I am dropping a note about the “issue” of Athlon XP 2100+ CPUs. I know why this is happening, and I’ll explain it.
Effectively what has changed in the 2100+ CPU is the bit value of the L10 bridges, nothing else.
The Palomino core support multipliers all the way up to 18x. These are implemented in the mobile versions of the XP. That is why there are 5 L1 bridges in Palominos compared to 4 L1 bridges in the Thunderbird core. This is also the reason why the L10 bridge was introduced. It provides a default 8X multiplier to the L3 and L4 bridges when enabled.
(Ed. note: For a more detailed explanation as to how the Palomino multipliers work (and a equivalent table), see http://www.beachlink.com/candjac/PaloDecode0.htm).
I’ve put together a table from AMD spec sheets and independent research from a few websites that decodes the value of each bridge for every possible multiplier up to 18x.
Now, the reason why unlocking an XP 2100+ doesn’t work is because of how current motherboard’s BIOSes address the issue of multiplier switching. So far there no one single BIOS which let’s you select a multiplier higher that 12.5x.
Those that offer 13x, 14x and 15x are actually remapped multipliers that actually correspond to 12.5x, 5x and 5.5x. These remapped multipliers were introduced with late Thunderbirds (1.3 and 1.4 GHz) because there weren’t any other multipliers available (again, remember those CPUs only had 4 L1 bridges). Current BIOSes ignore the fact that there is a fifth L1 bridge in Palominos which unlocks the L10 bridge which in turn is what enables higher multipliers.
This causes some additional difficulties with an XP2100 with a modified L10 bridge, because changing the L10 bridge changes the multiplier from 13X to 5X. BIOSes often don’t support the 5X multiplier, or they’ve remapped it to provide a 14X multiplier.
These problems wouldn’t occur if motherboard makers updated their BIOSes to also consider the fifth bridge. We should start to demand this to be fixed in new releases.