Hyper memory was not the tactic that worked well in the graphic world. Nvidia came out with an idea to have a card with 16MB of physical memory and to address additional memory from your system memory and simply brand it as 128MB or 256MB, depending on the chip. Nowadays you can get a hyper memory card that will get close to 1024MB.

The market wants more memory, and 1024MB graphics card has a higher chance of success even if it its more expensive and marginally faster than 512MB brother. The 256MB iteration of the same card might be limited in some cases, but at the same time it is not two times slower than 512MB version. Still, in the eyes of the public it looks that way.

We are, of course, talking about 80+ percent of the market who don’t know every codename of every upcoming and released chip, and don't know what the pipeline or Shaders are. Most of them don’t even know what a pixel is, but at the same time they spend money on graphics cards, mostly as a part of a selected notebook or PC.

The future of graphics is rather simple - the more, the merrier, and you can easily expect to see desktop cards with more than 2GB. Memory is getting cheaper by the day and 4GB sounds much better than 2GB of graphic memory. However, most of you already know that 512MB of graphics memory is enough for most games, and although 1024MB might prove to be useful at ultra high resolutions, it certainly doesn't provide a huge performance benefit.