Reykjavik (Iceland) - Considering the big news coming out of Intel this week (Larrabee) and the expected big News from Nvidia within the next two weeks (x86 CUDA), AMD is under pressure to match its rivals: AMD is making substantial changes to its GPGPU software strategy and announced at its GPG CTO Technology Day that it will ditch its Close-To-Metal platform in switch to OpenCL.
In his speech GPG CTO Technology Day held in Iceland’s capital, Raja Koduri, CTO of AMD GPG (ex-ATI), announced that AMD believes that the time for proprietary software solutions such as AMD’s own Close-to-Metal and Nvidia’s CUDA has passed.
As a result, AMD will throw its efforts behind DirectX 11 Computational Shaders and the OpenCL GPGPU language and will focus on standardized solutions only. Koduri highlighted the GPGPU advances made by companies such as CyberLink, PeakStream (which was acquired by Google), RapidMind, RogueWave, CAPS, ImageScan, Telenetics, Neurda and many others. It is apparent that many companies are bringing GPGPU-accelerated products to market, but AMD’s is going a somewhat different way as the company’s stream products will be aligned with DirectX 11 and OpenCL.
Koduri noted that a first product showcasing this strategy will be available in the first quarter of next year. Also, AMD is working on APU (Accelerated Processing Unit) at full speed, which is scheduled for debut in first half of next year.
The decision to go with OpenCL could be a critical step for AMD to compete with Nvidia and Intel’s GPGPU and cGPU products that are capturing the headlines today. AMD’s low-level programming approach was one of the main reasons why developers preferred Nvidia’s (high-level) CUDA version over the company’s stream processor cards. OpenCL is widely considered to be a possible solution of GPGPU programming that could bridge Nvidia, Intel, AMD and other products and we are hearing more and more developers requesting support for OpenCL.
