Intel, IBM Squabble Over 'Open Chip' Claims

Intel recently blasted IBM’s efforts to develop an open chip architecture alternative to its own leading x86 server processors, calling Big Blue’s attempt a costly strategy that runs the risk of turning the clock back to the 1970s when it comes to datacenter chip development.

Leading up to the first meeting this week of the group formed to support open chip architecture, Intel's criticism was aimed at IBM’s OpenPower Foundation, which since it was created in 2013 has amassed 113 members, including industry heavyweights such as Google, Nvidia and Rackspace.

On Wednesday, the group holds its first annual OpenPower Summit in San Jose, Calif., where members are expected to debut a swath of new OpenPower hardware and discuss platform momentum.

[Related: Rackspace Puts Intel Aside, Makes IBM OpenPower Pledge]

id
unit-1659132512259
type
Sponsored post

"Every company needs to understand what their options are. But we don’t want to revert to the mainframe days of 1970s and '80s when software and hardware were tied to specific CPUs," said Shannon Poulin, vice president of the Data Center Group and general manager of the Datacenter Marketing Group at Intel.

The OpenPower Foundation’s mission is to build an open database server architecture around IBM’s Power8 microprocessor, allowing foundation members to design custom versions of the chip for use in task-specific applications such as big data and cloud computing.

For IBM’s part, it says because itsOpenPowerprocessors are open and customizable at the core level, its chips perform better for many task-specific workloads compared to Intel’s datacenter Xeon chips. IBM asserts that Power systems offer better performance and equal-to or better-than cost per watt.

’The era of one company trying to meet all the needs of their clients by themselves is coming to an end,’ said IBM's Brad McCredie, IBM fellow and president of the OpenPower Foundation, in an interview Friday. ’The only way to keep pace is open innovation. We are focusing on making that a reality around the Power architecture.’

For the past year, the OpenPower Foundation has focused development of its chip architecture around both the scale-out datacenter market and high-performance computing solutions for customers. McCredie argues that Intel’s ’general purpose processors’ don’t offer the same type of customization, performance gains and cost savings.

’These are markets where developers like to overclock, customize and tune there hardware for their software,’ McCredie said. As for partners, they say opening up the Power chip to the development community has given it new life.

’For years people were saying Power was too expensive and too closed,’ said Michael Gray, COO at Champion Solutions Group, an IBM partner based in Boca Raton, Fla. ’Power is still a small part of our business, but with Google and Nvidia investing in this technology, that gives my customers a lot of confidence in the Power platform.’

The crux of the dispute between Santa Clara, Calif.-based Intel and Armonk, N.Y.-based IBM involves which chipmaker is more ’open’ than the other and allows third parties to customize software and hardware solutions.

In 2013, IBM announced it would license its struggling Power chip core architecture in what many viewed as a Hail Mary pass to save its Power chip from extinction. The licensing agreement included the formation of the OpenPower Foundation, which sought industry support for the open chip technology. Members have access to the Power chip’s core intellectual property, allowing it to be customized for specific workloads.

Intel, on the other hand, said it prefers to lock down a CPU’s core. It argues that doing so increases a chip’s value when it comes to OS and app compatibility across the x86 ecosystem. Poulin argues Intel does support microprocessor customization at every level -- just not the chip’s core. Where IBM touts its chip’s core as open, Intel considers its platform open -- allowing anyone to build off of it.

’Our competitors seem to be using Intel’s strong market position to mean that x86 is more ’proprietary’ in nature, and defining ’open architecture’ as only those with licensable IP/cores,’ Poulin said. ’We believe the definition of an ’open architecture’ goes well beyond just licensable IP. Openness must include the flexibility and interoperability of the entire platform stack -- from chips to systems to operating systems to applications.’

Intel has addressed criticism it has faced in the past in which the chip giant was accused of a one-size-fits-all approach to processors. In September, Intel released 32 SKUs of its Grantley-based E5-2600 v3 Xeon processors. In addition Intel said it has deployed 20 custom Grantley SKUs designed for customers and their specific workload needs.

’If you ask developers how many choices do I have with this ’open’ technology verses that, I think the world would agree there is no more open architecture than the Intel-based x86 platform,’ Poulin said.

Intel asserts that the lure of OpenPower platform is diminished by the fact that there isn’t yet broad adoption of OpenPower industry standards, which limits the development community to just a tiny fraction of IT professionals.

’We want to make sure that when you think about x86 hardware and software development that developers are not dealing with 40 different types of cores that have been modified needing 40 different versions of software,’ Poulin said. ’Companies need to look at the cost of these customized solutions.’

Nevertheless, OpenPower Foundation members say they are impressed with IBM’s Power8 performance and say within specific deployments, they are seeing significantly greater performance than Intel's Xeon chips. For example, Google and Rackspace say the OpenPower architecture is attractive because they can have input into how hardware is developed, so as to better tailor it to run their massive scale-out datacenters.

For IBM, OpenPower architecture gives it a way to enter this scale-out market dominated by Dell, HP and now Lenovo. It also gives Big Blue an opportunity to secure the mid- to high-range $7 billion RISC server market, which it dominates over Hewlett-Packard's Integrity servers and Oracle's Sparc servers, according to research firm IDC.

’For a long time, everyone considered Power too expensive and a niche solution,’ Champion Solutions Group's Gray said. ’Since IBM opened up the platform, now customers are interested in custom solution options, performance and security. We are seeing growth in our Power business.’

It’s unclear what, if any, market share gains or losses IBM has seen with its Power processor. But over the past 12 months, IBM has announced several key customer wins. In November, IBM unveiled a $325 million deal with the U.S. Department of Energy to sell it next-gen supercomputers based on its Power8 CPUs. In December, Rackspace said it was committed to data center development around IBM's Power8 processor. In April, Google showed off a prototype server based on the Power processor.

’It comes down to definitions,’ Poulin said. An open Intel platform is more valuable than an open core chip architecture, he said. ’It might take OpenPower Foundation members years to do with Power, what they can do tomorrow with Intel,’ he said.

For its part, IBM says, it says it won’t take years, and when you factor in performance, customization and price per watt, the advantage goes to Power.

PUBLISHED MARCH 16, 2015