SAN JOSE, Calif.—The OpenPower Foundation was launched 15 months ago with the promise of expanding the reach of IBM’s Power processor architecture and addressing the needs of scale-out computing environments.
More than a year later, and armed with more than 110 members, OpenPower officials are ready to unveil the first fruits of their labor. At the inaugural OpenPower Summit here March 18, OpenPower is announcing more than 10 hardware solutions—from servers and system boards to cards and a new chip customized for the Chinese market—that leverage the Power8 architecture.
The OpenPower Summit is being run in conjunction with the GPU Technology Conference 2015, which is hosted by GPU maker Nvidia.
The solutions give the group much-needed proof points as its looks to gain more traction in a market dominated by Intel and its x86 processors, and that includes growing competition from ARM and its chip-making partners.
“Here we are, a year into the [organization’s] life, and we’re going to be showing … 14 different pieces of hardware,” Brad McCredie, an IBM Fellow, vice president of IBM Power systems and president of the OpenPower Foundation, told eWEEK before the summit began. “This is not a set of wallflowers. There is a lot of innovation there. A lot of products are being built.”
Included among those pieces of hardware are Tyan’s TN71-BP012 systems, which will be the first commercially available OpenPower servers when they hit the market in the second quarter. Tyan rolled out an OpenPower reference architecture in October 2014. IBM will be among the first customers, leveraging the systems as part of a new bare-metal service offering within its SoftLayer cloud environment.
In addition, Cirrascale will show off the RM4950, the first GPU-accelerated OpenPower development platform created in collaboration with Nvidia and Tyan. It can be ordered now and will ship in volume in the second quarter, with the aim of helping the development of big data analytics, deep learning and scientific computing applications that can take advantage of systems with GPU accelerators.
IBM and Wistron are using GPUs from Nvidia and networking from Mellanox Technologies in developing the first OpenPower-based high-performance computing (HPC) servers. The systems will be part of what IBM delivers to the Lawrence Livermore and Oak Ridge national laboratories under a $325 million contract from the Department of Energy awarded in November 2014 that calls for two supercomputers that will be five to 10 times more power than the systems being used currently, according to IBM officials.
In addition, managed cloud provider Rackspace is showing off an open server design and prototype motherboard that combine concepts from OpenPower and the Open Compute Project that will run in Rackspace data centers and will target services using the OpenStack open-source cloud orchestration platform.
Other OpenPower offerings include solutions that leverage the Coherent Accelerator Processor Interface (CAPI) found in the Power architecture that lets vendors build solutions right on top of Power. These include the ConnectX-4 adapter card from Mellanox and shared virtual memory between a Stratix V field-programmable gate array (FPGA) and a Power8 CPU developed by IBM and Altera.
OpenPower Unveils Hardware Solutions Based on Power8 Architecture
OpenPower also is being embraced in China, where chip designer PowerCore is introducing CP1, the first Power chip for the Chinese market. Zoom Netcom will bring a new line of CP1-based two-socket servers—called RedPower—to market this year, while other OpenPower members, including ChuangHe, this year will roll out shared designs for China-branded OpenPower systems. Last year, the Chinese government created a public-private partnership called the China Power Technology Alliance to help leverage OpenPower with local technology for projects in the country.
OpenPower comes at a time when Web-scale businesses like Google, Facebook and Amazon—which run massive data centers filled with millions of servers—are increasingly influencing the server market by the sheer number of systems they buy, and as the shift in demand continues toward more workload-optimized servers. Intel is aggressively expanding its portfolio to meet the changing needs, from growing the number of models with each new chip release to expanding its custom chip business to increasing the capabilities of its low-power Atom systems-on-a-chip (SoCs). Earlier this month, Intel launched the Xeon D, the first SoC for the Xeon chip family and which is aimed at the hyperscale market.
In addition, ARM is looking to push its low-power 64-bit ARMv8 architecture into the server market, with the help of such chip makers as Applied Micro, Cavium, Advanced Micro Devices and Qualcomm.
IBM’s McCredie said the creation of the OpenPower Foundation will help drive the Power architecture’s presence in the competitive space. The group’s adoption of accelerators, including GPUs from Nvidia and FPGAs from Altera and Xilinx, has helped members meet demands in scale-out environments for heterogeneous infrastructures that are workload-optimized. The foundation also is making strides in networking and memory.
“An open ecosystem is a faster way to innovate, and I can see a day when an open ecosystem will be the only way to innovate fast enough,” he said.