That's a really handy chart, and something I forgot MikeC does. It could always use some more websites, if the idea is to find a consensus choice, so I tabbed the #s from ComputerBase.de, Digit-Life.com, Hardware.fr, Hexus.net (only at 1920x1200), and ZDNet.de (they had some "obscure" benchmarks).
MikeC's literal interpretation of these new sites yields an XTX record of 40-30-2. Considering every difference under 5% a draw, I came up with a record of 36-19-15 for the XTX: 36 wins, 19 losses to the GTX-512, and 15 "draws" (10 rounded down for NV, 3 for ATI, and 2 literal ties). If I wanted to be really strict and say only 10%+ counts as a win, that's 27-12-32 W-L-"D" for the XTX. Everything under 15% a draw? XTX is 21-10-42. Yep, a whopping 42 "draws."
But I'd be hard-pressed to consider 4% at even 100fps, or 4fps, a noticable win. 4% at 30fps is just 1.2fps, probably equally impossible to notice. Heck, even 10% seems a low bar for a win at either 30 or 100fps, meaning the faster card would be running at a breathtaking 33 or 110fps. Buckle up for whiplash, kids!
You can see a general pattern with the new sites: the higher I raise the bar for a (IMO) clear win, generally the lower a percentage the GTX-512 wins. I didn't intend that to happen, especially with the persistent "fanboy!11" vitriol, but I've mulled 15% as a minimum win differential for quite some time. Yeah, it's one more way of deciding for everyone else, somewhat like a HardOCP or Bit-Tech review (or any review that doesn't painstakingly compare IQ, really), but the raw #s are there if you need 'em. (FYI, that 15% bar is also why I don't really bother with OCing GPUs, as it's the rare card that nets you more than 15% return for your risking the warranty. If IHVs/AIBs tried to separate their cards by 15% performance--be it with clocks or pipes--it'd sure as heck make our lives easier when it comes time to buying new cards. )
Combining my tally (XTX is 40-30-2) with MikeC's (XTX is 43-38-2):
w/ literal draws: XTX is 83-68-4 (W-L-D), or wins 55% of the time. (Seems a bit low, no? I'm dividing W/L, just like I did to compare framerates.)
w/ <5% "draws": XTX 72-53-30 (w-L-"D"), or 58%. (This actually helps the GTX-512, as the XTX drops more "wins.")
w/ <10%: XTX 54-46-62, or 58%. (XTX loses 8 more, GTX-512 loses 7.)
w/ <15%: XTX 42-36-80, or 56%. (XTX drops another 8, GTX-512 drops 10.)
So, it's still fairly--surprisingly, IMO, considering R580's 48 pipes--close. I don't want to lose focus on price, so keep in mind that there's a >15% price premium for the GTX-512. Speaking of price/performance, I'd like to see a similar chart for the 1900XT and 7800GTX, but I'm not willing to make one (especially so close to the 7900). Even tho the cheapest XT is ~$50 more than the cheapest GTX, you *are* getting an extra 256MB, which might prove useful at >= 16x12 with >= 4xAA and max texture quality.
Boo-yah. Hope this helps. I can post a chart like Mike's, if anyone wants to see the actual bingo table.