Post a CommentNews and New Products
Tensilica, Lawrence Berkeley lab tackle energy-efficient supercomputing for climate modeling
Core-based architecture aims to achieve 1 quintillion FLOPS while providing an order-of-magnitude improvement in FLOPS per watt, compared with systems using desktop and server chips.
By Matthew Miller, Editor-in-Chief, EDN.com -- EDN, 5/5/2008
Processor-core maker Tensilica and the U.S. Department of Energy's Lawrence Berkeley National Laboratory today announced a collaboration to develop new designs for supercomputers that, by operating with greater power efficiency, will deliver up to 1000× more scientific number-crunching horsepower than is practical with existing systems.
In an irony of this environmentally conscious era, the supercomputers used to study issues such as climate change themselves impose a significant carbon footprint—consuming megawatts of electricity both directly and for the elaborate cooling systems that are required to deal with the excessive heat they generate. Even so, scientists wishing to tackle leading-edge research need 100× to 1000× more computing throughput than today's high-end systems can provide, according to Tensilica.
|
The team will use Tensilica's Xtensa LX extensible processor core as a basic building block. Each core dissipates just a few hundred milliwatts while churning out billions of FLOPS, representing an order-of-magnitude improvement in FLOPS per watt over traditional desktop or server processor chips, according to Tensilica. A supercomputer harnessing millions of such cores, tightly integrated at the chip, board, and rack level, will achieve the exascale goal within a power budget of "a few megawatts," the company stated.
Among the challenges the effort will have to address:
- Optimizing memory and communication bandwidth across such a massive array
- Efficient distribution of application functions
- Development of effective prototyping and software-development methods.
"Our studies show that energy costs make current approaches for supercomputing unsustainable," Horst Simon, associate laboratory director, computing sciences, for Berkeley Lab, said in a statement. "Hardware-software co-design using tiny processor cores, such as those made by Tensilica, holds great promise for systems that reduce power costs and increase practical system scale."
Tensilica claims its configurable cores are well suited to the project's demands. The effort will use automatic generation of processor designs, FPGA-based hardware implementation, and software tools that enable rapid prototyping and evaluation of processor instruction sets, interfaces, and multiprocessor-communication mechanisms, according to the company.
