The most likely challenge to Intel's hegemony in data center server processors is supposed to come from ARM licensees, but IBM is moving to supersede the ARM merchants with a deal to pair its Power processors with Xilinx FPGAs.
No one has yet been able to loosen Intel's stranglehold on the data center processor business, but IBM has been feverishly positioning Power as a legitimate option, especially at the high end. A strategic partnership with Xilinx announced today will help.
The newest data center applications need a combination of extremely powerful processors and workload acceleration, IBM argues. These applications include machine learning, network functions virtualization (NFV), genomics, high performance computing (HPC) and big data analytics.
The knock on ARM processors is that while they might be able to compete with Intel's x86 processors on the low end, the ARM community has yet to produce much that can compete with Intel's high-end Xeon processors.
IBM is positioning its Power architecture as a match for Xeon-based products. Xilinx field-programmable gate arrays (FPGAs), which can be configured by a customer or designer after manufacture, will provide the workload acceleration that IBM and Xilinx argue will benefit these newer applications.
IBM and Xilinx said their relationship will be a multi-year strategic partnership. Moving forward, IBM Systems Group developers will create solution stacks for Power-based servers, storage and middleware systems with Xilinx FPGA accelerators for data center architectures such as OpenStack, Docker and Spark, the two companies said.
IBM will also develop and qualify Xilinx accelerator boards into IBM Power Systems servers. Xilinx is developing and will release Power-based versions of its software defined SDAccel Development Environment and libraries for the OpenPower developer community.
Though the agreement announced today is specifically between IBM and Xilinx, they were careful to note that they will continue to collaborate with other members of the OpenPower Foundation, which include Altera, Mellanox, Nvidia, Google, Hitachi and ZTE, among many others.
An element that IBM and Xilinx say makes their combined approach competitive is IBM's Coherent Accelerator Processor Interface (CAPI). Server processors and server memory need to be tightly coupled. IBM designed CAPI to enable equally tight integration with other solutions right on top of the Power architecture. One example is workload accelerators, such as Xilinx's.
Meanwhile, the two companies said, independent software vendors are leveraging IBM flash storage attached to CAPI to create very large memory spaces for in memory processing of analytics, enabling the same query workloads to run with 1/24th the number of servers compared to commodity x86 solutions.
"The combination of IBM and Xilinx provides our clients not only with a new level of accelerated computing made possible by the tight integration between IBM Power processors and Xilinx FPGAs, but also gives them the ability to benefit directly from the constant stream of innovation being delivered by the rapidly expanding OpenPower ecosystem," said Ken King, IBM's GM of OpenPower, in a statement.
— Brian Santo, Senior Editor, Components, T&M, Light Reading