Nvidia Partner Colfax Trains Its Next Best Tesla Customers

programming language

And early next year, Colfax will be conducting what might be called Ecosystem Building 101 at its Sunnyvale, Calif. headquarters, where the company will host four training sessions for would-be CUDA programmers in January and February. The two-day classes offer a mix of GPU computing theory and hands-on CUDA programming with instructor Chris Mason, a product manager with Calgary, Canada-based Acceleware, a developer of software for the oil and gas industry.

"This is our first time partnering with Colfax," said Mason, who manages Acceleware's linear algebra product line. "The great thing about these training sessions is there's always a point where a light bulb goes on for the students, where they really figure out what they need to do to make the GPU work for them."

For Colfax, a charter member of Santa Clara, Calif.-based Nvidia's Tesla Preferred Provider program, the benefits of hosting CUDA training are many, Shah said. The biggest benefit may be the creation of more potential buyers of GPU servers and supercomputers built by Colfax, which offers Intel Xeon 5500 series systems jam-packed with as many as eight Nvidia Tesla graphics cards.

A cut of the $1,999 per-student enrollment fee charged by Acceleware is a nice bit of icing on the cake. Also attractive is creating more CUDA buzz ahead of Nvidia's highly anticipated 2010 release of its Fermi product line, the successor to Tesla at the high end of the chip maker's product stack.

Sponsored post

"The excitement for Fermi is at fever pitch," Shah said. Nvidia demonstrated its future-generation GPU at October's GPU Technology Conference in San Jose and Fermi is expected to arrive in March or April of next year, according to sources.

Acceleware is also "very excited" about Fermi, according to Mason, who was especially pleased about advances in double-precision performance and the addition of ECC memory correction -- a first for a GPU -- on the Fermi chips.

Nvidia itself does a good deal of CUDA training outreach, largely through the growing number of universities around the world that are now teaching the programming language. About 270 universities have CUDA programs, according to the chip maker.

The chip maker has also certified eight universities in the U.S., Europe and Asia that "leverage GPUS pretty extensively throughout their various research groups" as "CUDA Centers of Excellence," according to Nvidia spokesman Andrew Humber.

James Huang of AMAX Information Technologies, another Tesla Preferred Provider, said that China in particular was a hotbed for CUDA training at the university level.

"It seems like the number of [CUDA] programs at Chinese universities has grown five to 10 times in the past year," said Huang, a product marketing manager. AMAX hasn't yet signed on to host CUDA training at its Fremont, Calif. headquarters, but the system builder does offer potential customers a fee-based Tesla test bed for their own CUDA applications, he said.

For now, Colfax is the first system integrator partner to have organized a CUDA training event, according to Nvidia's Humber. The company is "actively reaching out to our channel partners right now and encouraging them to follow suit," he said.

"With the increased momentum we are seeing every day for GPU computing, it's a great time for system integrators serving the enterprise computing, education, defense and other markets to get involved and help educate their customers on the capabilities of GPUs," he said.