While I agree ATI have dominated this time around, I still can't get the mindless hate for nVidia at points, I will agree lately they have shot themselves in the foot several times, but there just seems to be a mass bandwagon (OCUK GCard forums used to be like this especially) of people mindlessly slating Fermi over issues that are either minor or just not there.


The heat issue, admittedly that's a half baked fan profile in the GCards BIOS causing the fan to only ramp up at v high temperatures where the heat can't be stopped, but I LANed earlier this year with Nikolii and had his 480 exhausting onto my legs, and with a custom fan profile and OCed (I think it was to around 830Mhz core) it just wasn't that hot or that noisy. Hell my 280 was chucking out more heat and noise. Plus, bear in mind the actual physical size of the GPU Die in fermi's case (529mm^2), it's significantly larger than Evergreen (334mm^2) with 3.2 Billion transistors in a GTX480 compared to 2.154Billion in a HD5870, so naturally it will run slightly hotter, plus you can't expect next-gen performance hardware to take the crown (Ok, tiny marginal amounts but still there) and still run very cool can you?


Also, note how the SLI scaling of Fermi seems to be largely forgotten about along with how well it scales when OCed.


Prices I'll agree are on the steep side, although I'm pretty sure (Although no source) that's down to GloFo's yields on the 40nm process (Again no source but iirc ATI are using TSMCs 40nm process, which they cleverly tested with the "pointless" 4770). (I could have the chip manufactures mixed up here, will do some research later when I have more time and edit with correct facts and sources)

Power consumption imho is moot, as if you have a rig that would normally accompany such cards (You're not exactly going to stick a GTX480/5870 in with an ULV CPU are you?) then I should imagine power consumption isn't exactly a main concern for you (At load anyway.)


I won't throw in the PhysX and CUDA arguments, mainly as PhysX, and frankly any GPU Physics API, is useless due to consoles (Although nVidia are quite stupid with the restrictiveness of PhysX I'll agree on that front) and CUDA isn't really an argument to be used on a gaming grade card, if you're looking at doing something seriously that involves that (Other than distributed computing e.g. F@H which many people run at home for reasons such as family etc) you'd most likely be comparing Quadro vs FireGL as apposed to GeForce vs Radeon.




And before I get accused of being a fanboy, I've already said that my next card will either be Southern Islands or a cheap 5870 (When SI are released), although GF104 and the GTX460 look like OK value for money but I fancy giving the red team another go . Plus the last build for a friend was completely AMD/ATI based, and the last person to ask me about GCards was given a 5770 as a recommendation.