Nvidia to Blaze Path to Exascale??

One of the biggest hurdles to exascale computing is power and heat. As we amass the vastly large number of processors required to break the exascale speed limit it becomes apparent that it will take an enormous amount of energy to power an exascale system and that an exascale system will give off an unbelievable amount of heat. Of course this assumes that we build an exascale system with current modern-day processors from Intel or AMD. But what if we didn't...? What if we used something else...?

ARM processors, developed by ARM Holdings, are 32 bit RISC chips that are significantly simpler than X86 based chips. This simplicity leads to less power consumption, which in turn produces less heat. While it is true that ARM chips consume less power than X86 based chips it is also true that they are computationally less powerful. Maybe it might not be the brightest idea to try and build an exascale supercomputer out of them. What if we turbo charged the ARM chips with Nvidia GPUs? Heterogeneous supercomputer designs, combining x86 chips with Nvidia GPUs, are all the rage these days. If we were to utilize a vast number of low powered ARM chips to drive vast number of high powered GPUs perhaps we could break the exascale speed limit without having to build a nuclear plant to power the system.

Well that's exactly what the good folk's at Nvidia are doing for us with their "Project Denver". According to Nvidia, "NVIDIA's project Denver will usher in a new era for computing by extending the performance range of the ARM instruction-set architecture, enabling the ARM architecture to cover a larger portion of the computing space. Coupled with an NVIDIA GPU, it will provide the heterogeneous computing platform of the future by combining a standard architecture with awesome performance and energy efficiency. An ARM processor coupled with an NVIDIA GPU represents the computing platform of the future. A high-performance CPU with a standard instruction set will run the serial parts of applications and provide compatibility while a highly-parallel, highly-efficient GPU will run the parallel portions of programs."

Now all they need to do is deliver on their promise and the road to exascale might be just around the corner.

ClusterChimps © 2009