The core of the matter

An Industry Insight by Computer Chips and Processors

AMD’s Raed Hijer gives BM the lowdown on the processing power behind data-intensive industries today.

High Performance Computing (HPC) applications entail solving complex scientific and engineering problems. The petroleum industry, in particular, is an example of an HPC area were massive amounts of data is required to be processed efficiently. The oil exploration process usually involves seismic acquisition, processing and imaging. The process is then continued by analyzing an excess of terabytes of data, which is a time- and resource-intensive process. This type of application and other compute-intensive applications are addressed by High Performance Computing (HPC) supercomputers or cluster-based computers. These systems often deploy hundreds of thousands of CPU cores and involves efficient architecture to handle the compute-intensive and high memory bandwidth requirement with minimal latency.

Built on AMD’s unmatched Direct Connect Architecture which helps eliminate the bottlenecks associated with front-side bus architectures, Quad-Core AMD Opteron processors deliver the world-class performance, scalability and overall system efficiency necessary to power the world’s largest high-performance computing (HPC) systems.  Additionally, AMD technology for these highly advanced systems provides the customers with much-needed energy efficiency in the form of performance-per-watt and innovative processor-level power management features such as AMD CoolCore Technology and Independent Dynamic Core Technology.

With total cost of ownership (TCO) in mind, AMD Opteron processors can be deployed on the same infrastructure and same thermal envelops from one generation to the next. This enables the customers to adopt new technologies based on the next generation Opteron without the hassle of redesigning the datacenter space, power and cooling requirements. As an example, customers can migrate from Dual Core Opteron processors to Quad Core Opteron processors with the same rack space and power requirements needed to power and cool the rack. By so doing, they are able to double the number of cores in the same infrastructure.

The TOP500 list ranks the world's 500 fastest high performance computers. Seven of the top 10 in the list, including the top two, leverage the balanced platform of AMD Direct Connect Architecture. The list reaffirms AMD’s leadership and the tremendous performance capability that HPC customers have enjoyed for years. Last year, the petaflop barrier was broken by both IBM’s ‘Roadrunner’ supercomputer at Los Alamos National Labs and Cray’s XT4 and XT5 ‘Jaguar’ supercomputer at Oak Ridge National Laboratory. While the former supercomputer adopts a heterogeneous architecture based on AMD Opteron and IBM PowerXCell processors. The latter is considered the first ever wholly x86-based supercomputer to achieve such a performance milestone.

The ‘Jaguar’ supercomputer runs on over 45,000 Quad-Core AMD Opteron processors. The petaflop milestone is significant in terms of solving problems that were never possible before. It reduces dramatically the computational time. It could produce results in a week time that used to take 20 years just a decade ago. In oil and gas applications, this could enable processing the seismic data several times faster than a year before. In addition to seven of the premium top ten, AMD Opteron processors have helped catapult 53 other global supercomputers on the TOP500 to new heights of HPC performance.  The list highlights industries of all types that are harnessing AMD’s world-class computing capability including IT service providers, financial institutions, automotive design, and researchers in energy, geology, meteorology, social sciences, astronomy and many other disciplines.

Recently, AMD announced widespread availability of its 45nm Quad-Core AMD Opteron processors, codenamed “Shanghai.” These latest Quad-Core AMD Opteron processors already hold multiple HPC workload-related performance records including:
• 34 percent higher floating point throughput performance than the competition on SPECfp_rate2006 for two-socket servers.
• 51 percent more memory bandwidth performance than the competition on STREAM benchmarks for two-socket servers.
• Approximately 65 percent faster job runs than the competition on FLUENT 12 beta (ANSYS) for two-socket servers.
• More than 30 percent better performance than the competition on LSDYNA MPP971s for two-socket servers.

www.amd.com/hpc

Raed Hijer is a Senior Field Application Engineer with Advanced Micro Devices (AMD), Inc. in Dubai. Prior to joining AMD, he worked for IBM Corp. in North Carolina in a variety of roles including desktop motherboard development, product engineering technical lead, and system lead engineer. While working at IBM, he published several papers and patents in the areas of image processing, pattern recognition, neural networks and PCB design.

Contact details:
Raed Hijer, Senior Field Application Engineer
T: +971 (0) 50 524 3387, E: raed.hijer@amd.com
www.amd.com