I feel about this the same I felt about the GTX 460 from nvidia when every argument was "hey, but if you OC it..."
My stance then was if nvidia wants the GTX 460 to have that level of performance, they should clock it accordingly or have a different SKU (560, evidently.)
If AMD has a 7970 Black Edition that runs at those clock speeds out of the gate, then great, let's take it into account.
Believe it or not, there are plenty of people for whom buying an add in video card and getting the drivers installed and working is the achievement. Perhaps this is more true in the south of the $200 segment than in the stratosphere where the 7970 currently resides, but I think the point is still valid. The only guaranteed clocks for the 7970 are the ones it ships with. Replace "7970" with any card you want and I'll stand by it.
That's just my 2c.
Ok. Sounds pretty good. If 28nm o/c's this well, then I'm sure nobody would mind any SKU comparison when o/c'd later on.
I understand your point, guys. The 7970 sold as it is, has 925MHz clocks, yes. A person shouldn't have to look only at benchmarks with overclocked cards and be forced to overclock the card to get such results.
Let me evolve my argument.
Let's ignore the 7970 itself for the moment, and talk graphic cards in general. Let's consider the fact that what we're trying to do here, is have a valid idea of how all of these products compare to each other.
The generally accepted stance is that graphics card should only be compared at stock clocks, or overclocked to overclocked. You're right in saying that comparing an overclocked card to one that is not is wrong, and I see where I went wrong here. My point instead is that the ease of overclock should have a much greater value. This is where the 7970 comparison to the 580 comes in. Because the ease of overclock that the 7970 is showing is uncanny at best, and unseen at most, it should be not only considered, but also valued. That is why comparing both of these cards at stock clocks is a moot point to me. Yes, it might not be the case for the user who just wants to buy the card and use it as is right away, but to me and many others, the fact that the 7970 can be overclocked so easily and so much has an obvious monetary value.
Thus, the question is: who are benchmarks made for? A typical reference vs reference benchmark would be made for users who don't necessarily want to overclock, and therefore the performance/$ shown is valid. However, for people who will surely overclock whatever product they buy, it's not necessarily valid. This is especially true in the case of the 580 vs 7970, as there is a huge discrepency in the overclocking ability of the cards. The difference between the two cards at stock clocks is less than the difference between the two overclocked.
Therefore, to me (and others I'm sure), a stock to stock comparison is meaningless. True enough, that doesn't mean that the overclocked 7970 should be compared to the stock 580. However, what it does mean that in itself, it's a useless comparison, as for overclockers, it doesn't represent true values.
(If I'm going wrong somewhere, tell me )