The 2003 IEEE Solid-State Circuits Technical Field Award will be awarded to Daniel Dobberpuhl for "pioneering the design of high-speed and low-power microprocessors."

What made Dobberpuhl such a compelling choice for the nominator, endorsers, and evaluating committee is that it is hard to think of any other person who has contributed so much to the field and yet is so modest about his contributions. This modesty has made him a great team leader and a great circuit design teacher. Dobberpuhl's most important legacy is his ability to find good circuit talent, and then help make them better. "Most engineers who have worked with Dan would leap at an opportunity to work with him again, and today most of the current processor design teams have members that were originally trained by Dan's group at Digital," wrote Anton Domic of Synopsis in his nomination.

A short list includes Dirk Meyer, vice president of microprocessor design at AMD in Austin; Jim Montanaro, who worked on some of the PowerPC designs in the Apple-IBM-Motorola alliance; Bruce Gieseke, now a Fellow at AMD; Ed McClellan at Motorola after its acquisition of C-Port; Mike Leary, who worked on SPARC designs at HAL-Fujitsu; and Liam Madden, at MIPS. Dobberpuhl's influence continued to be felt in the Alpha designs of the former Digital team in Massachusetts that became part of Compaq Computer (recently acquired by Intel Corporation). Looking at the commercially important microprocessor families, Dobberpuhl's "alumni" have directly influenced work at AMD and Intel for the X-86 architecture, the PowerPC chips designed at Motorola, SPARC designs, and the ARM and MIPS processor cores widely used today.

Bill Herrick, Director of Microprocessor Design at Compaq, endorsed the nomination and remembers his early career when he was first hired by Dobberpuhl at Digital. They worked together on the development of an NMOS standard cell library and custom disk controller chips. Coming from ten years of LSI experience working for the government and General Electric, Dobberpuhl quickly defined the cell functions and place and route methodology needed for the products under development. Herrick recalls, "We worked well together. Dan had a clear grasp of what was important and what was not. He designed an efficient back bias generator cell and the strategy for a depletion load logic family that could quickly and safely be assembled to implement our product chips."

Working on memory management of the chip set for the PDP-11/23, Dobberpuhl developed his own MOSFET model for his programmable calculator. "Circuit characterization and simulation used a slow time-share service and Spice-deck coding was tedious," Herrick commented. Dobberpuhl worked out parameters from the manufacturer's transistor data, calculated IV curve families, and used that to drive a time-delay model to size the transistor circuits. "This innovation saved Digital a lot of design time," said Herrick.

PDP 11: Fewer Transistors Made a Better Design

Dobberpuhl has designed a very large number of microprocessors. These processors were of tremendous commercial importance to the companies that produced them, and were substantial improvements to the current state of the art. The first of many processor designs which he led was the T11. This was a small PDP 11 processor that was going into a number of low-cost machines. Mark Horowitz of Stanford University puts this design in perspective. "It had become the fashion to brag about the number of transistors in a design, with the 68000 claiming to have 68K transistors (if you counted all the possible ROM cells as transistors). It was during this period that the T11 came out which used only 16K transistors and implemented a more complex machine. It was really a wake-up call that more transistors did not mean a better design." Adds Herrick, "Dan basically did the entire circuit design himself."

The next design was the MicroVAX, the first chip to get a complete VAX instruction on a single chip. According to Herrick, "Dan had developed an approach to VLSI design where he strove to introduce one significant advancement with each project. On MicroVAX he focused on the ALU speed and the virtual address translation path. These operations would be key to the chip's performance. Dan implemented the concept of a mini-translation buffer for caching addresses on MicroVAX to save cycles during this operation. His design involved techniques that pushed the technology limits of our 3-µm NMOS process. Dan also pioneered clocking methods on MicroVAX with the introduction of local drivers utilizing zero-threshold switching devices." Horowitz continues, "He used the parasitic resistance of the source drain to his advantage in his designs. He was one of the few people who could do this, since he understands the transistor physics and the computer architecture."

RISC Processors with Single Global Clock

Following the MicroVAX, Dobberpuhl started working on RISC processors: first PRISM, then Alpha. It was Dobberpuhl 's driving of the Alpha design that led to the fast clock processor, a technique that is now standard in the industry. It is important to remember that when the original Alpha processor came out, it was nearly two times faster than any other processor. Achieving this speed took both a vision that it was possible and the circuit experience to back it up. Dobberpuhl clearly provided both to this project. He realized that clocking was going to be a critical problem and created the single global clock solution for which Alpha processors became famous. This one global clock meant careful analysis of the clock skew would be needed, so he pulled in new analysis techniques that were being developed at CMU to help with the design. In the end he put together the needed pieces and produced the world's fastest processor.

At this time in his career Dobberpuhl also found time to coauthor, with L. A. Glasser, a landmark book published by Addison-Wesley, The Design and Analysis of VLSI Circuits. Stanford's Horowitz calls it "still one of the best circuit design books on the market, even if it is fifteen years old."

"After leading Alpha to its performance leadership position, Dobberpuhl realized that a major obstacle to continued advancement of microprocessors was power consumption. The Alpha 21064 dissipated a then staggering 30W. He helped form a team to find an even simpler architecture that would allow good performance at very low power levels. ARM was the architecture and StrongARM was the family of very low power microprocessors that emerged," according to Herrick. The StrongARM processor was a factor of two to three times lower in power consumption than anything else in its class and held this lowest energy/MIP rating for a number of years.

"Dan applied many of the high-performance circuit-design techniques he had developed on his previous microprocessors (e.g., dynamic logic, cascode logic, single-wire clocking and latching) with new clock control and power-management techniques to set a new standard for microprocessor performance in embedded applications," continues Herrick. The techniques that he pioneered in this design-threshold engineering, aggressive clock gating, and power modes-are standard practice today.

Dobberpuhl 's Greatest Legacy: Training and Inspiring New Circuit Designers

Focusing only on the circuitry that Dobberpuhl has developed misses his larger contribution in moving forward the art of circuit design. He always has chosen projects that really stretched his team and himself, and through this process both became much better at the art of circuit design. There was a rigor to the Digital circuit design teams in the late 80s and early 90s. They were a group of people that had worked together for a long time and had developed both a design method and a set of tools to support that methodology. While Dobberpuhl was not wild about new CAD tools, he quickly could tell when he needed a better hammer and was not afraid to go out and find it. While the design teams were not large, they consistently produced the highest speed chips. The industry finally caught on, and now you find these Digital designers leading design efforts at a number of processor companies.

"I think this is Dan's greater legacy-his ability to both train and inspire a new group of circuit designers to find the same enjoyment in building better circuits that he had when he started over thirty years ago," concluded Domic. "He is one of the reasons that many find circuit design fun, and I think others in my position would say the same thing. I know many people in the circuit design field, but if it were up to me, I would give the award to Dan Dobberpuhl. He is the best circuit designer I know."