Congratulations Chris Rowen, for He's a Jolly Good (IEEE) Fellow

So the lede is that Chris Rowen has been elected an IEEE Fellow. In a sense it is actually a group award, because he couldn't have achieved what he has without the teams that surrounded him, in most cases teams for which he assembled the key players himself. It is a big deal. There is a cap of 0.1% of the membership who can become fellows in any year.

Chris started out as a historian. When he went to Harvard he had to decide between history and physics and decided to do physics since it seemed like it would have more effect on the future. When he graduated, he wondered what to do so he decided to follow his father's footsteps going down the road from Harvard to MIT. He had also been admitted to Keble College Oxford (UK) to study the famous PPE course (philosophy, politics, and economics) that so many senior government ministers studied, including David Cameron, the present prime minister, although he didn't go (now this blog might have been very different if he had: Ten Downing Street instead of Ten Silica).

After one semester he decided to look for a summer job in engineering. In one of these strange alignments of the planets, his Mom sent him a newspaper clipping of a company looking for people that in 1977 nobody had ever heard of. It was called Intel. He worked on 4K-bit DRAM (the state of the art, and don't forget back then that Intel was a memory company). When he graduated, he looked for semiconductor openings and worked for Intel in Oregon on SRAM. Since he'd never had a class in engineering (remember, he studied physics), he went to Stanford to do a master's degree. That became a Ph.D. He fell in with John Hennessy (now the president of Stanford, as it has turned out), who was doing interesting work on RISC architectures. He joined the MIPS project, wrote the first optimizing compiler, and did his dissertation research on synthesis and place & route. He continued to work one day a week at Intel (and, until they noticed, even continued to vest his stock!). He met his wife there.

In 1984, Hennesy and two colleagues created MIPS (the company) along with Chris, who was technically a founder although without a big name at the time He worked on both software and hardware (he designed the ALU and the register file) and also did verification of the IEEE floating-point unit. He then moved to the system side, building small workstations and servers.

In 1992 Silicon Graphics acquired MIPS. In those days, workstation vendors thought they needed to own their own architecture (after all, HP, IBM, and Sun couldn't be wrong). Chris and his wife decided it was time for a change and so they moved to Neuchatel in Switzerland, where be was a sort of European CTO for MIPS.

But, unfortunately, SGI only wanted MIPS for their workstations, so they shut down its use in embedded, which eventually meant they were so far behind ARM they can never catch up. Unfortunately, SGI ran into its own problems and so MIPS followed SGI down even in the big iron category. MIPS was spun out as an independent company again (it would eventually be acquired by Imagination).

When he came back, he thought about a startup but was recruited to Synopsys to run their design re-use business. He concluded that the most interesting IP was processors. In that era, ARM was still private and small. He tried to persuade Synopsys to enter the processor business but their top 10 customers already had their own microprocessors.

So he left Synopsys with a clean sheet of paper, no history, no technology, no income, and a computer in his living room. To the left is a group photo of the founding team in the global headquarters! They had the idea for reconfigurable microprocessors from day one. Harvey Jones (Daisy, Synopsys) and Chris himself from his MIPS exit, put in the money. They recruited the core team. Of course, this was Tensilica. It only took 9 months from the point that they had 10 people to shipping a product. Language, compiler, ISA, optimized implementation.

The first customer was Silicon Spice (eventually bought by Broadcom for $2B). Also Zilog and Cisco.

He thinks the company was successful for a couple of reasons. One, they solved a hard problem, so they had a differentiated product that was hard to compete with. And secondly, it is really, really hard to establish a new architecture. Look how much money Motorola and IBM spent to try and establish PowerPC (and they even had Apple as a customer). They were (and are) in a different market from ARM and Intel since configurable processors could go places other processors couldn't go and, luckily, it turned out to be a fast-growing market. Application know-how was also important, especially in audio where they became the defacto standard due to a mixture of low power and fast time to market when audio standards were all in flux.

Eventually, Cadence acquired Tensilica, as you probably know.

The IEEE Fellowship reflects being part of two important movements, RISC and configurable processors.