Sign In

Communications of the ACM

ACM TechNews

Lack of Training Hinders GP­ in HPC

View as: Print Mobile App Share:
Nvidia GF100 GPU block diagram


Several factors, including the lack of training in parallel programming and support from independent software vendors, are major obstacles keeping general-purpose computation on graphics processing units (GPGPU) from being used in industries that utilize high-performance computing (HPC). For researchers, the main challenge is in the lack of training in parallel programming, says NVIDIA chief solution architect Simon See.

The use of GPGPU in HPC has gained momentum after it raised China's Tianhe-1A to the top of the Top500 supercomputer list. GPUs provide the fastest computing capability and the best results in terms of performance and cost, says Nanyang Technological University professor Bertil Schmidt.

However, Chinese Academy of Science researcher Ge Wei says GPU-based computing can be difficult to use if programmed incorrectly. Wei says some algorithms may need to be changed to take advantage of the technology.

See says research has shown how GPU-based computing can enhance enterprise databases, and he notes that researchers also are exploring whether it can be used to solve the database growth issue.

From ZDNet Asia
View Full Article


Abstracts Copyright © 2011 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account