It looks like you're new here. If you want to get involved, click one of these buttons!
Performance is competitive with a Radeon R9 285, while using considerably less power. The GTX 960 is arguably the best option for $200 right now if you don't care about power consumption, and definitely the best if you do care about power consumption. It's typically faster than a Radeon R9 280, while being cheaper than a Radeon R9 290.
With this launch, I'd expect the GK104-based cards (GeForce GTX 670, 680, 660 Ti, 760, and 770) that haven't already disappeared to go away soon. It's not clear how much longer the GK106-based cards (GeForce GTX 660, GTX 650 Ti) will hang around, but they're not really needed in Nvidia's lineup anymore. I'd expect the low end Kepler (GT 640, GTX 650) to hang around a while longer, at least until there's a Maxwell replacement for them--which may or may not be coming at all. GK110-based cards (GeForce GTX 780, 780 Ti, Titan) haven't made sense to buy for quite a while now, at least for gaming purposes, and seem to be mostly gone.
It's also interesting that the GTX 760 is based on the third Maxwell GPU chip, but the first to be aggressively priced. The GTX 750 and GTX 750 Ti have been very overpriced for their performance as compared to AMD's lineup ever since launch nearly a year ago. The GTX 980 is poor on a price/performance basis; the justification is that if you want top end performance, you have no other options. The GTX 970 is priced about the same as the faster Radeon R9 290X, and much more expensive than a Radeon R9 290.
There's still a big jump coming in GPUs in maybe a year or so with the move all the way from current 28 nm cards to 14/16 nm (probably TSMC 16 nm). That will apply to both AMD and Nvidia, of course. In the meantime, AMD could really use a more competitive architecture, as GCN was nice in its day but is really showing its age.
Comments
That is interesting, as nVidia hasn't made a huge attempt at being terribly competitive in the lower tier price brackets. They get in the same price tier, but historically they have never had an issue charging a 5-10% premium for their product, even when it didn't always make sense to in this middle and lower price brackets (I guess they can thank the Team Green Fan Club who always flock to nVidia no matter what for that)
Then again, I don't know if we have seen AMD's response, as AMD historically hasn't had any reservation with slashing their prices to remain competitive.
It's a nice card in itself. I'd like to see the price tag closer to $300 rather than $330 before recommending it on a price/performance basis. But it's not a terrible deal, especially if you care about power consumption.
AMD hasn't historically been at a cost of production disadvantage as compared to Nvidia, but they are today--and probably for the first time since they bought ATI. It's one thing to start a price war when you can build a widget for $10 and it costs your competitor $20 to do so. It's quite another to slash prices when it costs you $20 to build a widget and your competitor can do so for $10.
When's the last time that a GPU chip had a top bin for desktops, and then a next bin down that disabled some shaders, and then a third bin that disabled even more shaders, and that third bin was a decent value for the money? Has that ever happened at all? Ever? There are plenty of examples where that third bin existed and was a terrible value, such as a Radeon HD 5830, GeForce GTX 465, and GeForce GTX 660 Ti.
The problem is that there's already a GeForce GTX 970 with 13 of the 16 SMMs active. In order for a chip to make sense for use in a third bin, it has to be even more broken than that so that it doesn't fit the GTX 970 bin. If there are more than a handful of chips that fit that description for a new chip on a mature process node, then Nvidia really screwed up. And if there are only a handful of chips, you don't have to offer a great deal to sell out. All of those examples that I can give where the third bin even existed were AMD's or Nvidia's first GPU chip on a new process node, and that tends to lead to yield issues.