It will use them for the FastForward 2 development program

Nov 17, 2014 10:06 GMT  ·  By

Advanced Micro Devices hasn't done much on the supercomputing front, but this may finally start to change now that the company has received an actual grant intended specifically for it.

AMD has been mostly content with its consumer and professional graphics card series, the Radeon and FirePro series. There were also some attempts at supercomputing GPU modules, called FireStream, and it seems like FirePro, as a brand, is starting to cross into that territory as well.

However, overall, AMD has stayed out of the supercomputing industry. There are some installations with its GPUs, but not many.

Soon, though, this may change because of a grant from the U.S. Department of Energy (DOE) with the purpose of helping an Extreme-Scale Computing Research and Development Program.

AMD has received $32 million / €26 million from the DOE

It's not so much of an investment as it is a prize award with a specific research target. Although we suppose you could say that that's arguably the same thing. The sum is actually combined from two awards.

The grant is part of the "FastForward 2" Extreme-Scale Computing Research and Development Program we mentioned previously.

As part of the program, AMD will research integrated exascale node architecture involving HSA-capable APUs (heterogeneous systems architecture accelerated processing units).

Power efficiency, reliability, programmability, component/network integration and advanced memory architecture with efficient data movements are to be the primary hallmarks.

This is different from NVIDIA's approach, which revolves around add-in PCI Express GPU compute accelerators.

It doesn't feat AMD's MO though. The company did something similar (loosely) on the game console market, where it made special APU cores called Jaguar, which are somewhat modular.

If AMD manages to create a supercomputer architecture based on APUs, it might make the current GPU compute model obsolete. It will have to do really well though.

Memory technologies will have to be developed specifically for this, and AMD will be collaborating with the DOE and others to make them happen.

The hopes of the FastForward 2 program

Basically, the plan is to create supercomputers capable of 30 to 60 times more performance than the fastest ones in use today. A tall order, but that's exascale computing for you.

"We're honored to be selected once again by DOE to aid research efforts for exascale-computing," said Alan Lee, AMD's corporate vice president for Research and Advanced Development. "Our customers have long looked to AMD to develop innovative compute, graphics, and memory technologies and products that solve real-world problems. This research is focused on energy-efficient node architectures and memory systems to improve the capability of the world's fastest supercomputers."

Show Press Release