Nvidia Corp., the world’s largest supplier of graphics processors, said that it has no immediate plans to release graphics cards that offer higher speed than current top-of-the-range GeForce 8800 Ultra, but said that customers seeking for extreme performance will soon be able to install three graphics cards into one system to get incredible graphics rendering horsepower.

“We decided that the refresh that we would do this time was [GeForce] 8800 GT, and this is just a barn burner refresh. We are really proud of the 8800 GT and we are going to put our focus here. From [two GeForce] 8800 GT [graphics cards], you could obviously do SLI and soon you’ll be able to do three-way SLI, so you are going to be able to put a lot of GPU horsepower into your system, starting with a very affordable 8800 GT, and so – this is our focus for now,” Jen-Hsun Huang, the chief executive of Nvidia, said during a question and answer session with financial analysts when asked about a possible “halo” product refresh in 2007.

In previous years Nvidia refreshed its graphics cards lineup two times a year, in spring and in autumn, usually introducing a high-performance graphics product to create so-called halo effect. However, this year the company seemingly has decided not to release a single-chip high-end solution, but to enable proper support for its SLI multi-GPU technology from the drivers to make performance-demanding gamers buy existing GeForce 8800-series graphics cards and create multi-GPU systems with two, three or even four of them. Since three graphics cards that cost $249 will bring Nvidia about $750 in sales, while a typical high-end board costs $599 - $649, it does make sense for Nvidia to push such graphics solutions.

Performance of multi-GPU graphics sub-systems fully depends on driver and software optimizations. Therefore, not all games can benefit from two or three or even four graphics processing units (GPU).

ATI, graphics product group of Advanced Micro Devices, also said earlier this year that it would focus on development of multi-GPU solutions for high-end market instead of creating large graphics chips with roughly a billion of transistors, which are hard to produce and develop.