This is an odd about face by nVidia. They are on record as stating the RSX is faster--with the caviat that they would have faster GPU's on the market by the time the PS3 is available.
As for "slightly less powerful" all one can do is shrug. That is what one would expect nVidia to say, yet their slides from E3 clearly indicate that the RSX is more powerful than a 24 fragment pixel pipeline 7800GTX--even clocked at 500MHz. A 550MHz RSX, from the chip stand point, is going to be better than a 500MHz G70. I am not sure anyone would argue against that. So chip-for-chip, RSX is better. That is Strike #1.
From a system standpoint, well, I would give a nod to a 350MHz RSX in the PS3 over a 450MHz G70 in a PC simply because a closed system is going to make better use of the hardware with software designed with it from the ground up (versus the broad spectrum of support on the PC and the DX API). I think customers with PS3's will see more out of the RSX from a graphical perspective than a PC owner every will with a G70 in their box. From a consumer experience perspective, I believe the PS3 with RSX will do things that the G70 in the PC never will be capable of with the same experience just because of the platform it is used in. One may not call this power, but I personally would define power as what the end consumer sees; and in that scenario the RSX in a streamlined closed box market will dance around G70 because it will be more effecient and programmers will get more banf for buck from that design. Strike #2.
On the other hand G70 has, in real world scenarios, more memory bandwidth. Devs will optimize and work around this as much as possible, but there is no getting around the fact G70 has a ton of bandwidth to do MSAA or HDR (but not both, which is in a way good because MSAA+HDR would be very poor in performance). And the nVidia BS slides comparing G70 and Xenos and the "free AA" were a joke (as discussed before) because their focus was on CPU limited games--and yet even then picking out the new games like Doom 3, FarCry, and HL2 clearly showed significiant hits to performance.
Bandwidth is an issue on G70 with high levels of AA and AF, so I would expect this to be more pronounced on RSX. Not a ton mind you, but 1080p with high levels of AA (as NV was trying to suggest in some PR slides) is not very realistic.
So maybe NV was trying to spin this angle?
Another angle could be th 7800 Ultra. Still a 7800-series card, a 32 pixel fragment pipe part at 450-500MHz would be more powerful than RSX. Even more, it would be "slightly" more powerful, as NV is supposedly claiming.
So if nVidia is comparing a 7800 Ultra to the RSX then their claim may be true.
But is this not all irrelevant? Obviously NV does not want to give the impression that the new consoles are better, technically, than the PC--that is bad for business in many ways! Yet they already did... So why is everyone surprised that nVidia is backpeddling?
Anyhow, it is a moot point. RSX is going to run circles around G70 because the RSX is in a closed, effecient platform with specialized, streamlined, and optimized software is developed specifically with its strengths in mind. GPUs are very frequently about BRUTE POWER in the PC space. They are an open standard with ineffecient APIs and diverse hardware. The markets are almost impossible to compare--and for that reason comparing the two chips is almost pointless. They work differently within the design philosophies in their respective platforms. Basically the same use, but completely different approaches.
And for that reason comparing the two as competitors is kind of pointless IMO.