Top Guns: How a Team of Elite NVIDIA Engineers Helps the Best Developers Get Even Better
To the average gamer, “Navier Stokes” might sound like the suave, Bond-like main character from an espionage-oriented first-person shooter. Someone who’s as comfortable in a tuxedo as in a flak jacket.
They’d be half right. Navier-Stokes does indeed make an appearance in a growing number of games. Not as a character, though, but as an algorithm. Navier and Stokes were 19th century French and Irish mathematicians, respectively, who were among the first to develop a mathematical model for turbulence – say, smoke rising from a lit cigarette or the wake thrown off by a racing speedboat.
The algorithm whose name they share remains at the cutting edge of advanced mathematics. So much so that a proof of its efficacy is one of the famous “Clay Millennial Prizes” in mathematics and worth $1 million to whoever solves it.
At NVIDIA, Ashu Rege, VP of Game Content and Technology, tells the story of Navier-Stokes not because he’s a math geek (though with a Ph.D. in computer science from UC Berkeley, he could certainly play that role). He does so to point out the benefits that can accrue to game developers when they work with NVIDIA engineers.
The Mumbai native is an engineer’s engineer, who grew up programming simple games on a friend’s Sinclair ZX81 in the 1980s. His team was the first to develop an implementation of Navier-Stokes that was fast enough to work in real time – on an NVIDIA GPU. They called their graphics routine “Turbulence.”
“With smoke or with fluids in the real world, you get all these interesting eddies and formations,” he said. “If there is a smoke cloud and you walk through it, it should part. But it should part in certain ways, depending on the speed you’re walking. For a long time, in most games, smoke was quite fake. Navier-Stokes could simulate it, but it would take forever. With Turbulence, we did it in real time.”
Every company in the graphics business has in-house experts bristling with advanced math skills. What makes NVIDIA unique, says Rege, is that the company routinely sends its math heavyweights, who speak the language of Navier-Stokes, into the field to work with developers. They’re not just fixing bugs in device drivers. They’re sitting side by side with game developers helping inject NVIDIA tech and high-level math deep into the titles.
Rege is chief scientist, in a sense, for a group bringing cutting-edge science to the artists building the next generation of interactive experiences, and bringing some of that artistic sensibility back to NVIDIA.
“It’s very common at NVIDIA for someone to invent something, and then say, ‘I’m not happy just inventing it. I want to see it in a game,’” he said. “So instead of just sitting in an ivory tower here at NVIDIA, they’ll go and work with developers.”
“I’m talking about some of our best math Ph.D.s, who end up partnering with the very smartest people at our developers. Developers really appreciate it because they end up brainstorming with us about entirely new directions to take their graphics.”
Rege has been at NVIDIA for nine years. His main responsibility is to run the company’s developer support and outreach operation. In that post, he oversees about 100 people, most of whom also have math and computer graphics backgrounds from some of the world’s best universities.
At any given time, a significant percentage of Rege’s engineers are out in the field, spending time with developers. That was the case when Navier-Stokes first came out. It occurred again with the first version of “Crysis,” when half a dozen NVIDIA engineers camped out for months in Crytek’s Frankfurt office, helping get the game ready to ship.
Another triumph by his team involved ratcheting up the realism of the fur of Riley, “Call of Duty’s” bomb-sniffing German shepherd.
“Hair in general, and fur in particular, is highly complex,” said Rege. “To properly simulate fur, you have to model each and every strand, with each strand assigned certain properties. When the dog jumps into a pond, part of it is wet, its fur needs to look matted. Modeling that and making it look accurate is incredibly difficult.”
Turning Mathematics Into Money
NVIDIA also helped Ubisoft with the lighting on “Splinter Cell: Blacklist,” to make darkened rooftops and rain-soaked walls look real. And Snail Games got help making the clothes on the characters of “Age of Wushu” ripple and fold. Because Snail Games makes costumes available as an in-game purchase, the extra realism was worth hundreds of thousands of dollars.
Rege said that while his engineers love pushing the boundaries of graphics in a game, they also enjoy less glamorous challenges. For example, a game with just average graphics could benefit from a thorough review by a sharp programmer.
“One of our engineers might be able to tell the developers, ‘How about if you did your order of drawing in this way, so you don’t use up the battery quite as quickly?’ There is always something you can do to make a game run better.”
It’s not just game companies that Rege’s team supports. NVIDIA’s GPUs are increasingly common in the sorts of scientific and engineering operations found in the supercomputer world. So, one of his engineers who might be spending time with a new adventure game one month might be helping out at NASA’s Ames Research Center the next.