Nvidia Corp.’s new flagship graphics processing unit (GPU), which is due to be out sometimes in Q2 2009, is projected to feature substantially, but not dramatically higher computing performance compared to predecessors. In fact, the new code-named Nvidia GT212 chip may be suited very well for high-end dual-chip graphics cards, which may mean that Nvidia is shifting to multi-GPU configurations from “megachips”.
Nvidia GT212 (G212) graphics chip will feature 384 stream processors, 96 texture units and 256-bit GDDR5 memory controller, according to Hardware-Infos web-site. The processor will be made using 40nm process technology by Taiwan Semiconductor Manufacturing Co. and will contain about 1800 million transistors amid roughly 300mm² die size. It is rumoured that the product will be released commercially sometime in Q2 2009.
It is noteworthy that the GT212 chip has 34% smaller die size amid 28% increase in transistor count compared to GT200b chip (55nm version of that GT200/G200, which is also sometimes called GT206, G200b or G206) and is nearly two times smaller than the GT200 processor from Nvidia.
Nvidia is about to release its dual-chip GeForce GTX 295 graphics card with 480 stream processors in total, therefore, single-chip GT212 will hardly be able to achieve performance levels of this board without truly high clock-speeds.
If the information regarding the number of stream processors, die size, transistor count and positioning of the GT212 is correct, then the new GPU from Nvidia might be designed specifically to power high-end dual-chip graphics cards. Relatively small chips with moderate transistor count are ideal for multi-chip graphics cards as they do not require a lot of power.
Considering the fact that Nvidia could pack more stream processors into the new processor and create substantially larger chip with much higher performance, but decided not to do it, it is highly likely that the GT212 is made to compete in performance-mainstream market segment in single-chip configuration and in high-end market segment in dual-chip tandem.
Nvidia has always said that single-“megachip” configurations are more efficient e in terms of performance compared to dual-chip graphics cards as modern multi-GPU rendering techniques have a number of drawbacks. On the other hand, Nvidia has also said that provided that it has appropriate GPUs for dual-chip graphics which can provide truly high performance, it would release such products. Meanwhile, ATI, graphics products group of Advanced Micro Devices, has been saying that creating smaller chips for performance-mainstream segments and making dual-chip enthusiast-class graphics cards is more efficient than producing megachips for high-end market.
Nvidia did not comment on the news-story.