This year's Phil Kaufman award recipient is Wally Rhines. I already knew the big picture history that Wally did a PhD at Stanford, then worked half his career for Texas Instruments before joining Mentor as CEO. I called Wally up to get some more color (and some old photos from the TI era). Wally did his thesis titled Precipitation in Zinc-doped Gallium Arsenide . This was actually a key aspect of making red LEDs since the dopant for the P-type is zinc. His advisor was Craig Barrett, who would go on to be CEO of Intel, but was at that time the youngest professor at Stanford. Craig's work was actually on materials, not specifically electronics, and Wally's thesis was the first one in the electronics field. Wally joined TI straight out of Stanford. His first area of work was backside-illuminated thin imagers. They had 1-megapixel CCD imagers thinned to 25um in 1973. However, they were a bit ahead of their time and there wasn't really a market. Today, every smartphone contains backside-illuminated image sensors. After that he moved to DRAM design. He worked on the 4K and 16K DRAMs. The numbers seem trivially small compared to DRAMs today but that was the state of the art back then. Next, he became engineering manager for consumer products. They created about 30 chips per year for calculators and the Speak & Spell toy. These were all PMOS. In this era, MOS didn't get any respect since the received wisdom in the company was that bipolar and linear made all the money and MOS spent it. In 1978, Wally took over the microprocessor group. He says it was because nobody else would touch it. Intel had just announced the 8086, Motorola the 68000, Zilog the Z8000. TI had their own 16-bit microprocessor, but it only had a 16-bit address space (compared to 20 or 24 for the other designs). So there was no way they could compete with these other products for the general-purpose microprocessor market. Instead, they looked ahead and designed four specialized microprocessors, the 320, 340, 360, and 380. The 320 was a digital signal processor (DSP). The 340 was a graphics processor. The 360 was a mass storage processor. And the 380 was for the IBM token ring network. The 360 and 380 were failures, the 340 was a moderate success. But the 320 was huge. TI got into DSP products in a huge way. At the peak, DSP accounted for 50% of TI's revenue (including the wireless baseband chips which were heavily DSP-based). Of course, at the time it was not at all obvious that DSP would become so large; it is only looking back 35 years later that these things seem to be obvious. As a result, many people involved with DSP rose to senior positions within TI, including Wally himself. Wally was promoted to be in charge of all of TI's semiconductor business. He knew that if he stayed at TI then he had a reasonable shot at being the next CEO but figured that the then-CEO, Jerry Junkin, was not going to retire for another 10 to 15 years (in fact he died a couple of years later). Mentor had been calling him for a while since they were struggling with version 8 (remember, it got nicknamed throughout the industry as "version late"). Most companies don't hire a CEO who has not been a CEO before, they bring in a COO and let him rise. But Mentor wanted a CEO. So in October 1993, Wally joined Mentor and moved to Oregon. At the time, they had a few areas where Mentor had leadership products. Quicksim was their gate-level simulator and was #1 in the market. They were #1 in frameworks. They were strong in PCB (which didn't transition to version 8 and so didn't have the revenue falloff that the other products had). One key decision made back then was that, in January 1994, just a few months after Wally arrived, they kicked off development of a product that would become Calibre. Mentor OEMed TI's physical verification Checkmate, but there was a need for a new generation that supported hierarchy. The discussed a joint development but TI was not interested so Mentor did it on its own. Greg Hinckley, the COO/CFO of Mentor, once told me that Calibre was a product that many people tried to cancel year after year. Somehow it survived as some sort of skunkworks and avoided being killed. Of course, like the 320 DSP, in retrospect it is obvious that this was a good decision. Calibre and IC test are about 35% of Mentor's revenue. There was a VHDL simulator under development at Mentor called 1076 (that is the number of the IEEE standard for VHDL). But it wasn't very good. So the next step was to acquire Model Technology. That was another good move and the Model Technology technology has grown into Questa today. Mentor realized that they could not have a complete flow, instead they kept plugging holes in their product line, focusing on discontinuities in technology that created opportunities: hierarchy in DRC for Calibre, test compression for TestKompress, resolution enhancement technology (RET) for another aspect of Calibre, automotive where they could grow the business without needing to steal share from other EDA companies. Automotive, for example, is now around 20% of Mentor's revenue. Gradually they filled in the holes and they now do have a complete flow with simulation, synthesis, place and route, physical verification, RET decoration, test. Plus high-level synthesis, embedded software, emulation, PCB, and probably some other things I've forgotten. In 1993, when Wally joined, Mentor did $325M in revenue. Last year (which finished at the end of January 2015) it did $1.25B.![]()
![]()
↧