Practical Chip Design

There has been much angst about design outsourcing in recent years, the majority of which has come from US-based designers who have lost, or who fear for, their jobs. But as the industry gets more experience with the practice, there are other problems emerging as well, that impact not just designers but the outcome of designs, and quite possibly the competitiveness of the companies that outsourced the work in the first place. This is due to a natural evolution in the progress of outsourcing.

All of this came from an overheard conversation at Hot Chips this afternoon. A senior architect for a major US company was talking about the problems he has faced with outsourcing. He was describing how as the design teams on the other side of the Pacific get more sophisticated, the partition between the US and the—usually—Asian team has begun to shift.

In the good old days, when we were only sending routine bench-level jobs off shore, the partition was at a functional block, or even a task level. The bulk of the design team remained rooted in the US, with senior US people in control. Only some clearly-defined blocks, or specific tasks such as logic verification, went to the outsource team, and the process was tightly controlled.

But as the outsource teams gain more experience and more senior people, that practice is shifting. Today, often the architecture team is all that's left in the US, and the entire implementation team is on the other side of the Pacific. This reflects the rapid growth in sophistication and management skill of Asian design teams. But it creates a problem, according to my unintentional informant.

The problem is simple. Architects divorced from actual implementation tend to drift into Never-Never Land. Without drawing unflattering parallels to Frank Lloyd Wright here, these architects tend to create idealisms that are unworkable in the application, or are simply unimplementable. Conversely, implementation management separated from architects tends to lack vital information about the intent of the design—stuff that is very hard to capture in a specification but that would strongly influence how the design was implemented.

The result, according to the victim of my eavesdropping, is an increasing risk that designs would come back working perfectly, except not actually doing what the architects had in mind. Like that famous series of cartoons about the tire on a limb from many years ago, the design falls into a morass of miscommunication. Architects without recent implementation experience, and without the design team leaders sitting across the table from them, create a cloud palace. The design team, not party to the original discussions about the design requirements, by enormous effort comes as close as they can get: a palace on a mountain top. Marketing wanted a jumbo jet.