AMD:CPU+GPU United Marcouni fully oriented to HPC

In the current High-performance computing systems and applications, the vast majority of the x86 architecture of the processor and applications, with High-performance computing system performance continues to improve, multi-core and nuclear http://www.aliyun.com/zixun/aggregation/13440.html " > Heterogeneous systems Show excellent performance when satisfying certain special applications of High-performance computing. For example, for applications that pursue floating-point performance, GPU (graphics processing chips) are much faster than traditional CPUs (CPU), which is why AMD and Nvidia believe that GPU or CPU+GPU will eventually replace CPUs in future high-performance computing.

As a result, AMD put forward the concept of cohesion (fusion), the CPU and the GPU combined, and in the recent Taipei computer exhibition officially exhibited, the second generation Fusion series accelerated processor (APU). AMD says that APU symbolizes a major change in processor architecture and functionality, incorporating high-performance sequence operations and parallel graphics processing cores into a single wafer, effectively enhancing the processing speed of visual and data intensive operations in today's computing environments. AMD is expected to launch a two APU processor code-named Llano and Ontario by mid-April next year, which is expected to compete with Intel's Westmere and Sandy Bridge processor, which integrates CPUs and GPU.

Another AMD vital nuclear product is the recently launched Marcouni 12 nuclear chip, in the X86 architecture, the core to achieve 12 cores, although from the core number is still less than Sun SPARC 16 core, but for AMD this is also a major breakthrough in the nuclear technology. The chip has a performance of up to twice times that of AMD's 6 core processors, with overall performance up 88% and floating-point performance up to 119%. The mainstream 2-way and high-value-added 4-way server market can meet the clear needs of today's server customers-workload-specific performance, power efficiency, and overall value, while enabling customers to spend less money to gain more core and memory.