Why do Nvidia graphics cards have a faster memory clock, but lower FLOPS than AMD cards?

Question

AMD claims that the 6870 has around a 1 GHz memory clock, Nvidia says that the 560 ti has around a 4 GHz memory clock, and AMD claims that the 6870 can reach 2 TFLOPS, while I've heard that the 560 ti is only around 1.2 TFLOPS. Yet, both cards seem to bench almost exactly the same.

Why is this?

Also, an XFX 6870 with 2GB of VRAM is $100 less than a VisionTek one, what's the difference?

Answer 1

Graphics card performance figures are a boggy minefield of confused and conflicting information.

The memory clock figures you listed are probably equivalent in that the AMD is listed as 1 real GHz, while the Nvidia is 4GHz "effective clock" with a 1GHz real clock that is "quad pumped". Chances are the AMD is quad pumped too so the actual memory bandwidth figures are exactly the same.

The other problem comparing them is as you say the FLOPS figure, which again is pretty pointless to use to compare AMD to Nvidia as they are very different architectures and perform differently. The FLOPS figure is highly dependant upon performing very particular tasks which are not necessarily the same tasks used to generate screen graphics. They are also do not take into account the benefits of the specific architectural enhancements that may make one card do specific operations faster than the other, again which card does what better is nearly impossible to quantify.

The long and short of it is that you can only really tell if one card is better than the other by actually using a real game to compare whether one card performs better and even then that is highly dependent on how well optimized the game is for either card.

TL;DR: Check benchmarks to see if it will run whatever game you are wanting to play and ignore all the pointless marketing FUD