“Knights Corner” Puts Intel Back in the Co-Processor Driver’s Seat

November 17, 2011, By George Lang

Intel introduced its latest super co-processor, Knights Corner, this week at Seattle’s supercomputing SC11 conference. About the size of a credit card, it is more powerful than Intel’s original 1997 teraflop (a trillion calculations per second) supercomputer that filled the room with nearly 10,000 Pentium Pro processors!

Until now, competitors Advanced Micro Devices (AMD) and nVidia have led the near-teraflop race with 928 Gigaflops (via paired HemlockXT 5970 chips)  and 672 Gigaflops (GTX 480) respectively; with Intel trailing back at 109 Gigaflops (Core i7 980 XE). The interesting difference is that Intel’s new 22 nano-meter Knights Corner chip operates on the multi-integrated core MIC x86 architecture and is capable of performing current applications written on it, at the improved speeds, with little or no programming changes (Poeter). “This will allow scientists to use both CPU and co-processor performance simultaneously with existing x86 based applications, dramatically saving time, cost, and resources that would otherwise be needed to rewrite them to alternative proprietary languages” (Intel).

The AMD and nVidia graphic processing units (GPUs), on the other hand, are designed primarily for rendering images and require special programming for each chip configuration (BBC). This fact alone should attract software developers into Intel’s nest. The Intel co-processor will not, however, take the place of these GPUs any time soon.

Knights Corner will support the new PCI Express 3.0 bus specification, which doubles the speed of the older 2.0 standard; and, while it is primarily designed for the Xeon-based server platform, the side-matter of this high performance computing (HPC) technology will surely see the light of day in commercially available personal PCs in the not to distant future.


© 2008-2012 DeviceMag.com - All rights reserved | Privacy Policy