Officials with IBM, Nvidia and Mellanox Technologies are looking to accelerate the adoption of the OpenPower architecture in the high-performance computing space by establishing a second design center in Europe.
The three vendors, which helped launch the OpenPower Foundation in 2013 with the goal of extending the reach of IBM's Power chip architecture into such areas as scale-out computing environments, on July 2 announced the establishment of the Power Acceleration and Design Center in Montpellier, France. It comes eight months after the foundation created a similar operation at the Julich Supercomputing Center in Germany. The new facility in France will be led by software experts from Nvidia and Mellanox as well as the IBM Research and Development Lab in Montpellier and IBM Research Zurich, in Switzerland.
The OpenPower Foundation was launched when IBM officials announced they were opening up the Power architecture to enable third-party software and hardware developers to create products on top of it. Officials with the group have said they want the Power architecture to become the alternative chip architecture to x86, which through Intel and Advanced Micro Devices has come to dominate the data center infrastructure market.
"Our launch of this new Center reinforces IBM's commitment to open-source collaboration and is a next step in expanding the software and solution ecosystem around OpenPower," Dave Turek, vice president of HPC (high-performance computing) market engagement for IBM, said in a statement.
Officials with the foundation—which now has more than 130 member worldwide—want commercial and open-source software developers to leverage the expertise in the new center in France to develop HPC applications that can take advantage of GPU accelerators running in OpenPower-compatible systems. The center will offer access to IBM's Power processors, Nvidia's Tesla GPU accelerators and NVLink high-speed GPU interconnect technology, and Mellanox's InfiniBand connectivity.
HPC organizations are increasingly using accelerators like GPUs from Nvidia and AMD and Intel's many-core Xeon Phi co-processors in their systems to help improve performance while keeping power consumption down.
The OpenPower Foundation has been aggressive in pursuing the HPC and supercomputing spaces. The new center comes a month after IBM launched SuperVessel, an OpenPower-based cloud service that developers and university researchers can use for free to build applications for big data and cloud environments.
In addition, the U.S. Department of Energy in November 2014 awarded $325 million to IBM, Nvidia and Mellanox to build two supercomputers that will be five to seven times as powerful as the current systems. The two supercomputers, which will be fully operational in 2017, are part of the larger push in the industry toward exascale computing, which some industry observers expect will happen within the next six to eight years.
The systems, which will be based on the OpenPower architecture and run on IBM's Power9 chip and Nvidia GPU accelerators, will have peak performances of more than 100 petaflops.
The U.K.'s Science and Technology Facilities Council also is using OpenPower-based systems for big data research.
The OpenPower push comes at a time when organizations are looking for alternatives to Intel and the x86 architecture in the data center. Patrick Moorhead, president and principal analyst with Moor Insights and Strategy, has told eWEEK that the demand for a second source of chips is not only driven by the need for competition but also for protection against such issues as supply chain problems.
The foundation in March at the first OpenPower Summit unveiled more than 10 hardware solutions—from servers and system boards to cards and customized chips—that are based on the Power8 architecture.
OpenPower is not the only group looking to gain inroads against Intel. ARM and its chip-making partners want to take advantage of ARM's low-power designs to grab server business.