At what point is FSB a real bottleneck for quad chips?

[Speederlander]

Yes, it varies by application I know and I know there are other variables, but that said, when does FSB become a real-world issue with quad chips? I see over and over that "quads love high FSB". Ok, fine. But do they love high FSB in anything other than synthetic benchmarks or do we see REAL improvement (a couple % doesn't count) in apps? If I run one quad at 266 FSB and another at 400 FSB (both at same CPU speed), are there any apps that would see improvement? Assume 4 gig of fast RAM for the sake of argument so this doesn't become a discussion of that point.

[SLi_dog]

3D applications seem to like higher FSB, not sure about any diminishing return point or other "real world" applications though.

[Speederlander]

Anyone have specific experience looking at performance differences with only FSB variable?

[SLi_dog]

HeavyH20 was saying his Aquamark score increase by ~20,000 points running at the same CPU speed just by going from the eVGA 680i to the eVGA 680i RevA1 which allowed running at a higher FSB.