I have an original revision of the Abit KX7-333R motherboard, which apparently doesn't support the Thoroughbred processors properly.
Does anyone know categorically what the problem is with the initial revision of this board that causes problems with some or all of these processors?
I've been doing some trawling around trying to deduce as much information about these processors as possible, but it's a total minefield of conflicting information and conjecture on a lot of sites.
Here's what I've found so far:
With the introduction of the Thoroughbred, AMD changed the way the default multiplier is encoded on the chip package. Before, it used to be done using a combination of the L3 and L4 bridges, there being four of each.
Now on the Thoroughbred, all the L4 bridges appear to be open at all times, and the multiplier encoding is done using the L3 bridges only, of which there are now 5. From left to right, these bridges correspond to multipliers of 0.5x, 1x, 2x, 4x & 8x.
Sounds simple so far doesn't it, oh I wish...
What multiplier is actually set obviously depends on which of these bridges are open, but it's not a direct correlation. If all the bridges were open, you'd only get a 15.5x multiplier, which obviously isn't enough. For this reason there's a fixed offset used, which differs depending on the value set using the bridges.
For L3 bridge values of 0x to 1.5x inclusive, a fixed offset of 11x is added, giving you the 11x to 12.x multipliers.
For L3 bridge values of 2x to 7.5x and 10x to 15.5x, a fixed offset of 3x is added, giving 5x to 10.5x and 13x to 18.5x respectively.
Mad eh? Note that not all possible values are actually valid, and for some reason the 14.5x setting equates to 18x rather than the 17.5x you'd naturally assume. Oh well. Onto the L1 bridges.
As we all know, the L1 bridges are what lets you overclock the processor. If open, they allow the motherboard to set the multiplier and override the in-built default.
With the introduction of the Palomino, the L1 bridges were redesigned from the Thunderbird with the laser-cut pits between each one, grounded at the bottom of the pits, thus making simple connection of them rather difficult. I tried to do this on an XP1800 processor and failed miserably.
With the Thoroughbred, these pits are no longer present. Now I've seen conflicting information here, as some people claim these processors are shipped unlocked and a suitable motherboard can override the setting, yet others claim you still need to connect the L1 bridges using conductive paint or similar, although this process is obviously much easier without the pits in the way.
Personally the latter makes more sense to me, as I cant see how these processors can be unlocked as shipped, since the L1 bridges are not connected.
Now onto problems with motherboards.
As said above, the 5th L3 bridge controls the 8x multiplier, and from the figures above, you can see that this needs to be open in order to set the 10x to 15.5x range of values, which correspond to the 13x to 18.5x multiplier values.
If the processor is locked, then the motherboard doesn't need to do anything to set this multiplier, as the processor will default to the value encoded using the L3 bridges and all will be fine. Problems arise, however, if the L1 bridges are connected and the motherboard has control of the multiplier.
It appears that some motherboards do not have the ability to control the 8x multiplier line, as this 5th line was added by AMD to cater for higher multipliers. In this situation, the 13x and upwards multipliers cannot be set by the motherboard, and you're stuck basically. Even if you install a new BIOS which is capable of selecting the higher multipliers, as opposed to the generic >=12.5x setting, you will end up with the wrong multiplier as the 5th line will not be transmitted to the processor. In this case, the lines representing 10x to 15.5x will actually end up transmitting the low 4 lines only, producing the wrong (lower) multiplier.
Now I'm not sure if this is where the board modifications come in. Do the mods carried out by Abit enable this 5th line, thus allowing multipliers of higher than 12.5x to be selected? Frankly I'm dubious about this. If the processors are indeed locked as shipped, since the L1 bridges are open, then the motherboard doesn't need to set the multiplier at all, and all CPUs would work correctly.
For this reason I suspect the mods carried out are intended to fix another problem, but I'm not sure what. Whether it's something to do with subtle voltage or timing requirements of the Thoroughbred core I don't know, but this is pretty much the only thing I can think of. If indeed this is the case, then the problems would extend to all Thoroughbred processors, including the lower speed variants with this newer core.
I certainly cant see any differences between the 2200+ and 2400+ processors which would allow the 2400+ to work when the 2200+ doesn't. If my assumption about the processors being locked as shipped is correct, then the multiplier setting on the motherboard is irrelevant. If the processors are unlocked as shipped, then the motherboard would be incapable of producing any multiplier greater than 12.5x and even the 2100+ and 2200+ would fail to work correctly.
Any comments or corrections are more than welcome here. I'm trying my best to understand all this and I'm not sure how well I'm doing
Does anyone know categorically what the problem is with the initial revision of this board that causes problems with some or all of these processors?
I've been doing some trawling around trying to deduce as much information about these processors as possible, but it's a total minefield of conflicting information and conjecture on a lot of sites.
Here's what I've found so far:
With the introduction of the Thoroughbred, AMD changed the way the default multiplier is encoded on the chip package. Before, it used to be done using a combination of the L3 and L4 bridges, there being four of each.
Now on the Thoroughbred, all the L4 bridges appear to be open at all times, and the multiplier encoding is done using the L3 bridges only, of which there are now 5. From left to right, these bridges correspond to multipliers of 0.5x, 1x, 2x, 4x & 8x.
Sounds simple so far doesn't it, oh I wish...
What multiplier is actually set obviously depends on which of these bridges are open, but it's not a direct correlation. If all the bridges were open, you'd only get a 15.5x multiplier, which obviously isn't enough. For this reason there's a fixed offset used, which differs depending on the value set using the bridges.
For L3 bridge values of 0x to 1.5x inclusive, a fixed offset of 11x is added, giving you the 11x to 12.x multipliers.
For L3 bridge values of 2x to 7.5x and 10x to 15.5x, a fixed offset of 3x is added, giving 5x to 10.5x and 13x to 18.5x respectively.
Mad eh? Note that not all possible values are actually valid, and for some reason the 14.5x setting equates to 18x rather than the 17.5x you'd naturally assume. Oh well. Onto the L1 bridges.
As we all know, the L1 bridges are what lets you overclock the processor. If open, they allow the motherboard to set the multiplier and override the in-built default.
With the introduction of the Palomino, the L1 bridges were redesigned from the Thunderbird with the laser-cut pits between each one, grounded at the bottom of the pits, thus making simple connection of them rather difficult. I tried to do this on an XP1800 processor and failed miserably.
With the Thoroughbred, these pits are no longer present. Now I've seen conflicting information here, as some people claim these processors are shipped unlocked and a suitable motherboard can override the setting, yet others claim you still need to connect the L1 bridges using conductive paint or similar, although this process is obviously much easier without the pits in the way.
Personally the latter makes more sense to me, as I cant see how these processors can be unlocked as shipped, since the L1 bridges are not connected.
Now onto problems with motherboards.
As said above, the 5th L3 bridge controls the 8x multiplier, and from the figures above, you can see that this needs to be open in order to set the 10x to 15.5x range of values, which correspond to the 13x to 18.5x multiplier values.
If the processor is locked, then the motherboard doesn't need to do anything to set this multiplier, as the processor will default to the value encoded using the L3 bridges and all will be fine. Problems arise, however, if the L1 bridges are connected and the motherboard has control of the multiplier.
It appears that some motherboards do not have the ability to control the 8x multiplier line, as this 5th line was added by AMD to cater for higher multipliers. In this situation, the 13x and upwards multipliers cannot be set by the motherboard, and you're stuck basically. Even if you install a new BIOS which is capable of selecting the higher multipliers, as opposed to the generic >=12.5x setting, you will end up with the wrong multiplier as the 5th line will not be transmitted to the processor. In this case, the lines representing 10x to 15.5x will actually end up transmitting the low 4 lines only, producing the wrong (lower) multiplier.
Now I'm not sure if this is where the board modifications come in. Do the mods carried out by Abit enable this 5th line, thus allowing multipliers of higher than 12.5x to be selected? Frankly I'm dubious about this. If the processors are indeed locked as shipped, since the L1 bridges are open, then the motherboard doesn't need to set the multiplier at all, and all CPUs would work correctly.
For this reason I suspect the mods carried out are intended to fix another problem, but I'm not sure what. Whether it's something to do with subtle voltage or timing requirements of the Thoroughbred core I don't know, but this is pretty much the only thing I can think of. If indeed this is the case, then the problems would extend to all Thoroughbred processors, including the lower speed variants with this newer core.
I certainly cant see any differences between the 2200+ and 2400+ processors which would allow the 2400+ to work when the 2200+ doesn't. If my assumption about the processors being locked as shipped is correct, then the multiplier setting on the motherboard is irrelevant. If the processors are unlocked as shipped, then the motherboard would be incapable of producing any multiplier greater than 12.5x and even the 2100+ and 2200+ would fail to work correctly.
Any comments or corrections are more than welcome here. I'm trying my best to understand all this and I'm not sure how well I'm doing