Post Script: The GPU Technology Conference
Issue: April 1, 2013

Post Script: The GPU Technology Conference

I attended the GPU Technology Conference (GTC) in San Jose in March, which promotes awareness of the value of GPU computing. The conference/exhibition had a heavy focus on research and education, with hundreds of panels, papers and tutorials, but also featured a media and entertainment path that applied to post.

The show ( is presented in large part by Nvidia, and the company’s CEO/co-founder Jen-Hsun Huang was the opening keynote, reflecting on advancements in Nvidia technology and how it’s helping pros in the VFX community create more realistic animated oceans, via its Wave Works realtime ocean simulator, as well as more believable faces for animated characters, thanks to its Face Works technology.

Huang also used GTC as an opportunity to introduce GRID VCA, the company’s first end-to-end system, which he described as the “world’s first visual computing appliance.”

The 4RU solution features two high-performance Xeon processors and eight Grid GPUs, each with two Kepler GPUs, to create a remote workspace that can concurrently run up to eight applications and serve eight users. Base price is $24,900. A 16-GPU/16-user version ($39,900) was also in beta testing. And for the cost of a $4,800/year software license, studios can give their team of artists remote access to all of their tools, regardless of the computer they access them from. 

Huang showed how artists could access PC apps like 3DS Max or SolidWorks from a Mac laptop, all the while experiencing GPU-accelerated performance. When not maxed out by the eight or 16 simultaneous users, the GPUs could be combined to considerably increase performance for lower numbers of users. In time, Nvidia says it will also be able to split GPU performance to accommodate more than the current “1 user to 1 GPU” configuration.

Huang says computer graphics are still in their early days, and that new ground is being broken all the time. GPU performance has been doubling every two years, but he says with Nvidia’s current and upcoming technology, those rates could see 100x in the next five years.