Inaccessible 13.0x, 13.5x, and 14.0x

Hi all,

I just bought an Athlon XP 2500+ (Barton) for my KG7-RAID.
Since my RAM is not really good, I decided to keep the 133 MHz FSB
(or few MHz more), and raise the CPU multiplier instead.
The CPU comes with L1 unlocked, and L3 says 11.0x is the default
multiplier, thus the range I can manually set is from 5.0x to 12.5x,
which is a bit too little (1666 MHz, i.e. I'm using a 333 MHz FSB
CPU on a 266 MHz FSB motherboard).

So I modified the motherboard, by soldering a wire on AJ27, on the
back of the CPU socket, and then connecting it (via a switch and a
100 ohm resistor) to Vcc (Vcore).

Thus I get remapped multipliers.. as the following map shows:

5.0x - 13.0x
5.5x - 13.5x
6.0x - 14.0x
6.5x - 21.0x
7.0x - 15.0x
7.5x - 22.0x
8.0x - 16.0x
8.5x - 16.5x
9.0x - 17.0x
9.5x - 18.0x
10.0x - 23.0x
10.5x - 24.0x
11.0x - invalid
11.5x - 19.0x
12.0x - invalid
12.5x - 20.0x

Now here's the problem:

The CPU boots only if the multiplier is 7.0x (i.e. 15.0x), i.e.
all the settings of 6.5 (included) and below (i.e. down to 5.0) do
not work, while instead they should really work.

But how can I fix it, in practice? I want to be able to set also
the multipliers 13.0x, 13.5x, and 14.0x.

Please notice that I don't want to modify the CPU in any way
(unless it's ABSOLUTELY indispensible), since I would like to
keep the 3 years AMD warranty, and that's why I modded the KG7-RAID
instead of the CPU (by the way, is there really any way for AMD
to discover that a CPU was overclocked, when it wasn't physically
modified?).

Any hints to solve this "inaccessible 13.0x, 13.5x, and 14.0x" problem?

You help is much appreciated.

Thanks!
Arnold


PS: I discovered an interesting thing (maybe it was already known..
but, if not, here's my little contribution): the CPU reads the
multiplier settings at reset, and then latches/remembers them.
I.e. if one changes the multiplier inputs (AJ27, etc..) when the
CPU is already running, nothing happens/changes. At reset, instead,
the CPU reads those inputs and remembers them.
In substance, they do not connect directly to the multiplier hardware.

I must say however that I've been able to test this only with AJ27
(8x), so if this applies also to the other pins is just theory.

(Arnold Z.) wrote in message ...
Hi all,

I just bought an Athlon XP 2500+ (Barton) for my KG7-RAID.
Since my RAM is not really good, I decided to keep the 133 MHz FSB
(or few MHz more), and raise the CPU multiplier instead.
The CPU comes with L1 unlocked, and L3 says 11.0x is the default
multiplier, thus the range I can manually set is from 5.0x to 12.5x,
which is a bit too little (1666 MHz, i.e. I'm using a 333 MHz FSB
CPU on a 266 MHz FSB motherboard).

So I modified the motherboard, by soldering a wire on AJ27, on the
back of the CPU socket, and then connecting it (via a switch and a
100 ohm resistor) to Vcc (Vcore).

Thus I get remapped multipliers.. as the following map shows:

5.0x - 13.0x
5.5x - 13.5x
6.0x - 14.0x
6.5x - 21.0x
7.0x - 15.0x
7.5x - 22.0x
8.0x - 16.0x
8.5x - 16.5x
9.0x - 17.0x
9.5x - 18.0x
10.0x - 23.0x
10.5x - 24.0x
11.0x - invalid
11.5x - 19.0x
12.0x - invalid
12.5x - 20.0x

Now here's the problem:

The CPU boots only if the multiplier is 7.0x (i.e. 15.0x), i.e.
all the settings of 6.5 (included) and below (i.e. down to 5.0) do
not work, while instead they should really work.

But how can I fix it, in practice? I want to be able to set also
the multipliers 13.0x, 13.5x, and 14.0x.

a href="http://www.beachlink.com/candjac/index.htm"/a
http://www.beachlink.com/candjac/index.htm
Link to Workarounds article...esp section relating to the problem you
describe. Your mobo may have those buffered auto-matching circuits.
Try the workaround described, does not deface the CPU.
John C.



Please notice that I don't want to modify the CPU in any way
(unless it's ABSOLUTELY indispensible), since I would like to
keep the 3 years AMD warranty, and that's why I modded the KG7-RAID
instead of the CPU (by the way, is there really any way for AMD
to discover that a CPU was overclocked, when it wasn't physically
modified?).

Any hints to solve this "inaccessible 13.0x, 13.5x, and 14.0x" problem?

You help is much appreciated.

Thanks!
Arnold


PS: I discovered an interesting thing (maybe it was already known..
but, if not, here's my little contribution): the CPU reads the
multiplier settings at reset, and then latches/remembers them.
I.e. if one changes the multiplier inputs (AJ27, etc..) when the
CPU is already running, nothing happens/changes. At reset, instead,
the CPU reads those inputs and remembers them.
In substance, they do not connect directly to the multiplier hardware.

I must say however that I've been able to test this only with AJ27
(8x), so if this applies also to the other pins is just theory.