3D Graphics: John Brothers
Brutal world needs savage plan

Graphics is an ugly business, and few companies could tell you that better than S3 Inc. The company's fall from grace has become a textbook example of the pitfalls in the PC-graphics market, as a lack of cutting-edge 3-D capability dropped the company from the top slot in graphics.

The bad news is that the graphics market shifts continually as new features and algorithms get developed. But that's also the good news, as it gives S3 a shot at climbing back into the game with Savage3D, a high-end 3-D part that began shipping around Labor Day.

The man who led the Savage3D charge was John Brothers, S3 vice president and the leader of the architectural team. It was two years ago that Brothers, a drivers expert and veteran of Chips & Technologies Inc., started the work that ultimately would mark S3's attempt to rejoin the market.

"In the business that we're in, it's very cost-constrained and really schedule-driven," Brothers said. "If you miss schedule dates, you might as well have not done the product at all, because you'll be six months behind, and by then it'll be too late. It's very different from going for a small, high-end market."

And as updated versions of Savage3D were already being previewed behind closed doors at Comdex, Brothers and his team had already begun a two-year project that will become S3's next graphics architecture.

An electrical-engineering graduate, Brothers cut his teeth on graphics with Chips & Technologies, where he worked on what he felt were dead-end projects for little recognition. "My manager at the time said, try to find ways to make graphics faster. And I did, but nobody would listen. I got very frustrated in that group," he said.

Brothers eventually left Chips for consulting work, "got bored" with that, and hired on at S3. Here, he's found a less-contentious environment. "It's still at the size where anybody can talk to the president. It's not very hierarchical."

The problems that the company has had have stemmed not from a lack of talent, but from difficulty managing product development, Brothers said. With PC technology in constant flux, and graphics algorithms and features continually being invented, the architectural team has a tough job deciding which features will be important two years from now.

"You want that silver bullet to differentiate yourself," Brothers said. "The prize that everybody's after is, they want to do what everybody else does, but they also want that one thing nobody else has. They want to scoop the competition."

The process starts with brainstorming meetings facilitated by Brothers. "I'm the organizer type," Brothers said. "I'm trying to run the meeting, take notes, keep one person talking at a time. It's kind of hard refereeing. Actually, they work together really well, but it's pretty chaotic in those meetings."

And while engineers' hearts always gravitate to the spectacular technologies, S3's charter forces the team to lean toward the practical and the simple. "It forces you to find more elegant ways to do things," Brothers said.

Over time, the list of features gets pruned. Those that don't appear to be working out are scrapped, sometimes turning out to be problems the competition can't solve either. Others don't die off until deadlines begin to loom, when scheduling realities have to win out over cool technologies.

Underlying all these decisions is the fact that S3 intends to stay in the mainstream market rather than chase the high end as 3Dfx Interactive Inc. did. "Engineers want to go for the bleeding edge," he said. "There's kind of this tug toward the high end, but you've always got to keep the gate count in mind," he said.

Even for all its achievements in the end, Savage3D could get passed over because it doesn't play to the right audience. Benchmarking for 3-D graphics revolves around the Winbench rating, a single number that OEMs crave as much as megahertz performance from a microprocessor, but Savage3D's prized features include two that don't appear in the benchmark at all. One improves clarity of up-close images to a crisp precision, eliminating the fuzziness or blockiness of walls in games like Quake. The other is a means of using less memory space for textured images.

Neither feature contributes to Winbench scores, and what's more frustrating to Brothers, both have come under fire from the competition. It's a common practice, Brothers said-discrediting a feature that your own company can't do.

Uncertain future?
"It's not something you can predict two years ahead of time," Brothers said. "Looking at future things, I don't want our fate to be in the hands of something so unpredictable."

"I'm trying to pull back and work more on architectural stuff, which I'm better at than managing a large group of people," he said. "I would really like to focus on the new product. I think what I can do well is get things started, get people rolling on the right things." To that end, he hired a software expert in September who's about ready to take over that part of the development.

And despite coming off a tough year, Brothers still thinks that his company deserves a shot at redemption.

Return to: TimesPeople 98