Last weeks we repored about graphics circuit maker NVIDIA working on the launch of its new flagship GeForce GTX 590. With 1024 CUDA cores and dual GF110 graphics circuit it is an extreme graphics card demanding a lot from the other component, where NVIDIA only uses the finest of its GPU samples and availability will be limited.
NVIDIA had a successful launch of GeForce GTX 580 where it knocked down AMD with the new GF110 GPU that with high performance and relatively reasonable power consumption took over as the king of the market. AMD couldn't counter with a single Cayman GPU, but it still has the fastest card around with its dual-GPU Radeon HD 5970.
AMD's plan was to maintain this, for PR mostly, trump with the launch of Radeon HD 6990 "Antilles" sporting dual Cayman circuits, but NVIDIA's plan is to crash the party through an unsuspected launch of a Fermi-based graphics card with dual GPUs.
As we revealed we will be dealing with fully featured GF110 circuits with 512 CUDA active cores each. To make sure the graphics card won't go above any power consumption specifications NVIDIA has had to turn down both voltages and clock frequencies, but it has also starting sorting out the circuits at TSMC where the finest GF110 samples are removed from the belt and put on line for the flagship GeForce GTX 590.
NVIDIA wants the GF110 circuits with the least leakage to maximize the clock frequencies without power consumption skyrocketing. It comes as no surprise that GeForce GTX 590 will be in limited supply. NVIDIA has no hopes of making any real money from this monster card, but it sees an opening to whip AMD and launch what it could call the fastest card in the world - no bars.
NVIDIA's reasoning is most likely that the positive PR is going to boost sales overall.