Originally posted by: SerpentRoyal
Originally posted by: Replay
E2140 @ 3.208 Ghz (L2) 8x401 Prime95 stable.... Very happy to get a 100% overclock on this chip. Low multi and fsb wall are holding it back.
There shouldn't be a FSB wall on the MB. The E21x0 and E4xx0 chips will top out around 425MHz FSB. Try upping NB and SB by one notch. Also use 5-5-5-15-2T timing.
I appreciate how quick and generous you are to help others on Anand (and on the Abit forums). Cpu testing starts with low and relaxed memory settings, and memory tests begin with low cpu speeds. Then I test with tight ram timings, and the maximum stable cpu speed.
The FSB wall, and low multi, which I refer to, are limitations of the individual cpu chip (and not the motherboard). Perhaps my terminology is wrong? Bumping the bridges has helped wring significantly more speed from chips with higher multipliers, but these 8x chips refuse to go higher. Most of them run right up to a frequency wall, at relatively low core voltages, and go no further.
I believe the Intel cpu packaging for low-end products limits the stable front side bus frequency to around 400 Mhz +/- some Mhz. Looking in this thread, and elsewhere, it is a rare E21x0 which goes much higher than 400 Mhz. With a higher multiplier, and a lower fsb, these same cores commonly reach higher clock speeds.
The motherboard can do better, perhaps 500 Mhz. This ram tests good to DDR 1132 with 5-5-5-15 timings. And I think the silicon in the cpu would do better, but for the high frequency limitations of the packaging.
I still wonder if the BSEL mod to a 333 MHz initial bus speed would loosen some timings enough to squeeze a little more from these 8x chips. From what I've read it probably is not worth the effort. But, since I cannot stop tinkering with this stuff I may yet try that mod. I'm also toying with the idea of adding some more smd capacitors to the underside of the chip package.