The US Department of Energy (DOE) has awarded more than $40 million in funding to AMD, Intel, Nvidia and Whamcloud as part of its FastForward programme for the advancement of exascale research and development.
A jointly-funded collaboration between the DOE Office of Science and National Nuclear Security Administration (NNSA), FastForward is contracted through Lawrence Livermore National Security in conjunction with a consortium of seven US laboratories that includes Argonne National Laboratory, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Oak Ridge National Laboratory, Pacific Northwest National Laboratory and Sandia National Laboratories. Under the programme, partnerships have been initiated with multiple vendor companies to develop technologies, such as processors, memory and storage, that will pave the way for the first exascale supercomputer.
AMD has been awarded $12.6 million for two research projects. The first, AMD for processor-related research, will receive up to $9.6 million, while up to $3 million will go towards memory-related research. Many of the world’s leading supercomputers, including IBM’s Roadrunner computer at the DOE’s Los Alamos National Laboratory, use AMD Opteron processors and last year it was announced that 19,200 AMD Opteron 6200 Series processors will be used to help power the DOE’s new Titan system also at Oak Ridge National Laboratory when it becomes fully operational in 2013. The company is also responsible for developing the world’s first and only Accelerated Processing Unit (APU).
Two subcontracts totalling $19 million have been awarded to Intel Federal, a wholly-owned subsidiary, for research and development in a number of areas including memory and how to achieve more reliable and energy-efficient processor technology. Intel's memory research, in particular, will evaluate how next-generation memory architectures, combined with processing power, provide optimal, energy-efficient performance for a broad range of DOE applications and other HPC workloads. Intel has developed new generations of its Xeon processors with entirely new Many Integrated Core architecture-based Intel Xeon Phi co-processors. In addition, the company has recently acquired Infiniband and interconnect assets from QLogic and Cray, which will help increase the speed of data delivered on exascale-class platforms.
Writing on the Nvidia blog, Bill Dally announced that the company has been awarded a $12.4 million contract to conduct research and development in processor architecture, circuits, memory architecture, high-speed signalling and programming models to enable an exascale computer at a reasonable power level. ‘One of the great challenges in developing such systems is in making them energy efficient,’ he wrote. ‘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.’ Dally also commented that achieving this level of efficiency will require extraordinary innovation on a number of fronts and that heterogeneous computing offers the best approach to get there.
Although the exact figure is unknown, the subcontract for storage and I/O research and development has been awarded to Whamcloud. Beginning immediately, the subcontract incorporates application I/O expertise from the HDF Group, system I/O and I/O aggregation expertise from EMC Corporation and scale testing facilities from Cray, teamed with file system, architecture and project management skills from Whamcloud. All components developed in the project will be open sourced for the benefit of the entire Lustre community.
All FastForward contracts are scheduled for two years through the middle of 2014.
