AMD has unveiled its lineup of keynote speakers who will address developers at its upcoming AMD Fusion Developer Summit -- a list that includes ARM executive John Davies.
AMD on Tuesday said that Davies, vice president of technology in ARM's Media Processing division, will discuss ARM’s history of heterogeneous computing, its market strategy and, most of all, its support for AMD’s OpenCL and other open industry standards. The summit -- which will take place at the Meydenbauer Center in Bellevue, Wash. from June 13-16 -- will also include about 90 technology sessions on topics ranging from developer tools to high-performance computing, and, most notably, multimedia and professional graphics processing. However, according to AMD, the forum is dedicated to “furthering heterogeneous computing.”
In an interview with CRN on Monday, John Taylor, Director of AMD’s Client Product and Software Marketing division, said the heterogeneity of Open CL lies in the fact that it’s an open industry standard managed by Khronos Group as an industry spec.
"The spec allows software developers to run the same code, which is written for Open CL, on multiple hardware platforms, and support video and graphics capability that’s currently very popular among consumers,” Taylor said. “It allows parallel processing across a multicore CPU or a massively multicore GPU -- AMD makes both, and only AMD makes them as one piece of silicon: the APU.”
Taylor said AMD’s Accelerated Processing Units (APUs), which integrate CPU and GPU processing on a single die, underscore the value of graphics processing to developers.
“What’s often lost in the awkwardly-framed debate between X86 and the ARM platform is the importance of the GPU on either,” Taylor said. “To have software developers talk about how OpenCL works for them parlays into AMD’s advantage with Llano, which offers 400 to 500 gigaflops of compute power in a device that’s very thin, compared to 100 gigaflops of compute on a regular Intel or AMD CPU.”
Third-party developers write code for OpenCL that in turn provides a great deal of flexibility for AMD’s partners in the reseller channel, according to Taylor, as they look to avoid proprietary standards in order to leverage technology from different vendors and platforms. “APUs are arriving en masse to the channel, and partners will have access to our A-Series Llano chips soon,” Taylor said. “Open CL ties together a very efficient silicon platform with user-optimized software that allows VARs to create customized solutions with less complexity, requiring less energy and less cooling.”
Taylor said AMD’s OpenCL strategy in keeping with AMD’s approach to technological innovation, in contrast with rival graphics specialist Nvidia and its proprietary CUDA parallel processing architecture, also known as Fermi. “We bet on open standards like Open CL and then we engage a software developer community and bring multiple voices to bear,” he said. “CUDA is proprietary and is essentially about GPU computing.”
Next: More On CUDA Versus OpenCL,/strong>In addition, Open CL has the advantage of leveraging both multi-core CPUs and highly-parallel GPUs on sleek form factor devices whereas CUDA, according to Taylor, requires plugging-in a discrete GPU on the motherboard with separate wattage, and separate cooling.
As for ARM, which has earned a dominant share of the mobile market by licensing its low-power architectures to semiconductor manufacturers, the company’s core competency does not overlap, nor does it conflict with, use cases for AMD’s new low-power Fusion processors. “Our focus is not on the handset market with these APUs,” Taylor said. “Our focus is on desktops, laptops, tablet devices, set-top boxes, theater, and digital signage. That’s the market for this generation of APUs.”
In addition, AMD spun-off its former manufacturing division, Globalfoundries, in 2009 and finalized its separation from the chip manufacturer earlier this year, making it unlikely that ARM and AMD would partner in order for the former to design processors built by the latter, as ARM and Nvidia intend to do with their Project Denver server processors. However, AMD’s Fusion APU platform will eventually come to the mobile handheld market, according to Taylor.
“Mobile is still a growing market opportunity for AMD,” Taylor said. “We offer a differentiated design that’s very attractive to a software developer.”
AMD said the Fusion Developer Summit will also include a keynote from Herb Sutter, Microsoft principal architect of Native Languages, entitled “Heterogeneous Parallelism at Microsoft,” as well as presentations from AMD corporate fellow Phil Rogers and AMD Corporate Vice President and CTO, Graphics Eric Demiers on APUs and graphics processing, respectively.