The trick's I've seen so far, and this seems to be true in all the 3D software I use and CAD stuff....
1) the SECOND DIGIT of the Nvidia is the real one to watch:
=
a 7
900 GS is a LOT faster than a 8
400 GS.
Why?
The first simply designates the version of the process- a marketing ploy more than anything, but sometimes new things (version 8xxx is direct x10 compatible, the 7xxx series isn't- and the manufacturing process of the 8xxx series is a bit more refined, so they can bump up core speeds a tiny bit- but nothing noticeable in the real world yet, really)...
So a 6
800 GTX will rock a 7
600 or even possibly an 8
300-8
400 (regardless of faster memory and more pipes)!
And a 7900 will rock an 8600 no problem -it's got a much faster GPU on it!
BUT
a 7600
won't
rock an 8600, unless it is a GTX and the 8xxx is an entry GS or similar. HOWEVER, the newer 8xxx series has more features and is a bit faster than the older 7600...!
For sure an 7800 will toast a 6400, of any memory (GS/GT/GTX) type -
the GPU's are simply beefier the higher the second digit is
!!!!
This is great news for us budget-workers!!!!
(NOTE: the GS/GT/GTX means the type of memory they put on the card- GS is the norm/cheaper stuff, GT is the fast and GTX is a little bit faster/overclocked +stable. A 7600 GTX will compete or match a 8600 GS in real world performance
in almost all cases
except some benchmarks. in the REAL WORLD though they look the same, for now. Eventually the 8xxx series will be the bottom cutoff when looking for older gear in a few years, as the ancient-by-then 7xxx series won't do DirectX10. Get it?)
The ATI's run the same, but with the second digit ALMOST comparable to the similar 2nd digit of the Nvidia line...
I.E....
x1
400/x2
400 = 6300/7300-7400/8400
ATI seems to be struggling against the giant GF though... in the above example, the equivalent model is noticeably slightly less capable than it's GF rival. For instance I compared a x1400 against a gf7400 today on the same 10 floor GDL and window heavy model. I had to reduce everything for the ATI to get as smooth as the Nvidia (no highlights, emission, and the always display everything within...rendering down to 500).
So despite the touts of ATI being more OpenGL capable, perhaps that once was the case 3-4 years ago when Radeon series came out? I love what they're doing as a firm but it's going to get harder to keep up with the giant Nvidia in everything; simply ATI lacks the resources and budget to compete, and perhaps the OpenGL market/ CAD applications is one area where things are sliding a bit...
The
Quadro series is set up supposedly for faster CAD-based work in OpenGL. I'm not sure what the visual difference is in the real world, but the 2500 is similar to the 7900 Go GS from what I understand- even with it's double the onboard memory. This probably holds true across the board; the Quadro 340x-350x's will be the same as the 8900 GS/GTs in real-world working results, but at 2x the price. And BOTH will be even less noticeable of an upgrade to get the latest GPU's from last years x1xxx and 7xxx series than from AC10 to AC11.
Sooo, on that note...
(no really I do love AC11 but maybe we coulda skipped it till AC12?)
Looking at GPU's, you want the SECOND DIGIT the highest you can afford-
6900 GT/GTX, 7800-7900 GT/GTX and Quadro 2500's, as well as the ATI x
1600-x1800; but skip on the ATI's again for now from the reports I'm reading- Nvidia is problem free for sure.
This way you're not paying R&D for something that at 2-4x the cost gives you 5% better results.