I use my machine for graphics work using AutoCAD, Solidworks, Pro-E, etc. No gaming at all.
I am building a new i7 machine and am looking at graphics cards. Due to the cost I cannot consider the new Quadro 4000 (Fermi) and up cards. My less expensive alternatives are to use an older pre-Fermi card the Quadro FX 4600 or something like the GeForce GTX 460 (Fermi) 2Win 2GB 512-bit card. The first is built for the graphics I am doing and the second for gaming.
How does one go about trying to judge an older card versus an newer, but different application card? I guess maybe the first question one should ask is what is the difference between a gaming card and a dedicated graphics card? And how well does a gaming card handle my kind of graphics?
There really is no difference between the two cards except the drivers and a small string on a chip in the card. There is a program out there that will take your average GeForce card and turn it into a Quadro and back again. At the hardware level, they are the exact same thing. You pay a premium for a Quadro to get the application specific optimizations.
It's up to you whether you are bothered by the ethical issues raised by converting a cheap desktop card into a more expensive one without paying. Some of that money obviously goes to developing the application specific optimizations, so if everyone just converts desktop cards, pretty soon nVidia and AMD won't bother making the workstation drivers, or you'll have to buy them separately.
I advise against converting a gaming card. You will have driver recognition issues. Drivers are the real issue anyway. Gaming drivers and professional drivers are 180 degrees out of phase. There are currently seven Quadro boards on Newegg under $150. One of them is 600 which I am currently using to my satisfaction. You can't really buy a gaming card any cheaper.
Many CAD programs, including Soldworks, will bitch at you if you don't use a workstation card.
There are at least 4 "big" differences between the workstation and gaming cards
1. The workstation cards are statistically less error prone due to higher binning on the chips.
2. The firmware on the workstation cards is designed to complete the entire rendering process regardless of visual quality. Gaming cards can sometimes approximate or terminate rendering early in order to sustain performance. Additionally, the Double Precision Floating Point performance is capped at 25% of its theoretical performance on the Fermi and Kepler based gaming cards for marketing reasons. The Fermi and Kepler based workstation and Tesla cards are uncapped.
3. The workstation drivers contain application specific driver optimizations that the developers paid good money for. I've run SolidEdge on a gaming card and noticed visual anomalies that were not present when workstation cards were used.
4. Workstation cards can be equipped with much more SGRAM than desktop cards. Fermi and Kepler based Workstation/Tesla cards have ECC SGRAM which helps with reliability in CUDA/OpenCL/DirectCompute applications