You probably know that Harry Potter's birthday is July 31st. As it happens, that was the same day that Tensilica was founded. Furthermore, both started 20 years ago. The first Harry Potter book was actually published first, on June 26th 1997. On July 31st a little over a month later, on July 31st 1997 Chris Rowen incorporated Tensilica. That was 20 years ago today. Chris was on vacation in Hawaii but was gracious enough to get on the phone with me last week to tell me the story. MIPS, SGI and Synopsys Chris had been working in microprocessors and systems for a long time. His PhD was actually more in EDA, an early version of synthesis, place & route. He was at MIPS from the very beginning, but that was the era when every workstation company thought they needed their own processor (Sun had SPARC, HP had Prism and so on) so Silicon Graphics acquired them. After working there for a time Chris was recruited to Synopsys as the GM of their nascent Design Reuse Group, what today we would call an IP business. In looking at strategy, Chris came to the conclusion that most of the value would be processor-centric. He tried to convince Synopsys management to enter the processor business, but failed. Aart said, "every one of our top customers thinks they are in the processor business so we are not going to compete with them." So after less than a year at Synopsys, Chris left. He threw away all his notes and started with a clean sheet of paper. Founding The incorporation of Tensilica was complete on July 31st 1997, as I said in the opening to this piece. At that stage Chris had a clear idea of the mission of the company, to pursue processors that would be adaptable to different applications, consisting of both hardware and software. Here is the initial business plan. In that era, this was ambitious. The received wisdom then was that there were two ways to go. You could design your own processor, in which case you struggled with compilers and other tools that would need to be ported or created. Alternatively, you could go with a standard processor and use the same architecture for every application, whether or not it was a good fit. It was just Chris on his own at the beginning. Some people thought it was an interesting idea, one of whom was Harvey Jones. One complication was that Harvey was chairman of the board of Synopsys at the time, but there was no notion that this was competitive in any way since Synopsys didn't do processors and their IP was foundation libraries, PCI, USB and a few other interfaces. The landscape had a few other processor IP companies. ARM existed but was still low profile. ARC existed, recently spun out of Argonaut in England, with some very simple notions of extensibility. There was research going on in academia, especially in European universities. But the idea of an IP company with an extensible microprocessor was a wide-open field. Moore's Law was in its "happy scaling" phase when you could get a faster processor at the same power just by waiting, so you could get a lot of performance out of a standard instruction set without worrying about special processors for special tasks. However, Chris realized that power would become a big issue, and you couldn't get low-power microprocessors for SoCs just out of Moore's Law. The big early breakthrough was the proof of concept that you really could generate hardware (the processor) and software (the compiler) together. Seed Money A few weeks after incorporation, Chris went to Japan to talk to contacts he had over there and get calibration. The good news was they didn't think he was crazy. This also helped to refine, not just the technology roadmap, but also the pitch: who would use this and for what. About three months after Chris got started, Bernie Rosenthal, who was a key business guy in Synopsys's Design and Reuse (IP) group left the company. He was the first true employee of Tensilica. Discussions with Harvey Jones continued. He was well connected in financial circles and made lots of introductions to venture capitalists. However, in the end, Harvey and Chris decided they would provide the seed funding themselves. However, they also took in small amount of money from a list of gold-plated luminaries in the valley: Chris's dad, John Hennessy, Kurt Keutzer, Richard Newton, and Andy Bechtolsheim. In total that was $2.2M. With the seed money they could rent an office, next to the Duke of Edinburgh pub in Cupertino. In fact when they had all hands meetings, they would hold them in the pub. Some other key people signed on in this period. Monica Lamb was halfway through her sabattical from Stanford, and she agreed to take the second half of her sabbatical at Tensilica. She was very well plugged-in to the compiler community. Two key compiler people joined, John Ruttenberg and Woody Lichtensteijn, based in Waltham Mass. Other early employees were Ashish Dixit and Gülbin Ezer (who is still at Cadence), Earl Killian, and Albert Wong. First Investment Round The first year was very creative with an excellent team and clear goals of what they wanted to achieve. Each time they made a milestone they would open a bottle of Champagne. The more advanced the milestone, the better quality the Champagne. After 6 months, Chris said the Champagne had got "really good." They also spent that time refining the message and filing patents. In that first year, all the pieces that are still there today came together: the processor architecture, the compiler, the TIE (Tensilica Instruction Extension) language. In June 1998, roughly a year after founding and 6 months after funding, they raised their first round of VC money. It was actually the B round, since they'd called the seed round the A round. They raised $10.6M from Oak Investment Partners, Worldview Technology Partners, and Foundation Capital. By their first birthday, they had a team in place, the beginnings of the technology, a clear message. And 10 million in the bank. They were on their way. Early Years The above diagram shows the history of the first 4 years of Tensilica. Of course, the story of Tensiilca as a company ended up with the 2013 acquisition of Tensilica by Cadence. But Tensiilica as a product family lives on, of course. For information about the current Tensilica family, which is rather richer than it was at the end of this slide 17 years ago, start at the Tensilica Product Page . Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
![]()
↧