Haha, the 7950 GX2? You do know that you need a SLI board to run that card, right? Probably one of the most useless things made by nVIDIA. First, they have TWO cards sandwiched together like no other, so the heatsinks hardly get any air (especially the middle two when put together for Quad SLI). It's not THAT fast, bascially a 7900 GTX but slowed down a tiny bit. Two X1900 XTX's in CrossFire can almost beat four 7900 GTX's in Quad SLI. Then again, it's mainly driver problems. nVIDIA has been having so much problems with it's Quad SLI drivers lately. If you're actually foolish enough to buy one, wait until the drivers get better. Prices haven't been released for the 7950 GX2, but expect each card to be in the $500-$600 price range. At least it's better than the 7900 GX2's which cost almost $1,000 per card.
I would wait for nVIDIA's G81 and ATI's R600 instead. Those cards support DirectX 10, meaning you can run Windows Vista efficiently and get all it's benefits, and it's also backwards compatible with DirectX 9 so if you still use Windows XP, there will be no problems either. No words have been released about the G81 so I won't say much about it, but there are rumors that it'll use 48 pipelines and carry the traditional independent shader processors unlike the R600.
However, there have been some reports from X-bit Labs and other sites that the next ATI core, the R600, is claimed to be "the fastest DirectX 9 card in the world". It's reported specficiations show it too. There will be 64 Unified Shader processors and 16 texture units. Clock speed will be guaranteed to be 650mhz or higher, no less. Memory speeds are unknown, but it will be using GDDR4, so expect something fast. Keep in mind that the R600 is very unique and isn't like any video card today. However, it is very similar to one GPU. The Xbox 360 Xenos GPU. The R600 will use the Unified Shader Architecture like the Xenos GPU, but the Xenos still has some advantages like it's 10mb of DRAM for buffering uses, and things of the sort that ultimately make the Xenos one of the most efficient GPU's ever made.
So, you ask what is the Unified Shader Architecture? Most traditional cards use the pixel shader and vertex shader combination where they are independent and work in their own ways and this causes a lot of inefficiency because some pipelines won't be used and it just remains unused until needed again. However, the Unified Shader Architecture is essentially just the number of shader processors the GPU has, like the R600's 64 for example, is basically 64 pipelines. However, each pipeline can be used as a pixel shader or a vertex shader, so there will be no "hanging" shaders that won't be used. So efficiency is increased and speed is as well.
Bascially, the R600 will blow the 7950 GX2 out of the water, heck maybe even Quad SLI'ed ones too.