OPTIMIZED SYSTEMS

GPUs Play Central Role in Next-Gen Supercomputer

Salvatore Salamone | Date: 10-12-11 | Comments

Plans to upgrade the world�s third-most powerful supercomputer, the Oak Ridge Lab Jaguar system, were announced this week. The new system will rely on GPUs (paired with traditional CPUs) to enable an energy-efficient boost in performance. It is expected to be completed by 2013 and could deliver 20 petaflops of processing power.

Oak Ridge National Laboratory (ORNL) this week announced it has awarded a contract to Cray to boost the power of its top-performing Jaguar supercomputer, transforming it into a next-generation system dubbed Titan.

The upgrade will replace all of the current Jaguar AMD processors with the latest AMD Opteron processors, code-named Interlagos. This will transform the system from two six-core processors per node to one 16-core processor per node, and it will allow integration of graphics processing units (GPUs).

The newly proposed Titan system will extend the range of work that is currently carried out on Oak Ridge�s Jaguar supercomputer.

In the first phase of the upgrade, scheduled to begin this year, all of the processor boards, memory and system interconnect network components in Jaguar will be replaced with the new Cray XK6 processor board. This upgrade will also include approximately 1,000 current Nvidia Tesla 20-series GPUs. Researchers will begin modifying their scientific applications to gain a performance boost from the new technology.

In the second phase, to be performed in the second half of 2012, between 7,000 and 18,000 of Nvidia's next-generation Tesla GPUs will be added to the system, increasing the peak performance to between 10 and 20 petaflops. That represents an order of magnitude more performance than the current configuration. (In the most recent release of the list of the world�s top 500 most powerful supercomputers, the Jaguar system ranked third, with a peak power of 1.75 petaflops.)

The upgrade will deliver one-third more cores in the same physical space compared with Jaguar while also doubling the memory and outfitting the system with Cray's powerful "Gemini" network, which will improve performance on scientific applications.

As noted, the processing power of the AMD CPUs will be complemented with the NVIDIA Tesla GPUs, which are energy-efficient processors that accelerate specific types of calculations in scientific application codes. The use of GPUs in supercomputers is a growing trend. Currently, 19 of the world�s top 500 most powerful supercomputers use GPUs.

"[The] game-changing new technology will be the inclusion of advanced GPUs," said ORNL Director Thom Mason in a statement. "These Tesla GPUs are able to perform many more calculations for the same amount of power as conventional microprocessors, and [they] will work hand-in-hand with CPUs to deliver new levels of energy-efficient application acceleration."

Titan's initial simulation projects will include studies of the commercial viability of producing biofuels and biomaterials; investigations into combustion in order to burn fuels cleanly in efficient engines; and simulations to aid in the development of new materials for photocells and new battery technologies. Additionally, Titan will be used to study ways to extend the lifecycles of nuclear power plants safely, nuclear reactions, the flow of contaminants in the ground, and the impacts of energy use on climate.

"All of these areas of science will benefit from Titan's enormous increase in computational power," Mason said. "Titan will allow for significantly greater realism in models and simulations and the resulting scientific breakthroughs and technological innovations will provide the return on this national investment. Discoveries that take weeks, even on a system as powerful as Jaguar, might take days on Titan."