Athlon XP-M 3000+ on 2390B

nSurferx wrote:
> Wow, a detailed reply! Thanks for those helpful hints. I'll start with
> software kit first and see how it goes.

UPDATE: CrystalCPUID worked! Thanks for the link.
 >> Stay informed about: Athlon XP-M 3000+ on 2390B 

nSurferx wrote:
> I bought this hard-to-find Athlon XP-M 3000+ 266 fsb and installed it on
> a good ol' Tyan 2390B (latest bios 1.13)
> and both posting and Sandra reports the processor as 800mhz.
>
> How do I make the system recognize it as 2.2ghz (3000+) ??
>
> TIA.

Mobiles require multiplier control to work. They also need the user
to set the clock speed to something useful as well. I run a
2600+ Mobile in a desktop board, at 200x11, and the clock
speed is set in the BIOS, as well as the multiplier. 200MHz
gives FSB400, which helps with bandwidth to the Northbridge.

I didn't look on the Tyan site, but your board seems to use
KT133A chipset. You are limited there to FSB266, which means
you need a "high" multiplier to get all the way to 2.2GHz.
2200/133 = 16.5 .

A Mobile has two speeds. It has the two speeds, as in a laptop,
you need a "slow" speed to save battery power. When a Mobile
starts, it starts with the "slow" multiplier. There is also
the "normal" multiplier, which would be 16.5 in your case.

Mobile chipsets provide the easy ability to set the multiplier
value. I believe the BIOS makes an ACPI object available, so
the OS can switch modes. On Desktop chipsets, there is likely
no official support, but there are software utilities that
can load a multiplier, with the right chipset. (For example,
the Nforce2 chipset doesn't support software setting of the
multiplier, but other chipsets do. I'm not aware of a
comprehensive list of which ones do.)

There are also hardware methods of setting the multiplier.
To use these, you need access to the top of the socket,
where you insert tiny wires. Or, you remove the motherboard
from the case, and solder tiny wires to the bottom of the
socket. Both techniques do the same thing, but the soldering
method has the benefit, that the connections won't change on
you. The tiny wire method is a bit more fiddly. If you use
single strands of copper wire, for example, the copper can
oxidize with time, leading to a flaky connection. The best
wire, would be nickel or tin plated, rather than raw copper.
But stranded wire with a nice nickel plating, is harder
to find than it used to be. You would be hard pressed to find
anything other than flaky copper wire, at Radio Shack.

http://www.ocinside.de/go_e.html?/html/workshop/pinmod/amd_pinmod.html

On the above web page, select "PCB View", "Barton", and
for "Multiplier Selection" use "16,5 x FSB". The picture
returned by the web site, shows the five wires that would
be soldered to the bottom of the socket. Instead of using
wires directly, it would also be possible to buy switches,
so the multiplier can be programmed when the system is
reassembled. When using the wire method, if there is a
multiplier setting offered in the BIOS, you set the
BIOS to [Auto], so that the motherboard does not fight
with the wires. If you used a manual setting in the BIOS,
then some GPIO pins on the motherboard, may try to drive
the same pins as you've attached the wires. A setting of
[Auto] means the GPIO signals would be disabled.

The other method, is the software utility route. You run
a utility like this in Windows, and try to set the multiplier
while the OS is running. That means you go from 800MHz
to 2200MHz in a flash. If that high a clock is not stable,
you may crash instantly (especially if the processor needs
a bit more voltage).

http://crystalmark.info/software/CrystalCPUID/index-e.html
http://crystalmark.info/download/CrystalCPUID48.zip

The other issue is voltage. Be aware that the Mobile VID
signals are interpreted differently than on a desktop chip.
This is not all a bad thing in this case. For example,
my chip has a Mobile voltage rating of 1.45V. On my desktop
board, the Vcore interprets the bit pattern for VID as
1.575V. The extra voltage helps to reach the high clock
of the normal mode. On my Mobile, I needed to set the
voltage in the BIOS to 1.65V (which is normal for a
3200+ anyway), to get stability. If your BIOS did not
have the option to set the voltage, then you might
end up with the 1.575V type value. Note that excessive
voltage on the Mobile chips doesn't help - mine won't
POST with more than 1.70V applied to it.

The table here "L11 : Code to CORE Voltage Definition" shows
the correspondence between the Mobile coding for voltage
and how a desktop motherboard handles it. My "1.450(Q)"
gives me 1.575V in the BIOS. But I can also set the voltage
in the BIOS, and the BIOS settings are normal and are not
misinterpreted. So the following table only applies if your
Vcore is in [Auto] mode - the desktop voltage value is
applied when the motherboard adopts a "hands-off" approach
via the [Auto] BIOS setting.

http://fab51.com/cpu/barton/athlon-e23.html

So, see if you can change the multiplier with CrystalCPUID.
If not, you always have the option of using wires. Lots
of experiments to do

These forums are another place to look for more help. The
A7V family is similar in vintage, to your KT133A motherboard.

http://www.nforcershq.com/forum
http://www.nforcershq.com/forum/search.php

http://www.a7vtroubleshooting.com/forum/cgi-bin/yabb/YaBB.pl
http://www.a7vtroubleshooting.com/forum/cgi-bin/yabb/YaBB.pl?board=;ac...n=searc

HTH,
Paul
 >> Stay informed about: Athlon XP-M 3000+ on 2390B 

nSurferx wrote:
> nSurferx wrote:
>> Wow, a detailed reply! Thanks for those helpful hints. I'll start with
>> software kit first and see how it goes.
>
> UPDATE: CrystalCPUID worked! Thanks for the link.

You can test stability with Prime95 (mersenne.org). There
is an option called the "Torture Test". Running mixed FFTs
for four hours or more, without seeing any errors, is proof
everything is in good shape. If you see a "roundoff error"
in under 10 minutes, then you have more work to do to stabilize
your new setup. It could be that just a little more Vcore
would help, if you see problems in Prime. Either that, or
accept a lower core frequency setting.

If you can pass the Prime95 test, then chances are you won't
see any crashing or misbehavior later. Prime95 is also a
thermal test, and you should open a temperature utility
if you have one, to display the CPU temperature. Since
Prime95 creates a 100% CPU load, it will simulate conditions
when the computer is working hard, and allows you to see
whether the CPU cooler is adequate or not. If I remember
correctly, my Mobile runs at 43C under load (in a fairly
cool room temp). If you don't need a lot of Vcore to keep
the processor stable, a low Vcore helps keep the temperature
low.

Have fun,
Paul
 >> Stay informed about: Athlon XP-M 3000+ on 2390B