by Bill Dally

I’m happy to announce that the U.S. Department of Energy (DOE) has awarded NVIDIA a $12.4 million contract to research and develop technologies to achieve exascale computing. We’re very excited to work closely with DOE scientists to advance the frontiers of science.

The two-year contract calls for NVIDIA to conduct research and development in processor architecture, circuits, memory architecture, high-speed signaling and programming models to enable an exascale computer at a reasonable power level. The concept is to use thousands of efficient, throughput-optimized cores to perform the bulk of the work, with a handful of latency-optimized cores to perform the residual serial computation. We’ll work with scientists at seven DOE laboratories to ensure our design meets their needs and runs their scientific workloads.

The award is part of the DOE’s FastForward program, which is funding a handful of technology companies to accelerate R&D on exascale computing, the next great challenge in supercomputing. The DOE recognizes that a highly parallel, heterogeneous computing model is well suited for processing demanding scientific and technical computing loads.

Exascale systems will perform a quintillion floating point calculations per second (that’s a billion billion), making them 1,000 times faster than a one petaflop supercomputer. The world’s fastest computer today is about 16 petaflops.

One of the great challenges in developing such systems is in making them energy efficient. Theoretically, an exascale system could be built with x86 processors today, but it would require as much as 2 gigawatts of power — the entire output of the Hoover Dam. The GPUs in an exascale system built with NVIDIA Kepler K20 processors would consume about 150 megawatts. The DOE’s goal is to facilitate the development of exascale systems that consume less than 20 megawatts by the end of the decade.

Achieving this level of efficiency will require extraordinary innovation on a number of fronts. However, we believe that heterogeneous computing offers the best approach to get there.

Why is the DOE Doing This?

The U.S. is eager to develop exascale systems for reasons of national security and economic competitiveness. Supercomputing systems help accelerate discovery and innovation, benefiting a large number of industries. Supercomputers are also necessary to help solve the world’s most difficult scientific challenges, including finding cures for disease, studying climate change and developing more efficient engines.

This effort will build on earlier work we began in 2010 for the U.S. Defense Advanced Research Projects Agency (DARPA). This earlier work, known as Project Echelon, laid much of the groundwork needed to achieve resilient, highly efficient throughput processors.

One of the reasons I joined NVIDIA is because the company pioneered the use of massively parallel accelerators for supercomputing. We’ve made a lot of progress — there are already more than 50 GPU-based systems on the latest Top500 list of the world’s most powerful systems – but the research we will conduct in the next few years will enable profound scientific breakthroughs. I’m very eager to see them.