Mentor chairman Wally Rhines crusades for emulation

Now the company wants more. Rhines, who is also Mentor's chief executive officer, is positioning the company to take advantage of what he sees as an upturn in the emulation business starting this year. "[Capital equipment] budgets have been restricted so long that people are finding they don't have any more capacity around," he said. "In 2001 if you had five projects and three go t canceled, you freed up an emulator, but that's not happening anymore."

Rhines said that as mainstream IC design starts reach millions of gates, simulation and simulation farms can't verify designs fast enough. "If you look at the verification gap, you have to conclude that there is such an enormous chasm between what is available in general-purpose computing farms and what has to be done in verification. Something has to give," said Rhines. "I contend it will have to be emulation and that you will need a whole range: high-end multiuser systems and individual systems."

From the 1980s, when he led Texas Instruments Inc.'s design group for consumer products, Rhines has advocated emulation. "We built emulators for every chip we did, and we built about 30 chips a year," he said. "We were able, in an era when design automation couldn't do things, to get remarkable results."

Emulators are PC-to-refrigerator-size boxes containing a field of commercial or custom FPGAs, into which engineers p rogram their IC designs. The idea is that a design programmed into the emulator will run exponentially faster than it would in simulation or in a simulation server farm. Designers link the emulator to the rest of a system design to verify that it works with the rest of the system.

Replicates — basically cheaper, read-only versions of the emulator — give software engineers an early jump on developing code for the design under verification.

When he took Mentor's reins in the early '90s, Rhines said, he saw a need for commercial emulation systems and pursued it, acquiring French emulation vendor Meta Systems in 1995.

Exceptionally tough

But Rhines quickly found commercial emulation an exceptionally tough business. Indeed, the acquired company was seen as a huge threat to then market leader Quickturn, which sued Mentor on the grounds that Meta's SimExpress emulator violated a Quickturn patent. In winning the suit years later, Quickturn got U.S. courts to bar Mentor from selling SimExpress in the United States. The company continues to sell Meta emulation products abroad.

But Rhines persisted, even attempting a hostile takeover of Quickturn in the late '90s. That company went to Cadence Design Systems Inc. instead.

So when it got the chance, Mentor snapped up Ikos. "In Ikos, we gained the manufacturing and sales we needed in the U.S., but better technology [than Quickturn's]," Rhines said.

In outbidding Synopsys Inc. for Ikos, Rhines said, Mentor has gained on (but not bested) Quickturn in U.S. market share and topped it in worldwide hardware-assisted verification market share. He said he derived his figures by subtracting Mentor's and Ikos' combined hardware-assisted verification revenues from the numbers given in the 2001 EDAC Market Statistic Service report to claim the lead.

"The reality of the [emulation] market has changed," said Gary Smith, chief EDA analyst at Gartner Dataquest. In 2000, the overall acceleration and emulat ion market was $236 million and some were predicting it would reach $633.2 million in 2005, he said. But as the economy nose-dived, emulation followed. Smith said he must now recalculate those numbers.

"This year we are going to see 12 percent growth in this area and by 2003 we are going to see demand in the 19 to 21 percent range," he said. "I would expect growth to be 30 to 35 percent after that. Emulation has typically been used by verification teams, but we are going to see growth in design team emulation. As we merge hardware and software verification teams together, these boxes will become a must."

Bad reputation

To capitalize on this market, Mentor (Wilsonville, Ore.) and other vendors must first overcome the bad reputation emulation has gained over the years. Early commercial emulators were hard to program and buggy. Some broke, causing design groups to miss verification deadlines after spending millions on a given emulator system. In the late '90s Ikos' own VStati on 2M had problems with the FPGAs popping off the printed-circuit board. The company fixed the emulator and went on to score big wins in the graphics chip market. Most notably, it had a hand in verifying Nvidia Corp.'s GeForce3 chip set.

Another problem: emulator price tags are exponentially larger than those of simulation tools. Mentor's high-end Meta Celaro Super-C box, for example, sells for $6 million to $10 million. Simulators, by contrast, typically run from $4,000 to $25,000.

The cost is well worth it, Rhines said, because emulators run so much faster than simulators. Rhines pointed to an Infineon Technologies benchmark that shows simulators running a design at five cycles per second and emulation running at 650,000 cycles/s.

Rhines said the cost of an emulator is decreasing in terms of dollars per gate, noting too that emulators are expensive to design and manufacture and are not a high-volume item. Companies in good years will sell about 100 units of a product. In a year lik e 2001, vendors were lucky to sell half a dozen.

Tough 2001

Last year was tough, as customers slashed capital equipment budgets, where emulation systems often fall. "The 2001 weakness was not caused by people stopping their use of emulation," said Rhines. "[It] was caused by a combination of freezing capital budgets and projects that were killed or delayed."

Customers still need to design, Rhines said, and are running out of emulation capacity, so business is picking up again. "People are going to break the diet they've been on and someone is going to say, 'If you want this project out and shipping on time, we've got to get more emulation capacity,' " he said.