[citation][nom]youssef 2010[/nom]OK, Maybe you'd call AMD a slower GPU than the latest Nvidia has to offer. Maybe you'd criticize them for their lack of PhysX support. But then, you have to admit that they offered some technologies than Nvidia only offer in the 600 series like Eyefinity on a SINGLE board. I'm sitting here staring at my 1080p 23" display imagining the possible screen estate if I get three of them running in Eyefinity mode. AMD's price/performance couldn't be matched by Nvidia until the recent introduction of the GTX 680 and Nvidia had to pull the plug on any compute performance improvements. Also, AMD's tessellation is significantly improving with every generation while Nvidia is simply sitting there. Another thing is Nvidia's accelerated video encoding which is significantly worse than what AMD has to offer despite Nvidia's CUDA being SEVERAL years older than AMD's Stream /APP. Nvidia is also locking out the PhysX capability if it detects another GPU in the system. I can't understand this move since it will boost the sales of cheaper GPU which is a section dominated by AMD and most of graphics card profits are in the lower end of their portfolios.Last but not least, AMD never released a driver that fried GPUs. So, I think my money is way safer with AMD than it'll ever be with Nvidia.Please,my kind sir, look at The Best Graphics Cards for the money column before saying AMD's GPU's are a failure (that's also the right spelling for failure, not what you wrote)[/citation]
Nvidia's only faster because they were willing to sacrifice compute performance in an attempt to get us to turn to Quadro and Tesla. Furthermore, it's not even a big difference. There aren't many games where a difference between the 7970 and the 680 can be seen (at least in FPS). Granted, the 680 wins more often then the 7970 does, but it's worth every penny, just as the 680 is. The difference is that the 7970 has the memory capacity to last more than a year or two before AA/AF needs to be lowered to stop the VRAM capacity from getting overloaded, whereas the 680 already shows problems caused by it's low VRAM capacity for it's performance in some games and resolutions, settings, and AA/AF. AMD also has cards that can do six monitors in Eyefinity instead of just three and has had this for years. The list goes on, but I'll stop here before looking like an AMD fanboy.
Let's see what Nvidia did... They most certainly do have the most energy efficient and for the most part, the fastest single GPU and dual GPU cards in the gaming world right now. However, if games become more compute focused like so many say they will (including Tom's), then how long will that last? Well, that depends on whether or not such games can be released before Nvidia releases another compute focused architecture on their Geforce cards. I hope so because if not, then it would be a one-sided competition until Nvidia did. In that, I hope that the next generation of Nvidia cards have more compute performance just in case we get our compute heavier games soon. Nvidia does have one advantage in that with Kepler, the dual precision performance comes from cores that aren't related to the 32 bit gaming cores, so it should be able to keep it's regular performance unchanged when it adds in it's little bit of compute performance, whereas AMD will need to have some cores allocated to the 32 bit math and some allocated to the 64 bit math. If Nvidia simply adds more of the 64 bit cores, then it could have a winner in the next generation. Maybe AMD will take a similar approach in their next generation too by separating the 32 bit and 64 bit math into different cores. It definitely is an interesting concept, although I'm more partial to keeping everything in one for this so that if one type of performance is improved, it can all get improved at the same time.
It would be interesting if we could change the clock frequency of the 64 bit cores relative to the 32 bit cores. That way, we could overclock what really needs it more than what is already fast enough.
Facebook
Twitter
Instagram