The stars of the first-ever OpenPower Foundation Summit this week were 15 pieces of hardware, lined up onstage as proof points of the speedy innovation happening within the 15-month-old consortium.
The hardware—including servers, developer platforms, adapter cards and silicon—represented the broad range of companies that make up the 110-member group, which includes IBM, Nvidia, Google, Rackspace and Mellanox Technologies. Member vendors that may not be household names include Cirrascale, Inspur, Tyan and Nallatech.
OpenPower officials said the gear they were showing is a clear indication that OpenPower will become the primary alternative to dominant player Intel in the data center, and that the contest will be as much about business models as it is about products.
In addition, OpenPower intends to become a significant technology presence in the booming Chinese market, where the government is urging the country’s tech industry to leverage products from Chinese companies. Several of the pieces of hardware on display at the summit were from Chinese companies like ChuangHe and Inspur (servers), Zoom Netcom (server board), and Suzhou PowerCore, which worked with IBM to develop the CP1, a Power-based chip for the Chinese market.
The officials also spoke about the 31-member China Power Technology Alliance (CPTA), which is using the Power processor architecture to build tech products in China. An official with Suzhou PowerCore told attendees at the summit, which was held in San Jose, Calif., that the “CPTA’s goals are to bring Power technology to China, develop a processor technology that is locally ‘controllable,’ and transform the server business in China into a 100% local stack, as an alternative to Intel,” Roger Kay, principal analyst with Endpoint Technologies Associates, wrote in a column for Forbes.
It’s the push to become that Intel alternative that is the driving force behind OpenPower, which launched in December 2013 after IBM decided to open up its Power architecture as a way to extend its reach and address performance and power efficiency needs in hyper-scale and Web-scale data center environments. Meeting those demands is about speed and efficiency around innovation, and that is done best through open development and building a community, rather than keeping all design, development and manufacturing within a single company, according to Brad McCredie, IBM Fellow, vice president of IBM Power systems and president of the OpenPower Foundation. The technology on display at the event showed how well the open development model can work, McCredie said during the summit, which ran in conjunction with the GPU Technology Conference 2015 hosted by Nvidia.
“Here are 15 pieces of hardware that were put together in less than a year,” he said. “It’s an incredible amount of productivity in a short amount of time.”
Intel will have an increasingly difficult time keeping up with the speed of innovation that an organization like OpenPower will have, McCredie said. He expects organizations will understand this message.
“You no longer have to rely on x86 and one vendor,” he said. “Today, the limits of the x86 architecture are profound. The sole reliance on a tick-tock approach to processor development isn’t working anymore.”
The rise of what McCredie calls “warehouse-scale computing”—where companies like Google and Facebook operate massive data centers running a huge number of servers—is opening up opportunities for new architectures and business models to challenge Intel’s dominance. ARM and its chip-making partners are looking to push its low-power architecture—which is found in most smartphones and tablets—into the data center. Imagination Technologies is trying to do the same thing via the prpl Foundation to drive its MIPS architecture.
OpenPower Wants to Be Intel Alternative Here and in China
OpenPower may have an impact on them as well, according to analysts. There won’t be much room for more than one alternative to Intel and x86, and OpenPower has the backing of IBM and a more proven data center architecture in Power, Paul Teich, CTO and senior analyst with Moor Insights and Strategy, told eWEEK. In addition, longtime Intel rival Advanced Micro Devices—which intends to complement its x86 server chips with ones based on ARM’s architecture—hasn’t been able to keep up with Intel’s pace of innovation.
From “the size of the companies competing, their enterprise credibility, their business velocity, and their ability to sustain R&D investment for the hyperscale market and show that they will remain a credible market contender for the foreseeable future, IBM is clearly different from AMD and also from the first wave of ARMv8 (64-bit) server SoC [system-on-a-chip] licensees entering the market,” Teich said in an email to eWEEK.
OpenPower’s technology base “is clearly also a step above AMD and the ARMv8 server SoC licensees,” he wrote. “They are leveraging their HPC market ecosystem base, which is quite different from ARM’s embedded market base.”
Teich said ARM and its chip partners have to show how they fit into the server market and how their less powerful processor cores will help organizations with TCO “with respect to ‘brawny’ Intel and IBM processor cores. … I would guess that Intel and IBM will keep casting ARMv8 SoCs as ‘wimpy’ cores.”
Those sentiments were echoed by Charles King, principal analyst with Pund-IT.
“OpenPOWER may be an even greater danger to those hoping to leverage the ARM chip architecture in server and data center solutions [than to Intel],” King wrote in a research note. “After all, OpenPOWER offers many of the same flexible benefits that ARM proponents claim for themselves but the Power architecture has well-established RAS (reliability, availability and serviceability) features and a leadership position in business data centers that ARM can only dream of.”
In addition, Power has the “continuing support of thousands of business software developers and ISVs and their private and public sector customers. This is not to say that OpenPOWER has punctured prospects for ARM servers or deflated them entirely but we expect that keeping ARM’s prospects aloft will become that much harder,” he wrote.
Intel is making its own moves to develop silicon for the high-performance, energy-efficient systems that are increasingly in demand, from rapid development within its low-power Atom family to its new Xeon D—the first Xeon processor in an SoC model—to innovation around its Xeon Phi co-processors. The company is ramping its custom chip business and offer more versions of new chips. Whether OpenPower can detail any of those efforts remains to be seen.
“Competitively speaking, the obvious target for many or most OpenPOWER efforts is Intel,” King wrote. “Not only is the company the leader (usually by a long shot) in most business computing target markets but OpenPOWER’s development model (which allows Foundation members freedom to leverage the Power architecture as they wish and to engage chip fabs of their choosing) is markedly different than Intel’s tight control of its intellectual property and manufacturing processes.”
The push into the China market also will put OpenPower into competition with Intel in that country, where Intel has invested billions in companies and buildings and ramped up partnerships with other vendors, according to Endpoint’s Kay. In addition, because of its size, Intel has been able to drive down silicon prices for server OEMs.
However, it retains tight control of its technology, and given concerns of China lawmakers over whether the United States is using products from U.S. tech companies to spy on China (mirroring U.S. worries about Chinese companies like Huawei and ZTE). Intel’s “black box” approach means “you do not get to know exactly what’s baked into those microprocessors. And that is what the Chinese object to,” Kay wrote. “One can argue the reasons for China’s increasing insistence on an ownership participation in the technology that it uses, but the fact of it is undeniable.”
“In the OpenPower vs. Intel case, as of this moment, I’m convinced that, while financial motives play a role, the security principle is the main driver [in China],” he wrote. “Up through now, IBM has not been selling the least expensive technology. Its business model is based on supplying differentiated, critical capabilities that generate healthy margins. Intel has been the firm driving down the raw cost of computing. So, for the Chinese, OpenPower appears to be less about getting a better deal than about controlling the future of computing in China.”