Since 1986 - Covering the Fastest Computers in the World and the People Who Run Them

Language Flags
May 13, 2011

Roadblocks Ahead for GPGPU Adoption

Nicole Hemsoth

In many ways, this has been the year of GPU computing with the number one supercomputer in the world being powered by the new technology. However, according to the top purveyor of GPGPU, NVIDIA, there are still some major limitations for widespread adoption.

Chief Solution Architect at NVIDIA Simon See weighed in on the multiple challenges that are preventing greater GPGPU infiltration in industries that make use of high performance computing. According to See, the barrier lies in a lack of parallel programming talent as well as more general lack of support from ISVs.

See’s company has actively been trying to solve some of the parallel programming training challenges with the massive investments it has made in the last couple of years in CUDA Centers or Excellence, CUDA Research Centers and CUDA teaching Centers. However, if there is no clear value for ISVs to hire and implement GPGPU technology, there are still a lack of jobs to be found in this area, and thus the predicted explosion in GPU computing is destined to be hindered.

See told ZDnet that “for researchers, the main challenge lies in the lack of training in parallel programming. While it’s possible for researchers to write their own computational codes for HPCs, most are not able to ‘think in parallel programming’ which prevents them from applying the technology.”

A researcher from China who has worked within GPU-based systems told ZDnet that indeed, GPGPUs are beneficial for many areas of scientific computing but can very difficult to use if not been programmed properly. This was based on his experience using the technology to boost process engineering research as well as on advising other researchers to explore GPGPU possibilities by changing their algorithms to try it out.

He notes that another end of this issue is rooted in the fact that many HPC users in the oil and gas industry, for example, are still reliant on outside code for their operations. What this means is that the companies that might be able to benefit from advances in GPGPU technology can only make that leap if their ISV supports the technology.

Full story at ZDNet