The Department of Energy (DOE) continues to prime the R&D pump for the next-generation of supercomputers, exascale machines 50 times more powerful than today’s leading number-crunchers. In order to bring this goal to fruition, the Department of Energy’s Office of Science together with the National Nuclear Security Administration (NNSA) have awarded $25.4 million in research and development contracts to five influential HPC companies as part of a new initiative called DesignForward.
The select five HPC vendors who will share these exascale investment dollars are AMD, Cray, IBM, Intel Federal and NVIDIA. Their efforts will be instrumental to advancing the state of HPC beyond the petascale era by developing computing technology considered essential to national security, scientific progress, energy security and economic competitiveness.
“In an era of fierce international HPC competition, the development of exascale computing becomes critical not only to our national security missions but to the nation’s economic competitiveness in the global marketplace,” said William Harrod, FastForward Program Manager and Research Division Director for DOE’s Advanced Scientific Computing Research program. “This partnership between industry, the DOE Office of Science and NNSA supports the development of technology to overcome the obstacles on the road to exascale systems.”
DesignForward expands on another DOE exascale initiative called FastForward, which began in 2012. As part of that program, Intel, AMD, NVIDIA and Whamcloud were awarded tens of millions of dollars by the DOE to jumpstart work on extreme-scale supercomputers that satisfy the DOE’s computing needs. While FastForward is dedicated to accelerating the next-generation of hardware technologies at the component level, a major focus of DesignForward will be on the development of interconnect architectures and technologies required to support computations, data movement, and reliability at extreme-scale.
In support of this goal, the DesignForward partners have been tasked with the design and evaluation of interconnect architectures for future HPC systems. As these interconnects are on their way to supporting million and even billion-core parallelism it’s imperative that they be energy efficient, have high bandwidth, and employ ultra-low latency. As for how the work will be divvied up, the DOE notes that “Intel will focus on interconnect architectures and implementation approaches, Cray on open network protocol standards, AMD on interconnect architectures and associated execution models, IBM on energy-efficient interconnect architectures and messaging models and NVIDIA on interconnect architectures for massively threaded processors.” Vendors are expected to work in collaboration with DOE Co-Design Centers to assure that their designs aligns with DOE needs and applications.
A statement from the DOE Lawrence Berkeley Lab website further clarifies each program’s mission:
“FastForward focuses on component technologies in the three areas of Memory, Processor, and Storage technology to lay the ground-work for long-lead-time advances required to meet DOE mission requirements. DesignForward [the one announced today] focuses on fully integrated system design, and key technologies such as the interconnect that are required to enable efficient and scalable systems for extreme-scale computing.”
Both programs will run for at least two years and are aiming to generate production-ready technologies within a five-to-ten year timeframe.