Originally Posted by Ironcobra
But are gpu makers helping to cover up laziness by not giving enough texture memory or bad optimization, as memory is very cheap nowadays?
I don't think so. I think it is actually quite hard to implement enough memory while keeping the card under a certain length. Larger PCB makes the card much more expensive. Memory units have expanded to be used by both sides of the cards, but I think the larger memory units are a lot more expensive then what we would think. Think about it like going from a 2GB stick to an 8 GB stick 2 years ago and their price difference. They just don't make enough of the larger memory units that would make the cards super cheap. In part that is due to demand I would think. Either they make a larger card that increases cost while using the same sized units, or invest in larger units and place them on existing cards.
Furthermore, the current gains over more memory gives you shows how much of a bottleneck exists with the actual core performance. 1GB to 2GB offers maybe 3-5 fps on pretty large resolutions. If you GPU can't fill the memory fast enough, where's the point until you develop a much faster GPU? It's like shoving 2GB of memory on a GT220. The GPU just won't be able to fill (write) or use (read) the memory fast enough to offer any significant gains.
I'm not saying the same thing as the "why need a quad core when dual cores when games only support 2 cores" mentality. I'm just saying there could be a tech bottleneck that limits the gains of using such large amounts of video memory and that could be another reason why we see such lack of over 2GB per GPU cards.
I dunno, that's what comes to mind anyways.