Related Articles

Nvidia threw a big party in California last week, and invited many people to come. Okay, so perhaps it wasn't so much a party as it was a GPU Technology Conference. And they weren't your everyday, random people -- instead, some of the brightest folks from universities, research facilities, and tech companies from 150 different countries around the world. The event had a festive environment, and a strong focus on the cutting edge. In fact, just standing around, I thought a new scientific breakthrough could potentially break out at any moment.

I came in late-night last Tuesday to the Fairmont hotel-hybrid-conference center in downtown San Jose. After passing Nvidia's founder and CEO, Jen-Hsun Haung's lovely Ferrair outside the large glass doors of the hotel, this greeted me in the lobby:

A massive, massive paint-ball cannon. The one that guy from MythBusters used to create a Mona Lisa last year at NVISION '08. So I knew I had the right place.

Opening Keynote

To kick off a conference, it is proper form to have a big opening keynote speech to get things rolling.

The keynote was held in a giant ballroom -- not sure if the picture does it justice below -- and it went without saying that Jen-Hsun would be the main speaker, as the living embodiment of the Nvidia corporation. Being a member of the press I was able to bypass the keynote's long, long, long line that snaked through the halls of the Fairmont. Enjoying the privilege of my early entrance, I setup base in front of a giant screen a bit to the left of the main stage.

To get things warmed up, a M.C of sorts took stage, and officially welcomed everyone to the event. After the welcoming, he went on to compare recent developments in GPU technology, and parallel processing utilizing Nvidia's CUDA to the excitement of the space-race of his youth. Continuing the metaphor, the CEO-M.C expounded on how GPU technology was at a point of reaching escape velocity, and was ready to 'take off', so to speak. 

On each seat was a pair of 3D-enabling glasses. We were asked to put these on. We did; and the lights dimmed, and a video montage rolled.

The video showed the progression of GPU technology from 1995 to the present day. It was incredible to see the chunky, spot-the-polygon graphics slowly transform over the course of a decade to the near perfect ray-traced images that we can now create today. "The eye candy you can get, with a billion dollar's investment, is my favorite part of the job," joked Jen-Hsun.

The video then finished up with some very impressive 3D effects -- and for myself, sitting in the front row watching the display, I was quite struck. I've seen some 3D demos  -- but nothing on this scale. It was right then -- and not the later focus on 3D tech -- at that very moment where I became certain that the days of 3D have arrived. It's going mainstream. Within a decade, most of the big blockbuster movies will be enjoyed in 3D, many televisions will be in 3D, and the hardcore gamers will be playing only in 3D. I'm willing to put money on that now. One day in our life times 2D monitors will seem as quaint and as archaic as black and white television seems now.

"The last 16 years have felt like warp speed to me," said Jen-Hsun, after the last frame of video was projected. He recapped the development of the GPU into three major stages: the first was focused on horsepower -- getting the ability to run all those shaders; the second big phase change was the exploratins of the programming nature of the GPU -- the first programmable shader language, and the ability to alter each pixel. "This expanded the reach of the GPU," he said. You could see that Jen-Hsun was building up to declare the third big stage as being  upon us now, with the introduction of CUDA, which is Nvidia's opening up of the access to the parallel processing abilities of the GPU.

Regarding horsepower, Jen-Hsun pointed out that a GTX 275 that you can buy for $200 has more raw computational power than the a CM5 array -- an early '80's supercompuer that cost a briefcase-load of gold bars, weighed 18 tons, and was large as two homes.

If this NSA Cm5 (called FROSTBURG) fell on you, you'd be dead -- and it couldn't handle Left 4 Dead either

It was unveiled later by Dr. Jeff Nicols, the associate lab director for the Oak Ridge National Laboratory (originally founded to help with the Manhattan Project), that the next generation of supercomputer that will be a hybrid, CPU-GPU CUDA-enabled design. The system will provide somewhere on the level of 25 PetaFLOPS of computational power. Proof that it is becoming increasingly worthwhile to tap into the power of the GPU -- this was perhaps the foremost motto of the conference.

But what do with all this parellel processing potential locked up in the GPU?

CUDA of course, is the key proffered by Nvida to unlock this power.