Nvidia Working with Microsoft on GPU Computing for HPC
Nvidia is working with Microsoft to expand the reach of its graphics processing technologies into high-performance computing space.
Nvidia officials announced Sept. 28 that their engineers are collaborating with their Microsoft counterparts to tout the use of their Tesla GPUs in HPC environments that use Microsoft's Windows HPC Server 2008 OS.
The move is part of Nvidia's aggressive campaign to bring its GPUs into more mainstream, general-purpose computing environments as co-processors with traditional CPUs from Intel and Advanced Micro Devices.
AMD, with its ATI graphics business, also is pushing its graphics technologies into common computing environments, and Intel is beginning to ramp up the graphics capabilities in its products.
Nvidia's CUDA architecture gives developers a way of using both CPUs and GPUs at the same time in a co-processing fashion. The result has been rapid growth in the number of applications in a wide range of disciplines-such as business intelligence, data mining and scientific software-that can take advantage of the CUDA-based Tesla GPUs, according to Nvidia.
"The coupling of GPUs and CPUs illustrates the enormous power and opportunity of multicore co-processing," Dan Reed, corporate vice president of extreme computing at Microsoft, said in a statement.
Andy Keane, general manager of Nvidia's Tesla business, said researchers in HPC environments are looking at such technologies as Microsoft operating systems and GPUs as ways to improve their work.
"The scientific community was one of the first to realize the potential of the GPU to transform its work," Keane said in a statement.
Scientists are seeing speed improvements in their compute-intensive applications of 20 to 200 times, he said. In addition, researchers also are using Windows on workstations and in data centers more, thanks to such development tools as Microsoft Visual Studio and the operating system's ease of use and lower costs, Keane said.
Nvidia's Tesla GPUs support Windows XP and Vista in workstations and Windows Server 2003 and 2008 in the data center.
In addition to touting the work they've done with Microsoft in the data center, Nvidia officials also announced that it has released GPU drivers for Windows and Linux that conform to the OpenCL specification.
The company introduced the beta drivers in April. The newest drivers support the OpenCL v1.0 specification, including the OpenCL Images feature.
Like CUDA, OpenCL enables GPUs to be programmed like CPUs. The spec has wide support from such companies as AMD, Intel Apple and IBM, as well as Nvidia.
One of the features of the new Nvidia drivers is the support for all CUDA-enabled GPUs.