Mobile Masti - Technology and Celeb Blog

NVIDIA has strapped together more graphics processing power than you can shake a stick at, and it's named these pixel-pushing monsters after one of the gods of electricity, Tesla.

Now the company has its GeForce line of graphics cards for gaming and entertainment, Quadro for design and creation and now Tesla for high-performance graphics visualization for oil and gas exploration, brain research, weather prediction and other such intense tasks. And yes, it can certainly play Doom, but that would be like using a nuclear weapon to kill a mosquito.

Check out the gallery for some serious techno-porn of this awesome hardware, where you'll see the deskside unit, graphics card and blade servers in all their glory. Then jump for the specific and awe-inspiring particulars of these muthas.

Using NVIDIA's CUDA software architecture released along with some of the company's other high-end graphics cards a few months ago, NVIDIA has figured out ways to use massively multithreaded parallel computing to do some incredibly powerful computations. These Tesla processors can be lashed together as GPU servers, and then you hook them up to a PC using PCI Express x16 Gen 2 cables.

The GPU computing server blade, which NVIDIA calls the Tesla S870, will have a retail price of $12,000, and will be packed with four x8 series GPUs and suck up 550 watts of juice. That's good for over 500 gigaflops per GPU, and it's only one rack unit high. With four GPUs, do the math. That's some serious power. Stack a few of these up and it starts getting scary, in fact, according to NVIDIA four of those blades strapped together have enough processing power to qualify as one of the top 100 supercomputers.

That badass-looking deskside supercomputer, the Tesla D870, will have two x8 series GPUs and will also require 550W of power, and it's also good for 500 gigaflops per GPU. It''ll sell for $7,500 beginning in August.

The card will also be available in August, and it's called the Tesla C870. It has one x8 series GPU, which also can crank out 500 gigaflops, and it will sell for relatively cheap $1499.

[Source ]