Originally Posted by Mygaffer
You've "drank the Kool-aid" my friend. The whole "disappointed with 7970 performance" tack that Nvidia took is part of campaign to mitigate any losses from not having a competing product for several months after launch. They lost a lot of ground when the HD5000 series was out and Fermi was no where to be seen, that is why we have had all the "leaks" saying performance was going to be waaay beyond the 7970, that Kepler was "unbeatable", and why they decided to make this the GTX680.
Most likely we are going to see a GTX680 that trades blows with the 7970, beating it by small margins in TWIWMTBP games and performing closer to a 7950 in other games. As far as it being a "$500" gpu, this die is so small that even if they had yield problems they will still be making a lot of money on these cards. Performance and market demand are what drive price here more than the cost of the chips.
There's no need to get personal.
I didn't buy the whole unbeatable story (it even had facebook as source lol) but the fact is that the performance of the 7970 wasn't that overwhelming considering the performance of last gen GTX 580. If the 7970 would have been faster i.a.w. faster than 6990 and GTX 590 it would have been out of reach. They only can battle back with GK104 and that is what they are doing now. And according to leaks from Expreview which are pretty accurate most of the time the two are trading blows with GTX 680 taking a slight lead most of the time.
Fact is that they anticipated the competition's flagship card with GK100 but nVidia failed in making it a decent/finished product. So it does come in handy that now they know 7970 performance, they can counter it with their GK104 mid range part and sell it like it's a high end GTX 680. No Kool-Aid needed for that, thank you.
Originally Posted by grunion
Well of course, not the point.
I'm what iffing, what if the stock speed was 1.1-1.2, the conservative clocks may have saved us from extended price gouging from AMD.
Now NV can get a card out the door to at least compete while working on Big K.
Oh back to my point, would NV have sat on GK104 if it could not compete with a 1.1-.1.2 card?
Or went on and released it at $349-$450?
Let's face it, at 256/2gb of memory this card should not demand a $549 premium.
I think you misunderstood me on this one. If you read back a few pages you'd note that I made a similar statement as you're doing now. I simply meant that assigning a higher clock target at an ASIC means that you'll get a lower effective yield, because the clock skew in between the parts will be even greater.
However going by your point and assuming 1.2GHz was a nominal clock -- putting the performance 5-10% above GTX 590 and HD6990. In that case the HD 7970 would have probably cost ~$649- $699 and GK104 would not be able to compete and probably needed to be sold according to Pitcairn's performance, I think the estimate in your post would be a good one (~$359).
If you ask me if I find this pricing okay, of course not. But I already anticipated that this would happen when I saw the price tag of the 7970 and 7950 at their performance levels. I actually hoped for GK104 to be released as a GTX 660 while being 5-10% slower than a 7970 so that a price war would occur, but it didn't. After all, if nVidia did release the real GTX 680 based on GK100 back in December we wouldn't be paying $549 for a "GTX 660" and 7970. So no these cards should not demand this premium.