nVidia GTX 670 appearing in shops.

[Zebo]

Why don't they have that luxury? Fermi sold pretty well despite using a lot of power.

 

[aaksheytalwar]

And if you look at the Perf/w increase this gen, you'd understand that the gains are as large as ever but neither AMD nor Nvidia have the luxury of pumping twice the power in their cards anymore. You can still get the huge performance increases you're used to, just have to do it manualy by overclocking.

This

 

[SpeedyGonzales]

I don't understand this line of thinking. What was planned is absolutely irrevalent because for some reason or other Big Kepler is broken and won't be released without at least a base layer respin which would put it at a September release at the very earliest. The only difference is that is that instead of calling it the 670ti to save face, they are calling it the 680 because AMD lowballed the clocks on their cards to get better yields. Something that Nvidia is still seemingly struggling with.

And if you look at the Perf/w increase this gen, you'd understand that the gains are as large as ever but neither AMD nor Nvidia have the luxury of pumping twice the power in their cards anymore. You can still get the huge performance increases you're used to, just have to do it manualy by overclocking.

I do not take overclocking into account.

a) overclocking is a gamble (at least beyond 15%)
b) AMD and Nvidia can not count on overclocking, because only a minority of users are doing it.
c) we had very good overclockers in previous generations as well

For me it is very simple. Both GTX 680 and Radeon 7970 are not were they should be. Either performance is too low or prices are too high. And the experienced forum members have all spotted this and therefore recommend to wait (if you can).

 

[Arzachel]

If you're unwilling to tweak the hardware to suit your needs then you have to live with what the majority wants, and that's higher power efficiency. That isn't going to change with the next gen, nor the gen after. Big Kepler isn't going to be a miracle for gaming because it'll be mostly compute focused. And to my knowledge (since Gf6xxx series and what was it, rv400 on the Ati side?) There haven't ever been such gains from overclocking on the whole lineup of cards.

 

[SirPauly]

Good lord, always thought performance/dollar was the most important metric.

 

[Arzachel]

Good lord, always thought performance/dollar was the most important metric.

I personally agree, but if everyone thought like that, there wouldn't be such insane demand after Macbook Airs

 

[SpeedyGonzales]

If you're unwilling to tweak the hardware to suit your needs then you have to live with what the majority wants, and that's higher power efficiency. That isn't going to change with the next gen, nor the gen after. Big Kepler isn't going to be a miracle for gaming because it'll be mostly compute focused. And to my knowledge (since Gf6xxx series and what was it, rv400 on the Ati side?) There haven't ever been such gains from overclocking on the whole lineup of cards.

I agree, that Big Kepler's gaming performance will not scale linear with the die size compared to current Kepler and for a lot of people this will be a bad surprise.

You are probably right with the power efficiency argument. If you take this into the equation with probably 10-20% of the price still to be explained by the immature 28-process, it makes more sense. Let's wait and see, if people are willing to pay for power efficiency. Some will - for sure, but I think raw processing power was always easier to make people believe, they need the latest hardware (specifically in times of the "console syndrom").

 

[JTsyo]

Good lord, always thought performance/dollar was the most important metric.

Would you happen to be an engineer? There is a min performance criteria though. Would you take a free 450 vs a discounted 560? Seems 660Ti is in my price range but 6months is a long time, almost half a year... :'(

 

[entfy]

http://www.techpowerup.com/165252/MSI-GeForce-GTX-670-Pictured-Up-Close.html