Déjà pour le multiplicateur je viens de trouver ceci :
When 13X, 13.5X, 14X Multiplier settings from User-defined option do not boot, but 15X and up do.
May be caused by the Red bios driven buffered circuit used in some early mobos to auto-reset/match Northbridge's FID Multiplier "ID" signals when user/bios over-ride Duron/Tbird's 4 bit default Multipliers, because their default FID signals are set by separate/fixed L6 bridges (Blue), unlike Palomino/Tbred which auto-match 5 internal FID signals (Green, and no L6 bridges) directly from Multiplier signals.
But this Palomino/Tbred FID auto-match goes awry when resetting the Multiplier to 13X and up in mobos with those 4 bit buffered circuits, as the bios User-Defined option, which enables buffers A and bypasses the new FID Logic, substitutes incorrect data for the spec 0011 signals being blocked by disabled buffers B.
Example:- 13X bios reset sends 1st 4 bits of its 10100 code, 0100 = slower 5X code, to Northbridge while spec 0011 are blocked. So longer clock periods based on 5X input should result in Northbridge setting fewer/too few wait states for timing data transfers at 13X, data may corrupt/no boot. Boots at 15X suggest that timing issues ended there.
Those buffered circuits may also exist/cause problems in some later backward compatible mobos. Initial workarounds did not recall/consider these circuits and failed.
Pour info l'url : http://www.beachlink.com/candjac/Workarounds.htm
Il me reste à trouver pourquoi 120MHz et là...
