Looking to build a computer

2GB is better as games are now coming out that use 1GB of vram already, and this amount will only increase in the future. With next gen consoles launching next year there will be a big jump in vram consumption of games due to the new higher capacity that the consoles have. Right now we have been sitting around 768-1GB of usage for a long time because the current generation of consoles has been holding us back for far too long. Games are developed for console and ported to PC now days, so whatever the console has will be the limiting factor. Companies rarely release higher res content for PCs anymore.

Cores are used when they are needed, and anything not needed sits there idle. The more cores you have, the greater the card's capacity to process data. If you have less cores than is required to process something then you take a performance hit.
The faster the cores, the better performance you get. But it also matters on the design of the cores and other parts of the GPU. The "cores" are really programmable shader processors. Nvidia's design is very simple, 1 shader = 1 shader. Each thing that needs to be processed takes 1 shader core for each little part. Doesnt matter is it is a simple or complex calculation. AMD on the other hand uses very complicated shaders, each shader is made up of 3 simple and 1 complex shader processor (if I am remembering right). So lets take a 1536 shader graphics card as an example here. Really it only has 384 shader processors in the card. The difference is that AMD can process 4 things per processor core compared to Nvidia's 1 per core. This can be a very good thing, or very bad. On the one hand, each core can do lots of calculations at a time and group things together for high efficiency. But if a game requires lots of complex calculations and very few simple ones, then many cores go unused and the efficiency of each core drops by a lot. Also if the game is not designed well and it only uses 1 or 2 calculations per core instead of filling it up, then many cores are used when not needed while others have empty space, thus raising power consumption due to bad coding. This is why games designed for Nvidia perform really well on Nvidia cards and AMD always struggles, and games designed for an AMD architecture excell and Nvidia falls behind.

I do not know if Nvidia switched over to a different style more akin to AMD or if Nvidia kept the design that they have always used. But in their latest 6 series cards Nvidia massively increased the amount of shader processors they have, and also dropped the "hot clock" they have always used. This was the shaders core speed = 2x the actual core speed. This prevented Nvidia from having a lot of shaders because it would consume too much power and be too difficult to build, but made each shader core VERY fast. Now Nvidia has synced up the shader core to the graphics core and this allowed them to put a LOT more shaders on the card.



You really cant just look at numbers and be able to say the performance of something. You cannot compare between different architectures on spec sheets alone, and you cant compare between companies just by looking at the spec sheet. Nvidia had much faster shader processors, but they do not any longer. AMD has much more advanced pixel pipelines, but Nvidia overcomes this by having more. Nvidia has a little bit better texture mapping units, but AMD gets around this by having more TMUs. Nvidia increases memory bandwidth by using odd bus sizes slightly larger than average, but AMD overcomes this by using faster memory.