Nvidia Boosts HPC Efforts by Buying Portland Group

The Portland Group builds compilers and developer tools for HPC environments, and has worked with Nvidia on its CUDA platform.

Nvidia officials are making another move in the high-performance computing space, buying software compiler and toolmaker The Portland Group.

Nvidia announced the deal July 30, though no financial details were released.

The Oregon-based Portland Group (PGI) was founded in 1989 and offers compilers, debuggers and development tools for the increasingly competitive high-performance computing (HPC) industry, working in such languages as Fortran, C and C++. It has collaborated with Nvidia on Fortran and C/C++ for CUDA, the vendor's parallel computing platform. The company has also worked with ARM and IBM.

PGI also supports such languages as OpenMP and MPI, and works in Linux.

Nvidia has worked with PGI over the past five years in its push to have its graphics technologies used for general-purpose computing workloads, according to Ian Buck, general manager of GPU computing software for Nvidia. Though PGI will be owned by Nvidia, it will continue to work with other vendors, including Advanced Micro Devices and Intel.

"PGI and its exceptionally talented staff will continue to operate under the PGI flag—developing OpenACC, CUDA Fortran and CUDA x86 for multicore x86 and, of course, GPGPUs [general-purpose GPUs]," Buck said in a post on the Nvidia blog. "And they will continue to serve their wide-range of customers—including chip makers, research labs and HPC computing centers."

OpenACC is an effort involving, Nvidia, PGI and supercomputer maker Cray to create a standard for parallel computing. The move enables researchers, scientists and corporations to run applications in a parallel fashion on heterogeneous CPU/GPU systems, and parallel programmers can outline directives to the compiler, which will do all the work to optimize the applications for GPU-accelerated environments.

The HPC server space is a rapidly growing one due in large part to the growing demand for supercomputers, which organizations and countries around the world are leveraging for scientific and economic reasons, according to Earl Josephs, program vice president for technical computing at IDC.

"HPC technical servers, especially supercomputers, have been closely linked not only to scientific advances but also to industrial innovation and economic competitiveness," Joseph said in a statement in May. "For this reason, nations and regions across the world are increasing their investments in supercomputing, even in today's challenging economic conditions."

In June, IDC analysts said the HPC technical server sales grew 5.3 percent in the first quarter, reaching $2.5 billion. The firm expects the market to grow 6.8 percent annually, with revenues hitting $15.4 billion by 2017.

Major server vendors—from IBM and Hewlett-Packard to Dell and Fujitsu—are players in the space, and chip makers also see HPC has a growth area. Intel and AMD both have been aggressive in pushing their x86-based chips for the supercomputer market. In the latest Top 500 list of the world's fastest supercomputers released in June, Intel's Xeon chips were in four of the top 10—including the fastest, the Tianhe-2 system in China—while AMD's Opteron processor powered the second-fastest computer, the Titan system in the United States.

Nvidia has positioned its GPUs as accelerators for HPC systems that can run alongside traditional processors and drive performance while reducing power consumption, key considerations for supercomputer users. Nvidia's GPU accelerators were in two of the top 10 fastest systems on the June list.

AMD also offers its GPU technology as accelerators, while Intel is countering that with its Xeon Phi x86-based coprocessor offerings.

Nvidia also is working on its Project Denver to create a computing chip that leverages its GPU technology and ARM's CPU designs.