The average gamer doesnt even know what CUDA is (i dont even care), i just want to load up tf2/ gta iv and play at a high res/ with high settings at a decent framerate.
Im no ati/ nvidia fanboy and ill buy what is the best bang for buck at the time, i dont see the point in paying more than 20% more for a product if it doesnt deliver 20% more performance..
At the end of the day, competition between the companies is good for the end consumer
(Awaits gt3xx release to decide on what dx11 card to get and rig overhaul)
And there's nothing at all wrong with that, and your right. However that doesn't change the fact that CUDA is awesome and easily worth (IMO) the extra the cards cost.
It's not really that people don't understand CUDA that bothers me, it's more that people don't understand GPUs. They just want them to play games (fair enough) but on average they haven't got the foggiest that unlike the past, nowadays the GPU is often by far more powerful than the CPU itself.
The neat thing about CUDA is how simple it is to massively accelerate standard programming tasks, it's just that at the moment they're mainly used to address graphics tasks.
The difference is when you buy a graphics card you care about it making Crysis run nice and smoothly, when I buy a graphics card I'm buying it to massively increase the speed of a lower level operation.
Now you might think "Yeah but why do I care about programming and all that lark" - well the answer is that when some bloke writes a CUDA program to accelerate something you DO use (compression, audio transcoding, image processing etc) you WILL care about CUDA and more generally about GPU processing, as those little benchmarks you all run will suddenly get faster, A LOT faster.
The only way in the past to get performance anywhere *near* a CUDA capable card was to have racks upon racks of hardware.
Quite simply GPU computing, whether it be openCL or CUDA is massively important, and whilst it might only be affecting your games right now, it's only a matter of time before it's affecting photoshop etc..........
So to give it a real world spin. Say a CUDA card is an extra £50, but with it you know you can write a program that'll transcode a video from one format to another in 1 tenth of the time it does normally - that £50 starts looking a lot more attractive, then what about if you can decrease the time it takes to compress and uncompress data by a 90%, looking even better value now?
This is the realisation of major academic institutions - they get it, they get CUDA and openCL. Basically (as Nvidia are now saying) with a couple of CUDA cards in a box you essentially get your own personal super computer - now that is impressive - if you don't get that, then frankly I give up!