NVIDIA GPU Technology Workshop South East Asia 2014
Last month, NVIDIA held the GPU Technology Workshop South East Asia (GTW SEA) 2014 at the Suntec Singapore Convention & Exhibition Center.
The one-day GPU conference kicked off with keynote presentation by Marc Hamilton, NVIDIA’s Vice President of Solution Architecture & Engineering and John Taylor, CSIRO’s Director of eResearch & Computational and Simulation Sciences, followed by technology sessions and exhibits that showcased products and solutions from NVIDIA and its partners.
During his keynote entitled ‘GPU Technology: Past, Present, and Future’, Hamilton reflected on a decade in GPU technology, namely in the area of CUDA-enabled GPUs, followed by the Pascal architecture, Jetson TK1 development kit, as well as the GRID technology portfolio, which served as a recap of the company’s GPU Tech Conference that took place in March this year.
Hamilton further elaborated on the three key features of the nextgeneration Pascal GPU family: stacked DRAM chips, NVLink and unified memory. Stacked memory integrates several layers of DRAM chips vertically to deliver significantly more bandwidth, energy efficiency and more than twice the capacity of current GDDR5 chips. NVLink, on the other hand, provides fast, efficient data sharing between GPU and CPU at five to 12 times the data rates of PCIe 3.0. Meanwhile, unified memory allows the CPU to access the GPU’s memory, ensuring simpler programming and full bandwidth from each processor.
The second keynote, ‘Breakthrough Science on GPUs’ by John Taylor, was focused on the supercomputing capability of one CSIRO GPU cluster named Bragg, a world-class computational and simulation science facility in Australia that harnesses the power of 64 Tesla S1070s to accelerate analysis of large, complex research datasets in areas, such as computational biology, chemistry and geosciences.
Apart from the keynotes, the attendees were given the opportunity to learn the latest advancements in visual computing through a number of hands-on demonstrations and technical talks, which covered topics like high-performance computing (HPC), big data, virtualized desktop infrastructure (VDI), finance, and professional virtualization by NVIDIA experts, like-minded peers and industry figures.
The Jetson TK1 development kit is armed with a Kepler GPU with 192 CUDA cores, and a 4-Plus-1 quad-core ARM Cortex A15 CPU, for a total performance of 326 GFLOPS.
NVIDIA demonstrated the GRID Test Drive at its booth, where a notebook PC was connected to the virtual desktop. We were able to manipulate Ira’s face as part of the FaceWorks interactive tech demo, and stream YouTube videos in HD. The overall experience was smooth, with only the occasional frame drops over a 4G connection.