Nvidia closing in on 2 teraflops with graphics card
- — 09 January, 2009 01:07
In a bid to boost 3D gaming, Nvidia on Thursday announced a graphics card with 480 cores that can crank up performance to reach close to 2 teraflops.
The company's GTX 295 graphics cards has two graphics processing units (GPUs) with 240 cores each that can execute graphics and other computing tasks like video processing. The product was announced at the Consumer Electronics Show in Las Vegas.
Nvidia is targeting the card at gamers, enthusiasts, and visual computing application users. It will bring more realism to gameplay, including deeper graphics details and better character motion, said Bryan Del Rizzo, an Nvidia spokesman.
The card delivers 1.788 teraflops of performance, which Nvidia claims is the fastest single graphics card in the market. It outperforms its closest competitor, the ATI Radeon HD 4870 X2, by close to 15 percent when playing games, according to the company.
That claim isn't out of the realm of possibility, said Jon Peddie, president of Jon Peddie Research.
"I do expect it to take the performance crown -- at least until ATI pulls the next trick out of their hat," Peddie said.
ATI could not be reached for comment on the story.
For normal gamers, the card better enables some unique technologies to make gaming a more engaging experience.
The card enables users to play games on a three-dimensional display, although 3D glasses made by Nvidia and supported monitors are needed. 3D monitors are made by companies including ViewSonic and Samsung, and around 350 games support 3D gaming, according to Nvidia.
The graphics card also supports PhysX, a hardware and software engine that adds physical reality to games, like smoke billowing from an object after an explosion, or the behavior of a rock after it hits a target. Developers can use the PhysX software development kit to create interactive middleware for games.
Performance is also boosted by an improvement in Nvidia's manufacturing process for GPUs, Peddie said. The new chips are manufactured using the 55-nanometer process, while Nvidia earlier employed a 65-nanometer manufacturing process. The new chips are also more power efficient as a result, Peddie said.
The monster of a card supports around 1.8GB of graphics memory. It also supports quad-SLI (Scalable Link Interface) technology, which allows two GTX 295s, or a total of four GPUs, to work together to scale graphics in a wide variety of systems, Del Rizzo said.
"There are professional spaces where we support more cards and more GPUs, but in the consumer space, you are actually limited to PCI Express slots and restricted by thermal and power requirements inside the PC," Del Rizzo said.
Outside of gaming, the massive processing power across cores could boost Nvidia's push into the GPU-based general purpose computing space. The card can convert PCs to mini-supercomputers, scaling application performance by executing code simultaneously across many processing cores.
"Things like video encoding, video conversion, video algorithms, physics calculations, and distributed computing -- like Folding at Home and Seti at Home -- can now be run on the GPU, magnitudes faster than what could be run on the CPU," Del Rizzo said.
Nvidia for years has pushed its CUDA parallel programming architecture, a software toolkit that allows for the development of programs to be executed on its graphics processors. The company has also extended its support to OpenCL (Open Computing Language), a specification also supported by Advanced Micro Devices and Apple that enables programmers to take advantage of the processing speeds of GPUs and other processors for faster program execution.
The card fits into any normal PCI Express slot. Price of GTX 295 is US$499. The company has also launched the GTX 285, which includes one GPU, and is priced at $399.