Let's cut this one up:
Why do you expect Maxwell to be automatically faster than GCN 2.0?
I expect it to be faster because Nvidia has been consistently outperforming AMD in the high-end segment.
Naturally, there's no natural law that this will continue but the architecture of Maxwell, that we saw in GTX 750/750 Ti, leaves massive room for performance increases, especially when it comes down to 20 nm. I'd love to see AMD surprise all of us, and indeed they are more efficient on die area, but if we are to see a massive reversal in fortunes, I'd be the first one to celebrate. It's never good when one company has the performance crown for too long, as we've seen in the CPU space all too well.
As for $500+, people have shown a willingness to pay that much, the 7970 when it was launched was popular, as was the GTX680 at that price. Both are basically small mid-range dies. People were also happy to pay much more for the real big die, ie. GK110 on Titan and 780.
The top tier cards can command a high price because enthusiasts are willing to pay a premium for performance. Don't expect it to be better on 20nm or 16nm.
I think you misread the original post. I was referring to Saylick's assertion that high-end cards used to cost 500 dollars or so; now they are at 700.
And secondly, the question isn't if people
at all are willing to pay more than 500 dollars for a card. The entire debacle of GTX Titan has proved as much. I wrote that
most people are not willing to pay as much and you're more than welcome to try to refute that claim
But even more importantly, and the part that I think you missed, is even within the segment who can pay 500 dollars or so for a card, the higher up you go, the fewer people you have left that are still interested.
I don't think there's anything controversial to the notion that a lot more people would be on GTX 780 Ti's if they were costing 500 dollars a piece. And the whole conversation between me and Saylick was why it couldn't be like it used to be. I mean we're not talking about some ancient era. 2012 wasn't that long ago. Even accounting for inflation, 700 dollars isn't even close to a logical step from 500 dollars.
And this is the reason why I spoke about how Nvidia's revenue in the last three fiscal years has been basically flat, and add to that the massive disaster that is their Tegra business.
As for 16 nm and 20 nm, the cost savings aren't going to be massive, to put it mildly, but a bigger discipling factor will be how well AMD does. If it going to be growing their revenue by 30+ % next year and the year after that, they will force Nvidia's prices down.
I think you're missing the financial picture aside from the technological picture, in trying to explain the reason why prices has jumped as much. Especially as costs on a mature 28 nm process are lower than they were on an untested 28 nm process a few years ago. This shows the limit of that analysis.
Facebook
Twitter