CORE           I/O
  LT1086CT    BD 250   |  LT1086CT
             -------   |
   -----    |   O   |  |    -----
  |  O  |   |       |  |   |  O  |
  |     |   |XXXXXXX|  |   |     |
  |XXXXX|   |XXXXXXX|  |   |XXXXX|
  |XXXXX|   |XXXXXXX|  |   |XXXXX|
   | | |     |  |  |   |    | | |
 ADJ | Vin   B  E  C *)|  ADJ | Vin
     Vout              |      Vout

   E\                                  |
     \| B                              |
BD250 |------+                         |
     /|      |   ------------          |
   C/        |  | LT 1086 CT |         |
    |        |  |            |         |
5V -+---WWW--+--| Vin   Vout |--+------+- core
    |   10      |            |  |      |
    |   Ohm     |    ADJ     |  W      |
    |            ------------   W 125  |
    |                 |         W ohm  |
    |                 |         |      |
  ++++                +---------+    +++++
  10µF     1.25V+-O O-++             100µF
  ----          |     WW 2x10 Ohm    -----
    |      1.3V +-O O-++               |
    |           |     W 10 Ohm         |
    |      1.4V +-O O-+                |
    |      ...  |     .                |
    |      ...  |     .                |
    |      2.7V |-O O-+                |
    |           |     W 10 Ohm         |
    |      2.8V +-O O-+                |
    |           |                      |

*) The pin out of the BD250 is in conflict with my data book, but my DMM names the pins this way and it works this way!!!

Modifying the
ASUS  P/I-P55T2P4  Revision 2.x
to support split voltage to use K6 etc.

Disclaimer of warrantee :
I'm not responsible for any error, misinformation, or damage that may result from the usage of my page.
You use all at your own risk!

If you want to use new CPUs like all K6 (166,200,233,266,300,333,350,3d,3d+) in your T2P4 revision 2 you have two solutions.

1.) the easy one     : buy an Adapter socket that supports BF2 jumper and low voltages like 2.2V (and below!)
     - advantage       : quick, easy, can be done by everyone who can change a CPU
     - disadvantage  : the price is in the US$50 range.

2.) the tricky one   : build your own voltage regulation for your board.
     - advantage       : cheap (less than US$20), Every voltage between 1.25V and 2.8V, FUN!
     - disadvantage  : If you are qualified to do this there are no real disadvantages, because this is part of your hobby : experience in building electronic devices needed, more time, more work, not so safe

0.) buy an Intel overdrive CPU for the same price as an K6 and a new board together!

Gladly, when ASUS designed the T2P4 Rev 2.1 board they divided the power planes for I/O voltage and core voltage like its recommended for Pentium MMX and AMD K6 CPUs and linked them together with four bridges, called JP14 - 16. But sadly the core voltage and not the I/O voltage is fixed to the onboard regulation and can't be disconnected easily. So we have to remove the onboard regulation and have to build a new regulator for I/O and a new regulator for core!

I have checked this only with my T2P4 Revision 2.1 board!
Your board may be different!
Check every hint I give!

First, solder out the four bridges called JP14 - 16 near the CPU socket and replace them with jumper pins. You can cover the pins with jumpers to use the board as it was before, and you need them to feed in the new voltages of your own build regulation.

Its a nice Idea, to ground the BF2 pin and solder in the pins for the USB connector at the same time! I don't know if USB works or not, but its not much work if you do it the same time. I will check on my rev 3 board what additional resistors are needed, and will announce it here when I open that system the next time. (or if somebody is so friendly and do this for me :-)

On my board all the four pins of each row are connected together. (check this!)
The row on the socket side are connected to the I/O voltage pins of the CPU. (check this with a K6 pin out.)
The row on the regulator side is connected to the core voltage and the power transistor. (check this!)

That's all you have to check, before starting to replace it.

Because the max. current of K6's can be up to 10 Ampere the onboard regulation surely is not able to provide enough power, so sadly we have to remove the output transistor. (The one on the heatsink)
And I see really no way to separate them securely and use it for I/O voltage. :-( nobody can see inside a multilayer board! )

Now you have to make your own regulation. There are surely better designs than my solution, but mine is easy, cheap, it works and last but not least I was able to design it :-)

When you design your own solution search the CPU Data sheets what targets you have to reach.

My solution will do:
I/O   : 3.5 Volt 1.5 Ampere
core : 1.25Volt - 2.8Volt (not higher!!!) and 'enough' current

You need :
  2 x adjustable LOW-DROP Voltage regulators like the LT 1086 CT ( 1.25 - 30 Volt 1.5Ampere)
  1 x PNP power transistor like the BD250 (max. current 40A, max. 125Watt at ideal cooling!)
  1 x cigarette box sized heatsink (6 cm x 4 cm  with 2 cm fins or so, just to name something)
  thermal grease and insulated mounting material for the regulators and the transistor!

20 x 10 Ohm resistors and a handful of others, all ¼ Watt
     a handful of some capacitors
17 x 2 jumper pins
     some other stuff

Mount the regulators and the transistor as you can see on top of my scheme with thermal grease and insulation foil. (I don't know what its called, but you should really know what to use, as I recommend this only for serious experienced electronic hobby solderer!) Don't forget the insulation on the screws!
The left regulator and the transistor are for the core voltage and are mounted to be easy connected. Start with the 10 Ohm resistor in the 5V line.  Then make the direct connections between the transistor and the left regulator. Because there are so few items I builded it without an epoxy card (except the jumper row with the seventeen 10 Ohm resistors!)
That's an issue of personal preferences how you make it. You can make lousy design on a card, and excellent design without it! I often prefer 'free soldering' for smaller projects, because I hate using a 10x10cm card when all things fit well into a matchbox! And it depends on yourself if this 'free soldering' is like a provisory or if its build for eternity. (All fixed well, good locking soldering, nothing can be bend together, no insulation burned, all wires have right size, good thermal connection to heatsink, cables come to a fixed point and nothing gets broken if you bend a cable etc.) The only real disadvantage I see is, replacing something may sometimes be more difficult, because 'free' design is 3d design, while card design is only 2d and because in free design the wires of the items are often used for the connections.

How does the regulation work?

I/O voltage.
The I/O voltage is very simple. Nearly everything is done by the regulator itself. The input voltage is feed into the Vin pin. At the Vout pin the voltage gets out! The output voltage is always 1.25V higher than the ADJust pin. So when you put a 125 Ohm resistor between them there is always a current of 10mA. This current will flow through the resistor we put from ADJ to ground. We take a 220 Ohm resistor and because of the 10mA through this resistor there will be a voltage of 2.2V over the resistor. So the ADJust pin has 2.2Volt. The OUT pin is always 1.25V higher, so we will get 3.45V at the output. Because of item variations in the 1.25V and the resistors you have to 'play' around a bit with different resistors. Surely you won't hit 3.5 Volt with varying both resistors, so you can use two resistors to get the 125 Ohm (and a 125 Ohm resistor would be very hard to find, too! :-) At the input and the output there are capacitors connected to ground to stabilize everything. It would be a good idea to add something like 100nF or so, too, to improve stabilization. Also you can add a capacitor from ADJ to ground. And you must connect a 220 OHm resistor from output to ground, to allow the regulator start regulation when there is no other load on the output. (or you would be very confused about the high output voltages you will measure! Been there, done that! :-)

core voltage
The core voltage is nearly as simple as the I/O voltage. One difference is, that there is not one resistor from ADJ to ground, but 17 equal resistors of 10 Ohm and you can choose with a jumper how much you use. That's on reason, we used a 125 Ohm resistor, because this resistor must be fine tuned and therefor consists of two resistors and because it makes calculation so easy. (And because that's the value in the data book :-) Because the lowest voltage starts at 1.25V the first resistor is 5 Ohm (two 10 Ohm parallel to keep things easy :-) so we get 1.3Volt. Than every resistor adds 0.1Volt, until we finally reach 2.8Volt. Set it to 2.8Volt and then vary the 125 resistor, until you really get 2.8Volt! Now you think you are clever and add more resistors to get higher voltage, but as you can see soon that don't work, except with very tiny currents. Now we came to the more important difference, the high power PNP transistor. It uses a very common trick (My self-made bridged stereo amplifier for my home hi-fi is using the same trick with cheap low power integrated amplifier chip and a PNP and a NPN transistor in the +/- lines.) When current starts to grow, the regulator is consuming more current. This current is running through the 10 Ohm resistor in the input line. When the voltage that this current produces reaches 0.7Volt, the transistor becomes more and more conductive and feeds the output more and more. So the regulator itself provides only a maximum current of 70mA, the rest is provided by the transistor. The transistor I choose (because I have some of it in my grab box :-) can provide a maximal current of 40A, and that's much more than the power supply can deliver, so its enough! When you can manage to keep the transistor casing at 25°C the maximum power the transistor can destroy is 125W. If we set the lowest voltage it has to eliminate 3.75 Volt. That would allow 33 Ampere. Though we can't keep the casing at 25°C we have still a good safety margin with this transistor. As an estimation with a K6-266 the transistor would reach a critical zone when its outside is hot like boiling water. And to avoid that we use a heatsink. A cigarette box sized will be more than enough. And if you place it in a little air stream (from a fan for example) it would be always cool to the touch! (And I like to keep everything in my Computer cool to the touch!)

The 5 Volt will be provided by a cable to a Disk power connector, not the MB, because of the high current, but the ground connection runs direct to the MB, because there is very few current but it is used as the reference point for the regulation. And if the reference is influenced by long cables and other devices inteferencing with it, you can't expect a stable voltage! Near the core I/O connector (the jumpers you soldered in at the begin) are lots of ground points you can use. If you use connectors (fan for example), please make two independent ground connections, because if ground connection gets lost, the voltage will rise to the highest possible voltage! (It will act like a gas balloon with cutted ground wire!)
A good point to explain, what the highest voltage will be. With a very small current both will be in the >4V range. But when it comes to a remarkable current both can deliver different max. voltage. The I/O Voltage can 'theoretical' be stable up to 4V. (5V input - 1V LOW DROP = maximal 4V output) The core voltage can 'theoretical' be stable up to 3.3V. (5V input - 1V LOW DROP -0.7V transistor = maximal 3.3V output) But that's only a theoretical value! In practice I found out, that everything above 2.8V isn't stable with my K6-200. setted to 2.9 or 3.0 Volt it will work, but a simple DMM will show that the voltage always jumpers around in the 2.8V to 3.0V range. I could search why this happens, but because my K6-200 will be happy with 2.7V I don't waste my time and be happy with my solution!

I made a new jumper block for all BF pins, so they were free to use as ground points. I soldered in some other jumper pins in free holes in the area between fan and core-I/O-connector, and soldered connectors in an epoxy card to place it right over them. Then I made all ground connections (check with Ohmmeter twice which are really ground connectors!) and connections for core and for I/O and screwed the hole epoxy card under my heatsink. (with insulation of course and all not used pins soldered, but carefully connected to nothing!) That will fix it, but is of course not safe enough! But my CPU heatsink is self made and was cut out of the same material and is glued to the K6 and is secured to the socket by two springs, so I simply screwed a plastic bar between them, so everything is fixed well! And this design allows me to blow through CPU heatsink and voltage heatsink with one 80 mm fan. (running at 9V and sucking in fresh air, guided by cardwork) Because even, when my thermometer sensor under the CPU says 42°C both heatsinks themselves are cool to the touch. The high temperature is not produce by bad cooling, but by the thermal resistance between the chip itself and the chip casing. So using a full power 92 mm fan may reduce the CPU temperature to 40°C???

BTW: It would be a nice idea to put in some extra electrolyte capacitors into the soldering spots in CPU socket to stabilize Vcore and Vio.

This description is not so well written and detailed in all steps, because I think it will not be often done and I recommend it only to persons with experience, so I think it's enough in the moment. I posted a first announcement some time ago of my success of modifying my rev 2.1 board to use K6 and didn't got a single response or request. So I don't waste more time with this, until I see there is real demand!

to be continued...