"the reason? my crappy radeon8500 dies at with more than 104fsb. that's right.. a lame 70mhz AGP. it can't handle an extra 4mhz of heat."
That's strange, I've run my Radeon 8500 at up to 150 FSB, and that's on a 440BX chipset! It only has a 2/3rd AGP divider so I was running 100 mhz on the AGP bus no problem. It's one of the reasons I still keep the card since the Ti500 I had would not handle over 73 mhz or so.
I've tested a bunch of sticks that do DDR400:
2-256 MB DTL sticks (0224), 200, 2.7 volts
1-512 MB DTL (0221), 206, 2.8 volts
1-512 MB Corsair XMS3200, Ver. 1.1 (Samsung CTL), 213, 2.8 volts (haven't tried higher)
2-256 MB CTL sticks (0208), 206, 2.8 volts. One stick will do 213. Haven't tried higher with one stick.
All the results above are on a P4B266-E, which locks out at 2.5-3-3-7 over DDR340 or so, so I couldn't test faster settings. But because it is an Intel 845D board, Sandra scores easily surpassed 3k.
When I had a P4S533 board, the 2-256 MB CTL sticks dated 0208 (same as in the VR-Zone article!) actually did 200, 2-2-2-5-1 using a mem ratio like 3:5 or something (2 sticks!). And I mean stable in 3D. But if I tried to do near 200 mhz at 4:5, had to drop down to 2.5-2-2-6-2. But I think it was because my CPU was near it's limit at around 158-160 FSB, plus the board is infamous for instability over 155 FSB due to the lack of an AGP/PCI lock or 1/5th divider.
BTW, I tried a recent batch of 2-256 CTL sticks (0225) and they would not do 200 mhz at all on the P4B266-E. Best they could do was around 193 at relaxed settings on an MSI board. So the 0224 DTL sticks actually performed better.