Unfortunately there aren't a great deal of commonality amongst reviews (or reviewers). To get a good idea of overall performance you would need a large review db to work from.
You'll notice that reviews differ quite markedly depending upon what games are used, both in titles (some being ATI friendly, some TWIMTBP) and whether they use OpenGL, DX9, 10 or 11. Something else to consider is that some reviewers use settings either much lower most gamers would ever use (no AA/AF etc.) resulting in exaggerated fps, or maximum settings that produce bigger percentage variances (i.e comparing 4xMSAA in Metro 2033) at framerates that make the game unplayable.
As a system builder I tend to keep a database of current GPU's performance targeted for screen resolution, game engine and renderer for my customers- since many people will upgrade/buy graphics based on a particular game engine or game type. Frames per second is secondary in as much as usually just wants a definitive answer- Can I play it smoothly?
Taking the reviews from : Techspot, Tech Report, Guru 3D, Anand, Xbit, bit-tech, Hardware Canucks, PCGH, PC Perspective, Hexus, Tweakers.net, Tom's, Neoseeker, Hartware, OCC, Firing Squad, Computerbase, Legit Reviews, [H]OCP, Bjorn 3D, Tech Power Up, Extreme Tech, Benchmark Reviews, Motherboards.org, Inside Hardware, Tweaktown, Trusted Reviews, Legion Hardware, Ninjalane and Xtreme Systems (forum)...
Average fps.
DX9, 1680x1050/1200............HD 5870 (108.58fps).....GTX480 (129fps, +18.8%)
DX10.......................................HD5870 (71.45fps).......GTX480 (89.11fps, +24.71%)
DX11.......................................HD5870 (66.3fps).........GTX480 (75fps, +13.07%)
OpenGL..................................HD5850 (119.95fps).....GTX480 (125.6fps, +4.71%)
DX9, 1920x1080/1200............HD5870 (100.89fps)......GTX480 (116.69fps, +15.67%)
DX10......................................HD5870 (64.24fps)........GTX480 (77.4fps, +20.49%)
DX11......................................HD5870 (52.35fps)........GTX480 (59.34fps, +13.35%)
OpenGL.................................HD5870 (116.27fps)......GTX480 (108.17fps,
-6.97%)
DX9, 2560x1600....................HD5870 (70.9fps)..........GTX480 (80.25fps, +13.18%)
DX10.....................................HD5870 (44.98fps)........GTX480 (52.43fps +16.55%)
DX11.....................................HD5870 (37fps).............GTX480 (40.03fps, +8.17%)
OpenGL................................HD5870 (72.63fps)........GTX480 (66.13fps,
-8.95%)
As for the state of the play at this moment, you would need to realize that these numbers are only as good as the next driver revision, or in Metro 2033's case a bug fix.
The longer term depends on how much of a commitment to gaming AMD show-to benefit their own architecture (which seems to play fine with nVidia cards, the reverse is not necessarily true). If the development funds continue to flow from TWIMTBP, especially in AAA titles, then it's equally probable that nVidia sponsored titles will feature heavier tesselation options within DX11 to further distinguish themselves (at least in the enthusiast sector) from ATI's offerings. Add in the facts that whoever provides the dev funding is likely to have optimized drivers on game launch day (while the other is obviously playing catch-up unless the get a look at the code before launch), and most people do not upgrade their graphics solution every six months, then these are the factors that should influence a graphics purchase. If AMD follow through with funding then that particular playing field will level out.
Looking at the above figures it can easily be noted that the average framerates of both cards are well above the playable level for the most part...so after all these empirical arguments the choice still boils down to the subjective- For some buyers, the heat, noise, power consumption and price are offset (or are of little consideration) by PhysX, driver support on game launch day, generally better multi-gpu support, and I suppose for some, satisfaction in having the fastest single gpu. Likewise, for many consumers the price/performance point (which heavily favours AMD), noise (a valid concern), risk of heatstroke (?) and heat output/power consumption* have priority
* Something I could agree with if the person buying the card was also the kind of person that kept their car tuned, turns off the ignition when waiting at a drive-thru/pick-up lane/stalled in a tail-back and doesn’t crank the a/c when the temp gets 5º outside their comfort zone.
Even here in Minnesota, during the summer, the ole thermostat is always above 70F. so what happens when the dude from Calcutta,Perth, or Orlando buys one of these things?
Speaking from a place that can nudge 40ºC/104ºF in summer (with 95-100% relative humidity) I’d say that having air-flow through the house is a must in any event. Anyone that fancies sitting in a sealed room 24/7 gaming or running Furmark on loop probably needs a life more than they need a graphics upgrade. Worst case scenario I presume would be card failure > RMA > Get new replacement Fermi II (since the AMD fanboy scuttlebutt tells us that after the chips are used from the initial 9,000 risk wafers there will be no more production of GTX4xx –so no direct exchange replacement for bad cards).
HI guys,after reading this I don't care about Nvidia or ATI.I just want a graphic card that can deliver good frames while playing CRYSIS 2. I can't afford 5970,so which graphic card should I pick up?
2 x HD 5770