Untitled Document
Not a member yet? Register for full benefits!

 New Nvidia GPU: A Supercomputer in a PC?

This story is from the category Computing Power
Printer Friendly Version
Email to a Friend (currently Down)



Date posted: 05/12/2006

The GPU, or graphics processing unit is the heart of any 3D accelerator graphics care. It is a powerful workhorse, frequently more powerful than a PC?s CPU ? central processing unit ? and dedicated solely to graphics. The pace of GPU advancement has picked up substantially in recent years, with Nvidia and ATI Technologies, going at it hammer and tongs, to try and outdo one another.

On Wednesday the 8th of November 2006, Nvidia announced the GeForce 8800, a new graphics accelerator chipset, which is currently the most powerful available for the home market. The new GPU chipset has 128 individual processors on it, for massive multi-core processing power. In fact, it has power equivalent to the fastest supercomputer on the planet, circa 1999.

The sequel to the 8800, which Nvidia plans to ship in about a year, is slated to have double the mathematical power, making it actually rival modern mainframe systems.

GPUs are not just displaying graphics now. The architecture of a GPU board is far superior to the older architecture of a PC motherboard, and this lends the processing units far more efficiency and power, something that both Nvidia and ATI are noticing. Both feel the line between GPUs and conventional CPUs is beginning to blur and fade.

Earlier this year, the lessening of a distinction was shown crystal clearly when Microsoft Research Lab?s annual competition to find the fastest search algorithm was won in 2006, by a team that used a 3D graphics card to do all the processing, rather than a traditional PC system.

GPUs look like they may well be the heirs of Moore?s law. Whilst CPU design has become bogged down in recent years, with just the addition of new cores, GPU architecture ? which sees the processors, ram, and other components on the chip as part of a single whole rather than discrete components ? has continued to leap ahead, doubling in efficiency and power every year.

The 8800 was demonstrated to the public last Wednesday. One of the programs used to demonstrate it, was this frog below.

The frog can be pulled and prodded by the disembodied hand, stretched and slapped around. In each case it deforms organically, and correctly, in real-time. The ray-traced light in the scene is generated dynamically, also in real-time (the 8800 has processors dedicated solely to ray-tracing)

?We?re just entering an era where it is possible to capture photo-real human faces in motion.? Said Steve Perlman, a Silicon Valley technologist.

?The 8800 allows us to display that level of realism in real time.?

See the full Story via external site: www.nytimes.com

Most recent stories in this category (Computing Power):

19/02/2017: Printable solar cells just got a little closer

04/02/2017: 1,000x more efficient nano-LED offers possibility of faster processors

31/01/2017: For this metal, electricity flows, but not heat

26/01/2017: Google brings AI to Raspberry Pi

12/01/2017: Researchers turn memory chips into processors to speed up computing tasks

08/01/2017: Intel announces Compute Card – A full PC the size of a Credit Card

23/12/2016: Scalable energy harvesting of unused mechanical energy in the environment

28/11/2016: Japan kicks off AI supercomputer project