You have a 1.02g revision board. You can see the revision clearly right above the top RAM slot; just open the case and look right there.

That revision will NOT get high FSB with a quad. The 1.04g will get higher, but most people can't stablize 400 FSB with a quad. Weldzilla managed to reach 395 FSB, but 400 FSB was too unstable. And I think that was with a Q6600 that had no problems with 1.4 vFSB and 1.85 vmch. (45nm chips are usually recommended to keep vFSB at 1.3 or lower although people have used them at 1.4 VTT (vFSB), but I personally wouldn't go above 1.3v.

Mine does 364 FSB at 1.55vmch but 370 FSB gives about 24 errors in memtest per "test 5" loop but putting vmch at 1.65v fixes that (vfsb at 1.30 with vmch at 1.55v MIGHT also work, since as I mentioned before, increasing vfsb also raises the vmch a bit, like .06v for every .1 increase). So I managed to get stable at 370fsb with 1.65 vmch and vfsb at just 1.20v. But it's really hard to say if its better running 370 FSB x 10 for 3.7 ghz, with 4-4-4-12 timings, or 337 fsb x 11 for 3707 mhz, cas 3-3-3-8 timings (both with 2x2GB PC7200 Gskill Pi's). Too lazy to test.
But for 4 ghz, I just do 333x12, then if I ever want a little more, I have to mess around with multpliers and fsb; 0.5x multipliers would be very useful (cpu supports them; board doesn't....)

and 380 FSB requires 1.75 mch+1.30 vfsb OR 1.85 mch. Dont like running the mch that high though even with the fan I have on it and the Arctic Ceramique which keeps temps semi decent. (I shudder if it's getting 1.96v at 1.85vmch+1.40 vfsb).

Try running Prime 95 version 25.8 at your 370 FSB.
First try about 3 hours of small FFT's. Hopefully that will pass (that Q9650 should be able to do that on a lot lower than 1.36 vcore).

Then try "Blend". Also give OCCT RAM/Blend test a try as well.
You may also want to go back into memtest86+ 2.10 and try LOOPING test #5 in it over and over (test 5 finishes a lot faster than a few of the other tests....I think test 4 and 7 take like 10 minutes each?).

If you get errors in Blend or OCCT (RAM test) then it's the chipset.
Try raising your DRAM voltage by .1v higher, and try 1.40v on the vFSB.
That should get you stable. 370 FSB is pretty good for that 1.02g board...I've seen people that had problems at 340 FSB with a quad on 1.02's....nicely done.

Anyway as I said, get OCCT 3.0 and Prime 95 25.8 and try a combination of small FFT for prime/cpu test and Blend and OCCT Ram tests for chipset stability. 1.40 vfsb should hopefully get you stable.