AMD's New S10000 GPU, Opterons Power Top Energy-Efficient Supercomputers
That system, dubbed SANAM, is a Saudi Arabian build-out located in the King Abdulaziz City for Science and Technology. It's capable of 2.35 GFLOPS/watt and built on Intel's Xeon E5-2650 eight-core processors. There's a total of 38,400 cores in the system, alongside 420 GPU cards (a total of 840 GPUs). The machine's efficiency doesn't make it a lightweight; SANAM sits at a respectable #52 on the TOP500 list.
"There is a massive shift in the HPC industry from defining supercomputing performance in terms of FLOPS to performance per-watt and developing environmentally responsible technologies," said John Gustafson, senior fellow and chief product architect, Graphics Business Unit, AMD. “The SANAM supercomputer, equipped with AMD FirePro™ S10000 GPUs based on our Graphics Core Next Architecture, consumes a mere 180 kilowatts of power while running, which is incredibly low for a supercomputer. With solutions like the AMD FirePro™ S10000 along with innovative software from AMD and technology partners like PGI, AccelerEyes, and CAPS Entreprise, we enable institutions engaged in cutting-edge research, like the University’s work in Quantum Chromodynamics, to achieve massive compute performance and processor density, while staying within their power budget."
The #1 system on the Green500 is a Xeon E5 + Xeon Phi rig, while the 3rd and 4th spots are owned by the Oak Ridge National Laboratory (Opteron 6274 + K20X) and the Swiss Scientific Computing Center (confusingly abbreviated CSCS). That system is based on the Opteron 6272, backed up by Nvidia's Kepler. The last six spots are rounded out by BlueGene/Q configurations.
The need to improve computer efficiency and FLOPS/watt isn't just a passing fad, it's a major concern. The only way for computer systems to continue to scale to the level needed to break the exaflop barrier is for continued enormous improvements in total computer efficiency. Continuous improvements at every level are necessary, and the work AMD has done on the consumer side to bridge CPUs and GPUs will eventually make an appearance in these sorts of supercomputing environments.
AMD is continuing to develop software products that address the needs of HPC developers:
Accelereyes ArrayFire: Accelereyes is dedicated to delivering fast, simple GPU software. With the GA release of ArrayFire, a GPU software acceleration library that provides hundreds of functions already optimized for speed by top GPU computing experts, this allows for easy integration into C, C++, Fortran, and Python applications
Portland Group (PGI) Accelerator compilers: PGI Accelerator Fortran, C and C++ compilers targets the AMD line of accelerated processing units (APUs) as well as the AMD line of discrete GPU (Graphics Processing Unit) accelerators. PGI continues to work closely with AMD to extend its PGI Accelerator directive-based compilers. The goal is to generate code directly for AMD GPU accelerators, and to generate heterogeneous x64+GPU executable files that automatically use both the CPU and GPU compute capabilities of AMD APUs
CAPS Entreprise HMPP compiler: CAPS Entreprise is a leading provider of solutions for deploying applications on “many-core” systems. CAPS source-to-source HMPP™ compiler is based on C, C++, and FORTRAN directives and supports OpenACC® and OpenHMPP standards. With help from AMD, the compiler incorporates a powerful OpenCL™ parallel data generator
AMD CodeXL: AMD CodeXL is a comprehensive tool suite that enables developers to harness the benefits of AMD CPUs, GPUs and APUs. It includes powerful GPU debugging, comprehensive GPU and CPU profiling, and static OpenCL™ kernel analysis capabilities, enhancing accessibility for software developers to enter the era of heterogeneous computing. AMD CodeXL is available as both a Visual Studio® extension and a standalone user interface application for Windows® and Linux®
One of the changes since Rory Read came onboard has been a shift in focus to prioritize software development and working with developers across consumer and professional ecosystems. AMD has always done some compiler and optimization work, but the company is putting more effort into building tools that let companies harness the capabilities of its APUs and GPUs.