 60 are. In the end it’s all about moving electrons around faster and by adding different materials you can accomplish similar things without necessarily having to make them smaller.”

By Brian Skelly