Today NVIDIA is introducing two new Graphics Processing Units (GPUs) based on their GeForce GTX 200 Series - the long-awaited GeForce GTX 280 and GeForce GTX 260. Both GPUs are high-end products targeted at gaming enthusiasts and will be featured on graphics cards retailing for $649 and $399 respectively.
Graphics cards based on the GeForce GTX 280 are expected to be available to purchase tomorrow, June 17, while the GeForce GTX 260 is scheduled for availability next Thursday, June 26. In support of today's launch, nV News was provided with a reference GeForce GTX 280 from NVIDIA and a retail GeForce GTX 280 FTW Edition from EVGA.
The GeForce GTX 280 is NVIDIA's most complex and powerful GPU to date and contains over 1.4 billion transistors. Based on second generation unified architecture, the GeForce GTX 280 functions as a multi-purpose processor capable of accelerating real-time 3D graphics as well as computationally intensive parallel-computing applications. With a total of 240 shader processors, 80 texture processors, and 1GB of frame buffer memory, the GeForce GTX 280 will provide unprecedented levels of 3D graphics performance.
EVGA GeForce GTX 280 FTW Edition
Since its acquisition of AGEIA, NVIDIA has been porting the PhysX Application Programming Interface (API) over to CUDA. In fact, NVIDIA is in the process of finalizing a driver that will allow the existing 70 million plus GeForce 8 and GeForce 9 based graphics cards to hardware accelerate PhysX applications via CUDA. Also, be on the lookout for a Folding@Home Client and GPU-assisted video transcoding acceleration available with BadaBOOM Media Converter.
By leveraging NVIDIA CUDA Technology, the GeForce GTX 280 can transform itself into a fully programmable multi-processor with 240 cores, on-die shared memory, random read and write capability with 1GB of dedicated memory and the ability to deliver close to one trillion floating point operations per second (teraflop). NVIDIA refers to this mode of operation as computing mode, which can accelerate applications such as video transcoding, physics, and scientific computing.
The GeForce GTX 200 Series features a number of architectural improvements over the previous generation of GPUs. Some of the goals that NVIDIA's engineers set out to accomplish include the following:
Design a processor with up to 2x the performance of the GeForce 8800 GTX
Support future games that will feature increased shader complexity
Increase the amount of available graphics memory
Improve architectural efficiency per watt and per square millimeter
Improve DirectX 10 performance including geometry shading and stream out processing
Increase the computational capability for high-performance CUDA applications and GPU physics
Improve power management by reducing load when 3D hardware acceleration is not required
Some of these goals were achieved by incorporating the following improvements:
Increasing the number of simultaneous threads being processed by 3x (10,000 to 30,000 threads)
Incorporate a new scheduler design for 20% more textural efficiency
Increase the bus width of the memory to GPU interface
Improve z-cull and compression technology for better performance at high resolutions
Provide full-speed raster operation (ROP) frame buffer blending (vs. half-speed on the GeForce 8800 GTX)
Improve dual issue of program instructions for efficiencies with high performance computation
Double the number of registers to accommodate longer and more complex shader programs
IEEE 754R double precision for improved floating point computational accuracy
Hardware support for 10-bit color scan out (DisplayPort only)
The GeForce GTX 280 and GeForce GTX 260 both exceed NVIDIA's high-end GeForce 8 Series offerings in the amount of memory available and memory bus width. Compared to the 768MB of graphics memory available on the GeForce 8800 GTX, which has withstood the test of time, the GeForce GTX 280 features 1GB of GDDR3 memory while the GeForce 260 is outfitted with 896MB.
GeForce GTX Specs
In addition, the GeForce GTX 280 and GeForce GTX 260 feature a wider memory bus than the GeForce 8800 GTX. The GeForce 8800 GTX has a 384-bit memory bus, while the GeForce GTX 280 and GeForce GTX 260 have memory bus widths of 512-bits and 448-bits respectively. These two features alone will provide the next-generation GPUs with greater performance when high quality graphics and textures are used in conjunction with antialiasing at high resolutions.