"GeForce 9600 GT." The card we’re testing today is the very first GeForce 9! The first surprise is that this new generation of cards isn’t being introduced via the high end, as has always been the case, but via the midrange, mainstream price point. In reality, there’s a very simple explanation for this phenomenon, the architecture behind this card is the same as the one introduced on the GeForce 8s and hardly improved with the arrival of the midrange versions (the 8600 GT, then the 8800 GT, as has been nVidia’s custom since the GeForce 6 series).

A few weeks back we criticized AMD’s decision to call its cards based on the RV670 "Radeon HD 3000," but nVidia’s attitude is even more reprehensible given the extremely short list of differences (no new API supported, no reworking of stream processors) between this GeForce 9600 GT and the GeForce 8800 GT. After GeForce 7800s that, at the time, didn’t really seem to deserve the name GeForce 6900, the nVidia marketing department has gone even farther and seems to have taken us back to the darkest days of the GeForce 4 MX. It’s a disturbing sign of the new slowdown in evolution and technological competition the two manufacturers seem to be engaged in. Still, let’s give credit where credits due –the bang-for-buck ratio of the midrange solutions that have emerged since the end of 2007 is exceptional.

The G94: Half a G92?

Unlike all the GeForce 8800s introduced since the [link to your 8800 GTS 320 MB article here] early in 2007, the GeForce 9600 GT is not based on the famous G92 chip, but on its little brother, the G94. The essential difference? The G94 only has half the number of stream processors the G92 had – 64. You might well think that it pales in comparison to the 112 a "simple" GeForce 8800 GT 256 MB boasted. That’s not untrue, but the shader frequency is higher (though only compared to the 8800 GT, since the G92 on the 8800 GTS uses the same frequency), attaining 1625 MHz for these units (650 MHz for the rest of the GPU). But even at that frequency, the raw processing power of the G94 is still a full 38% lower than that of the 8800 GT.

Let’s look at the reasoning behind this choice. It’s often more interesting for a card maker to attain a given level of processing power by using fewer SPs, but at a higher frequency, rather than the other way around, because it lets them reduce the number of transistors on the chip, and consequently its size, thus reducing production costs. Of course, for that principle to work, the required frequency has to be able to be achieved without a severe drop in yields, in which case production costs can end up being higher despite the increase in the number of dies produced per wafer. Accordingly, the G94 has only 505 million transistors (that’s 33% fewer than the G92) and it has a surface area we measured at 225 mm², or 31% smaller, despite the same engraving depth (65 nm). That value is still 15% higher than on the RV670 used in the AMD graphics cards with which it will directly compete; but remember that the latter uses 55 nm given its higher number of transistors, and that that necessarily has a financial impact – either at the time of development of the process for the chips, in terms of yields, or in terms of production costs.