I'm sorry. The spread spectrum theory was disproved. It was just a bunch of events happening at the same time that I pinpointed it on the wrong item. For what it's worth though, it didn't really hurt that much either, but that's not really the point.

That said, try dropping the multiplier even more and working with 1:1 to start. Don't worry about NBCC. Some of your ram subtimings may be too tight as well.

If you want to rule out ram, you can try this first:
5-5-5-15-5-35-3-11-7-13. Or if you're more certain of it, 4-4-4-10-3-25-2-9-4-11.

How many volts are you pushing to the ram? 2.1v?

Stick with 1:1 divider for testing, and turn off static read control for now.

I take it you have a rev1.03g board and don't have vnb adjustments?

How much vcore are you pushing too?

Let's do 8x400 1:1 first to confirm that you can at least scale that high in fsb. Then work multiplier up to 9. You may indeed be needing more vcore.