Well there can be a difference beyond the OC, granted it's still a very limited region that benefits. Those general llittle boosts in other tests should be noted with the OC, but let's not dismiss the areas that do benefit significantly: like in the case of the Crysis high res/AA results, you have to admit, it's very unlikely that that big of a jump is due to the relatively minor OC in comparison.
But like I mentioned very limited areas where it would be worthwhile considering. Even in the biggest difference it needs to be noted the mode because in DX9 it does make it more into the realm of playable 27 vs 15 fps, but in DX10 we're talking 15.5 : 8.5 where 15.5 is much more, but just on the cusp of useable for some and without question unusable for others.
The thing that is interesting is that it's only a small 2 fps drop for adding AA to the 1GB model, whereas the 512MB model's performance is more than cut in half by the addition of AA with an 8.5 fps drop, where it was pretty close before AA was applied.
Anywhoo, it's still just one set of benchies, but it does confirm what other tests showed at higher resolutions when the smaller VRAM cards dropped off.
It's too bad they didn't include the HD4870 1GB in CF (not the X2 to avoid any design/bridge benefits-drawbacks) against the HD4870 512MB in CF just to see the impact. Having the HD4850 in CF is enlightening, but seing it often lose to the single HD4870-512MB in crysis puts more focus on memory bandwidth to me than VRAM size. In the other titles, usually the HD4850FCF is outfront.
It definitely takes it out of the realm of theory and makes it actual, but still debatable on the practical value.
To me it still comes down to the value, where if it's a small premium hey definitely it's a nice insurance policy for future titles and implementations, etc. But if it's a big premium, I wouldn't bother unless I specifically knew I could make major use of the benefit.