Unlike AMD/ATI, nVidia spent Computex Taipei 2009 in silence, not mentioning anything about the future of their hardware. That was to be expected, given the fact that nVidia was also silent during Computex Taipei 2008, and then briefing select journalists about Tesla D1060 and S1070 week after the show in Santa Clara, CA. Of course, this was followed by launching GeForce GTX280 and GTX260 the week after.
Thus, nVidia is keeping silent about their DirectX 11 plans, but don’t think that the timing of their "nVision Lite" [End of September 2009] is timed just because they want to talk about 15 month GT200 architecture. Our sources are confident that the part is on track, and that it will launch during Q3 2009. Again, according to the same sources, one of primary concerns was how to avoid all the thermal spots that plagued the GT200 architecture.
The GT300 part targets a thermal range of 225W and should feature two 6-pin PEG [PCI Express Graphics] power connectors, same as on the current GTX285 graphics card. The target was not previously rumored 300W, e.g. one 8-pin and one 6-pin PCIe. Then again, nVidia should hold off from opening the champagne, since ATI’s Evergreen uses just one 6-pin PEG. Dual Evergreen will probably use the conventional high-end arrangement off one 8-pin and one 6-pin PEG connector.
There you go. You have a 40nm chip targeting clocks of 700 MHz for the core, 1600 MHz for those 512 MIMD shader cores and nice 1.1 GHz QDR, e.g. 4.4 GT/s for the GDDR5 memory… sucking same amount of power as the actual GTX285. Expect Jen-Hsun and Ujesh to be all over the power features inside the chip, since the chip architects sweated blood over this one.