Thursday 11 October 2012

GPGPU Hardware

 GPGPU Hardware
Graphics Processing Units (GPU’s) designed to accelerate graphics applications are highly parallel processors capable of 100’s of Gflops/sec. Competition among the GPU vendors for market share in the PC gaming market has driven technological advancements in graphics cards, and the volume in this market has driven prices down.

These processors can be applied to non-graphical high-performance computing applications. Researchers have been investigating this usage for several years with success in a number of areas. This new market for their technology has been recognized by the main graphics card vendors, Nvidia and AMD/ATI. Both have introduced product lines specifically targeting high-performance computing, sometimes called the GPGPU market for general purpose computing on graphics processing units.
 



Nvidia Tesla Products
 
Nvidia introduced the Tesla product line in 2007. The first Tesla card is called the C870 and is based on their high-end graphics card, the FX 5600, but lacks video outputs and, at $1,499, is priced at half the $2,999 FX 5600. Like graphics cards, the Tesla C870 requires a PCI-Express 16x slot. It also draws a lot of power, about 170w, and spacewise it takes up two slots and is full-length. An important limitation of the Nvidia GPU’s is that they only support single-precision floating-point arithmetic


 
Tesla C870 – 170w, 1.5GB, 518Gflops peak
 
 AMD FireStream

 
AMD made FireStream GPGPU boards available for early adopters and developers in 2007. In November 2007, they announced the FireStream 9170, intended for production use, with general availability expected in Q2 2008. The board is priced at $1,999, draws less than 100w, but is similar in size to the Tesla C870. Unlike the Nvidia GPU’s, the FireStream supports double-precision, but at an estimated102Gflops peak compared to the single-precision performance of 500Gflops peak.


 
AMD FireStream 9170 – 100w, 2GB, 500Gflops peak
 

 




 




 

No comments:

Post a Comment