Sandy Bridge Spans Generations

Intel’s processor clock has tocked, delivering a nextgeneration architecture for PCs and servers. At the recent Intel Developer’s Forum (IDF), the company unveiled its Sandy Bridge processor architecture, the next tock in its tick-tock roadmap. The new CPU is an evolutionary improvement over its predecessor, Nehalem, tweaking the branch predictor, register renaming, and instruction decoding. These changes will slightly improve performance on traditional integer applications, but we may be reaching the point where the CPU microarchitecture is so efficient, few ways remain to improve performance. The big changes in Sandy Bridge target multimedia applications such as 3D graphics, image processing, and video processing. The chip is Intel’s first to integrate the graphics processing unit (GPU) on the processor itself. This integration not only eliminates an external chip, but it improves graphics performance by more closely coupling the GPU and the CPU. Sandy Bridge introduces the Advanced Vector Extensions (AVX), which double peak floating-point throughput. AVX will accelerate many 3D-graphics and imaging applications. The new processor also adds hard-wired video encoding. Sandy Bridge will first appear in desktop and notebook processors that will be announced in early 2011 and branded as “2nd generation Intel Core” processors. It will later roll into a family of server processors. Because these chips use the same 32nm process as the current Westmere-based designs, increases in top speed will be modest. A focus on power efficiency, however, enables Sandy Bridge to achieve more performance within the same TDP limits as its predecessor. Improvements in turbo mode enable even greater performance for short time periods. For notebook computers, these improvements can significantly extend battery life by completing tasks more quickly and allowing the system to revert to a sleep state.