The Nvidia GTX480 has a reported memory bandwidth of 177GB/s. Isn't the bandwidth of a PCIEx16 slot only about 8GB/s? If so, wouldn't the PCIEx16 slot be the "data bottleneck" so than all of the GPU's speed > 8GB/s be unrecognized?. And if that's so, wouldn't a video card with 8GB/s throughput give the same display as one with 177GB/s capability?
When producing 3d graphics, there are two major uses of the video ram in the graphics card - the first use is storing all the textures, which are basically bitmaps images, that are applied to the geometry in the 3d scene. The second main use is for storing the image data when anti-aliasing algorithms are applied to the rendered image - which for some reason uses a lot of ram. So, most usage of the PCIe bus is loading the textures into the ram on the graphics card. Most of the textures are loaded up in advance- it loads the whole library of textures so it can generate the images for any scene in the game (or a particular level in the game) - without needing to load more up from the HDD. That's why the ram in video cards is so large. The image creation and processing then all occurs on the card, nearly independently from the mainboard of the PC. So the video memory bandwidth is relevant to the processing of the GPU and what it is storing on the video ram - the PCIe bus is only needed to get the info into video ram in the first place. It isn't needed so much for the actual processing tasks that create the images that go to the monitor. It's basically equivalent to the memory speed on the mainboard in comparison with the HDD speed. Once everything that is needed is loaded into ram, the speed of the HDD is irrelevant.