Accelerated Computing Era Begins
Performance of consumer applications has been determined by performance of microprocessors for decades. But with the arrival of accelerated processing units (APUs) the situation may change drastically as at least some of the programs will begin to use massively-parallel graphics processing units and thus will deliver better customer experience thanks to new levels of performance.
The development of APUs began back in 2006, when Advanced Micro Devices acquired ATI Technologies. Even though initially it was considered that integration of a graphics engine onto a microprocessor would hardly be an issue, the process took a lot longer. AMD did not only have to ensure that the hardware actually functions, but had to develop proper OpenCL 1.0/DirectX 11-compatible graphics cores, appropriate x86 cores as well as wait till the industry actually adopts the standards required for efficient APUs. As a result, the outcome of the project will only be visible in 2011.
Both AMD and Nvidia has worked hard with numerous software developers to ensure that their programs could take advantage of ATI Radeon and Nvidia GeForce graphics processors' capabilities and many applications do use the GPUs now. The software that can utilize conventional graphics chips is naturally capable of taking advantage of APUs, provided that the apps use industry standards like OpenCL or DirectCompute, not proprietary things like CUDA. But a big problem is that consumers have to find programs that actually utilize graphics chips for general-purpose computing (GPGPU).
At present Nvidia offers a set of links on its web-site that help to find GPGPU-enabled applications for various tasks. AMD will also offer a similar kind of service in the very short-term future. However, it is obvious that it is not enough for general consumers.
AMD is currently working on a number of APU-oriented software advertising and delivery mechanisms. In addition, at least one computer maker, which is set to introduce AMD Brazos-based computers at the CES, will also offer an application store that will specifically mark software that take advantage of APUs.
Availability of programs that utilize graphics processing engines for general-purpose computing will play a critical role in popularity of AMD's APUs. Both Ontario/Zacate and Llano accelerated processing units will be slower than Intel's Sandy Bridge offerings in terms of x86 performance, but will offer tangible advantages in GPGPU-enabled applications.
But do not expect the majority of programs to be APU-aware in 2011. The most important thing next year is that the second largest designer of x86 chips will give a clear signal about its commitment to GPGPU-accelerated future, the era of accelerated computing.