NVIDIA Collaborates with Microsoft on High-performance GPU Computing

By DE Editors

NVIDIA has announced work with Microsoft to promote NVIDIA Tesla graphics processing units (GPUs) for high-performance parallel computing using the Windows HPC Server 2008 operating system.

“The coupling of GPUs and CPUs illustrates the enormous power and opportunity of multicore co-processing,” said Dan Reed, corporate vice president of Extreme Computing at Microsoft. “NVIDIA’s work with Microsoft and the Windows HPC Server platform is helping enable scientists and researchers in many fields achieve supercomputer performance on diverse applications.”

NVIDIA Research developed several GPU-enabled applications on the Windows HPC Server 2008 platform, such as a ray tracing application that can be used for advanced photo-realistic modeling of automobiles. Related to this, NVIDIA worked with Microsoft Research to install a large Tesla GPU computing cluster and is studying applications that are optimized for the GPU.

In addition, a range of enterprise applications—such as data mining, machine learning and business intelligence, as well as scientific applications like molecular dynamics, financial computing, and seismic processing—are taking advantage of the massively parallel CUDA architecture on which NVIDIA’s GPUs are based. The CUDA architecture enables developers to use the CPU and the GPU together in a co-processing model. Compute-intensive sections of an application use the parallel computing capabilities of the GPU, while the sequential part of an application’s code runs on the CPU.

NVIDIA Tesla high-performance GPU computing products support Windows XP and Windows Vista in the workstation and Windows Server 2003 and Windows Server 2008 in the data center.

In related news, NVIDIA has released the first public OpenCL conformant GPU drivers for Windows and Linux.

  In addition to the drivers themselves, NVIDIA has released a performance profiling tool and an OpenCL Best Practices Guide. This public release is fully conformant with the OpenCL v1.0 specification and supports the OpenCL Images features of the specification that, while optional for other vendors, provides performance benefits across many image processing disciplines such as medical imaging, video transcoding applications, machine vision, and facial detection, according to the company.

Key features include:

• Profiling of actual hardware signals, kernel efficiency, and instruction issue rate
  
• Timing of memory copies between system memory and GPU dedicated memory
  
• Customizable graphs to help developers focus in on problem areas
  
• Basic auto-analysis to reveal warp serialization problems
  
• Easy import/export to CSV for custom analysis
  
• Support for all CUDA enabled GPUs
 

For more information, visit NVIDIA and the company’s OpenCL developer site.