Report: GPU Technology Conference Workshop 2011, Singapore
Nvidia and Novatte jointly organised the GPU Technology Conference (GTC) Workshop 2011 on 12 May at Biopolis, Singapore. The GTC workshop brings together researchers, developers and GPU computing experts for knowledge sharing and exchange of ideas. VR-Zone reports.
The GPU Technology Conference was last held in September 2010 in the United States. The GTC Workshop 2011 brings this platform to Asia for the first time. Jointly organised by Nvidia and Novatte, a high-performance computing solutions provider, the GTC Workshop brings together researchers, developers and experts in the field of GPU computing. The GTC Workshop is sponsored by Nvidia and Singapore's Agency for Science, Technology and Research.
During the GTC Workshop, researchers from various institutes of higher learning, including the Chinese Academy of Science, Rostov State Building University, Nanyang Technological University and National University of Singapore, presented on how GPU computing has benefitted their research, as well as methods for optimizing GPU computing.
Novatte CEO, Yury Drozdov, also spoke on Novatte's complete GPU computing solutions. Aimed at research institutes and commercial users, Novatte's solutions provide not only scalable hardware clusters but also qualfied support packages and professional application porting services.
He also demonstrated Novatte's admin-friendly GPU cluster, which provides an easy to use graphical management system for administration of hardware clusters.
Novatte brought with them a working demo system.
Simon See, Chief Solution Architect, Nvidia, gave the keynote introduction to the conference. He presented on the various improvements to CUDA since its introduction.
CUDA version 4.0 introduces several new features. GPUDirect 2.0 allows multiple GPUs to share data directly without having to go through system memory. Unified Virtual Addressing provides a contiguous memory address space across the CPU and multiple GPUs. Another key improvement is Thrust, which makes parallel programming in C++ easier and faster.
Participants were treated to a buffet lunch, after which a basic CUDA tutorial was conducted for researchers and developers.
The next GPU Technology Conference will be held from 11 to 14 October 2011 in San Jose, California.