Originally Posted by Benetanegia
It makes no difference really and cards are still made oflots of small power of two chunks. The 384 bit memory controler is really 6 x 64 bit memory controlers each controling one memory module so there's your power of 2. Shaders both in AMD and NVidia architecture are composed of 16 shader wide arrays, SIMDs, which is what really does the hard and fundaental work, so power of 2 again, TMUs and ROPs are typically clustered in groups of 4 or 8... but really it makes no real difference. It's like that for convenience, until I hear the opposite. Rendering typically works on quads of pixels, 2x2 or 4x4 so that's why they tend to make it that way for GPUs. Other than that there's no reason that I know of.
ok i just came across these older posts in the forum while going through it, and there is a few things to note or take into consideration regarding the 256bit memory controller
1-gk104 was meant to be the mid/high range card and not the high end
2-usually cards like the gk104 might end up in the mobile segment therefore must be built with both worlds in mind(tho im not sure whether nvidia uses the second fastest desktop chip for mobile or the third fastest so correct me if im wrong)
3- considering the fact it was designed to be the med/high there are 2 reasons on why nvidia would choose 256bit, the first reason is to purposely limit the performance to allow for faster and more expensive cards to be released later or to place the card in the preferred place they desire in the market in terms of performance and price (similar to what they did with gtx460 768mb and gtx460 1gb) the second explanation could be that the card doesnt benefit from more bandwidth and would only make it more expensive for minor gains.
4- we dont know whether nvidia will call the gk104 gtx660 or 680, and if they do call it gtx680 is it because they failed to release gk110 due to yield issues or is it because the gk104 is sufficient to compete?
if it does end up to be gtx680 it would probably be the first time nvidia would have smaller die sizes than amd, tho after all the issues they had with fermi and manufacturing such a change in methodology is not all that shocking
but overall nvidia seems to have learned alot from amds strength(i wish amd would do the same from nvidia) as amd(the graphics division strictly) had more experience with facing manufacturing difficulties and knowing what to expect, and used to work according to due dates,i read an article by one of the chief engineers at ati where he explains how things work at ati(i will try to find the article and post it up, it was about rv670 and how amd jumped the wagon first for 55nm and how the process of releasing products works)