Next generation NVIDIA in the 40nm manufacturing process will share the basic design concept with the GT200 design. This architecture was introduced at 65nm and was rather hot, and forced NVIDIA to make very big chips, but eventually the architecture was flexible enough to shrink to 55nm and bring some new products such as GeForce GTX 285 and GeForce GTX 295 all based on 55nm GT200b chips. The original idea was carried from 65nm to 55nm technology. The future will bring a similar concept but at 40nm process and you can easily expect more shaders, higher clocks and less heat dissipation from these new GPU's. Though that is somehow the usual you can expect with new products. - FudZilla

NVIDIA 40nm Chipset To Be Similar As GT200 Design

Knowing NVIDIA there should be something innovative, but many sources have confirmed that the 40nm design basics will be close to their GT200 concept. Though NVIDIA bets a lot in 2009 on CUDA and PhysX and some innovations are expected in this field. NVIDIA will probably try to optimise its GPU's further for parallel computing. NVIDIA will still stick to big monolithic chips for the high-end, pushing the transistor count and hopefully performance limits to the next level, and it always has a safe net woven of two GPU's on a single card. This is exactly what they have done with the GeForce GTX 295 graphics solution.

If ATI's new generation 40nm dual GPU card, lets call it R800 ends up faster than a single 40nm high-end NVIDIA chip, NVIDIA will use two of its chips for its top range video card, and the game is on. According to some reports the specs of NVIDIA's coming flagship 40nm GT212 architecture have been identified and revealed from unnamed sources near NVIDIA, as usual. To start matters off, the GT212 is the successor to their 55nm GT200b which is currently rolling into production. The upcoming architecture is essentially following two footsteps, a die shrink to a smaller fabrication process and a decrease in memory interface.

The memory interface of the GT212 will decrease from 512-bit on the GT200 to 256-bit, which is quite similar to what occurred with AMD's RV770 chipsets. To compensate however, NVIDIA will follow suit by incorporating some Hynix 7Gbps GDDR5 memory into its GT212 as well as into the rest of its 40nm lineup. In addition, the stream processor count of the GT212 will increase 384 and the number of texture mapping units will increase from 96 as well. Because of the sharp increase in stream processors, the number of transistors will jump up to nearly 1.8 billion, a great 400 million more than their current GT200 chipset.

The new die surface will be reduced as well which should significantly reduce manufacturing costs for NVIDIA when the company needs it most. All in all, the complexity of the new architecture will be less than a fifth more than the GT200 card due to lower PCB complexity needed as a result of the lowered memory interface. The new architecture is confirmed to launch during the second quarter this year, most likely around April or a bit later, although the latter statement has not been confirmed.