Let's upgrade an early Socket 7 board beyond the limit (Part 1): RTC, VRM, MMX

hello and welcome i know that many of you guys  like my videos about repairs and restoration   but don't you also think that today is a nice day  for an interesting diy at least i think so and in   the end of today's video i hope that you feel  the same i have here one interesting main board   it is made by asus and the name is pi p55 tp4xeg  revision 2.4 this was a quite common mainboard   back in the days as painting was a new and shiny  cpu on the market it is based on the intel 430fx   chipsets also known as triton 1. although  this chipset was quite solid for its time   it is actually quite slow as well it was made  in the time before painting mmx was released   and so it officially supports only non-mmx  painting 1 cpus with front side bars of 50   60 and 66 megahertz and a core speed of up to  200 megahertz the pre mmx cpus needed single   cpu voltage of 3.3 volt which was not provided  by the at power supply so the big cooler with a   voltage regulator is located down here on the main  board up here we have the level 2 cache currently   there are 256 kilobytes of total cache installed  but this main board supports up to 512 kilobytes   of level 2 cache the cache can be extended either  using the dip ics or so-called coast module   unfortunately both types of cache  cannot be used simultaneously   and here on the left we have four pci slots  and three iso slots as this board was sold   the agp port was not yet introduced so this main  board is limited to pci graphics cards only well   ice are two but that makes no sense so why do you  think this main board actually caught my attention   as you probably see the lowest pci  slot has some kind of proprietary   extension it is called asus media bus there were  different versions of these slot extensions and   they all were incompatible to each other as  was masted completely up and today a main   board and such a media bus expansion card which  would fit to each other are quite hard to find   apparently i have here a card which does fit  into this main board it is a combination of   an adaptex kazi controller which hangs on  the main pci interface and a second part   a creative sound blaster vibra 16. i think that  i don't need to explain what a sound blaster 16   is but the interesting thing is that despite  of the compactness of the card it still has   the original yamaha opl3 chip which would give  a nice and genuine fm sound in the dos games   and furthermore it even kept the wave table  header for dotted board and general midi sound   on top we have here ultra wide scuzzy and 50p  normal scuzzy connectors for the hard drive   cd-roms etc and on the slot brackets we have the  usual sound card stuff joystick port audio in and   output and the volume control wheel this scassy  and sound expansion card can be plugged into this   slot like that and i thought where i have such a  rare combination at hand it would be quite cool   to make a retro pc based around it but it would  be quite boring just to take it and simply push   the whole thing into a case so i wondered if i  can make some interesting upgrades to this board   let's start simple this board has a dallas rtc  module which i modified a long time ago by adding   an external cr2032 battery i already showed couple  of times how this can be done it works just fine   but there is one slight problem with this mode the  rtc module is located in the end of the iso slot   and with the battery on top it is simply too high  when i'm trying to plug in a longer iso card into   the slot the modified rtc module with the battery  is in the way so the last iso slot is unusable for   a longer cards like this one and that is a shame  what can be done about it well i could put the   cr2032 on wires and remove it from the top of  the rtc module but fortunately i have this if   you watch my channel regularly you will remember  this rtc module replacement which i introduced   some time ago we can now simply replace the  original rtc module by this mw3287 and it should   be low enough so using long iso cards would be  possible again and as you see it is a perfect fit   ok the new rtc module is installed let's take a  look what else can be improved in this board a   painting 133 is currently installed a 3.3 volt  cpu which is powered by this voltage regulator   unlike atx the at power supply doesn't provide  3.3 volts but only plus minus 5 and 12 volts   so the voltage conversion down to 3.3 volts  need to be made directly on the main board   and as you see this regulator has a big hit sink  because it can get quite hard it is completely   responsible for the cpu power  which is actually quite hungry   as i said this is a pentium 133 megahertz cpu  which you can see here the cpu is a good fit   for this main board which supports pentium  cpus from 75 megahertz to 200 megahertz   but wouldn't it be nice to have a pentium  mmx 200 megahertz in this mainboard   beside the mmx instructions which can boost the  performance in some optimized applications this   cpu has also twice as large level 1 cache as the  normal non-mmx pentium 32 kilobyte instead of 16.   the mmx variant of the pentium cpu had some  more improvements and depending on application   it could provide very measurable performance  gain over the non-mmx version on the same core   frequency so let's see why we can't just drop in  a pentium mmx into this socket this main board   was made in the time where pentium mmx was not yet  released so officially it doesn't support the cpu   the actual problem is neither the mmx nor to old  chipset on this mainboard but the cpu voltage   you see original non-mmx pentium cpus used so  called 3.3 volt single voltage power to get   down with the temperature intel decided to lower  the cpu voltage for the pentium mmx cpus however   all the infrastructure around the cpu was already  implemented for 3.3 volts and switching everything   to a lower voltage would mean a huge amount of  re-engineering and electric incompatibility to   all the devices already available on the market  so intel introduced an interesting solution which   lives on until today a dual voltage one voltage  for the i o at 3.3 volt which was used for   communication with the infrastructure and one  core voltage which the cpu used internally   for the pentium mmx200 the core voltage which  is also called v-core was set at 2.8 volts and   the peripheral io voltage or vio remained at  3.3 volts just as for all the pentiums before   now to the problem with this mainboard the bad  news is that it was made for 3.3 volt single   voltage cpus and will not work with dual voltage  cpus out of the box the good news is however it   was prepared for a vrm a voltage regulator module  which can be added to provide additional cpu   voltages as you see the vrm connector is not yet  soldered it has two rows of 15 pins each and some   of them are hard wired the question is what is the  pin out well luckily intel described the vrm in   its pentium processor flexible motherboard design  guidelines it receives multiple voltages 3.3 volts   5 volts and 12 volts so if the cpu is a single  voltage model like all the non-mmx pentiums then   the pins a4 through a7 and b4 through b7 can  be simply shorted this is what we see here   on this main board here the same 3.3 volts  are directly sent to the vio and the v-core   in case of dual voltage cpus like pentium mmx  the 3.3 volt should be sent only to the vio   and the v core should be regulated from the 5  volt rail theoretically it is also possible to get   it from the 12 volts rail but this might be  too much of a current since we have only one   12 volts pin furthermore dependent on the  type of regulator it could probably be more   efficient to get low voltage from 5 volts than  from 12 volts and apropos voltage current and   efficiency the pentium mmx cpu which i'd like to  use in this main board needs 2.8 plus minus 0.1   volts and due to intel's documentation this  cpu draws up to 5.7 amps that's quite a lot   actually and using a linear voltage regulator as  i showed it for a 486 in one of my last videos   would need a big heat spreader to dissipate over  12 watts i think we need a more efficient solution   and to remind you a little bit a linear voltage  regulator provides a steady output voltage from a   higher input voltage dissipating excessive energy  as heat and that heat can be reduced using another   type of a voltage regulator let's imagine to  have a power switch between the power supply unit   which delivers 5 volts and the cpu if the switch  gets flipped the cpu will be powered directly   by the psu following graph visualize how the  voltage on the cpu would look like in this case   the blue line represents the voltage over the time  and due to physical restrictions the voltage will   not instantly rise to 5 volts when the switch  gets flipped but will take such a slope nearing   5 volts after a while please keep in mind that  this whole visualization is very inaccurate   simplified and exaggerated it serves only for  explanational purpose and is not a precise model   anyway this voltage would be just fine if the  cpu would need 5 volts but what if it needs   for example 3 volts let's install a voltmeter  between the switch and the cpu and turn on the   switch again but now instead of just waiting  until the voltage gets up to 5 volts we will   turn off the switch again when the voltage on the  voltmeter reaches 3 volts certainly the voltage   will drop to 0 after a short period of time again  so it doesn't sound too exciting so far however   what if we repeat the game and turn the switch on  again when the voltage reaches let's say 2.9 volt   now the voltage will rise until we switch  off at 3 volts again and so on and so on if   we keep switching on and off all the time we  will be able to keep the voltage between 2.9   and 3 volts and so convert 5 volts to nearly  3 volts as required this is the theory so far   but in reality such switching must happen very  often from hundreds to millions times a second   neither a human hand nor any mechanical switch  would stand such a stress so in reality something   like this will be used again this is a very  simplified representation which should just   explain the idea instead of a mechanical switch a  transistor is used which will be turned on or off   by a so-called pulse with modulation controller  or short pwm instead of the voltmeter there is   so called feedback connection which is used by the  pwm controller to measure the voltage on the cpu   and to decide if it's time to switch the  transistor on or off the principle behind   this circuit is the same as i explained before  with a switch and it is called switching power   regulator such a solution is extremely widely used  you will find it on main boards graphics cards   in every notebook and even in the usb power  supply which you use daily to charge your phone   it is very useful to know and understand how it  works because it is the number one reason for   failing of all the devices i just said but why  did it become so useful where is the benefit to   a linear voltage regulator which i presented  in another video well the transistor heats up   when it is turned on and cools down again when it  is off so instead of keeping it on all the time   like it is done with the linear voltage  regulator which has to dissipate a lot of heat   this switching voltage regulator gets always some  time to cool down a little bit between the cycles   it still gets quite warm but it is by far more  efficient than a usual linear voltage regulator   and so needs a lot smaller cooling solutions  however switching voltage regulators do also have   an important drawback they generate not a steady  voltage like linear voltage regulators do but such   a jigsaw pattern to work around this effect  a little bit usually a big capacitor is added   to the circuit which acts as a low pass filter  and flattens the jigsaw character of the output   again this is very inaccurate representation and  in the reality it is usually not as perfect as   the pink line shown here but you hopefully get the  idea in reality this line is still not as stable   as it sometimes needs to be that's why switching  voltage regulators are rarely used in the high-end   audio equipment furthermore i was talking only  about the voltage so far but flipping a switch   also produces high current peaks which are  usually filtered by an additional inductor   so if you need a hint if a voltage regulator in  front of you is a linear or a switching one then   search for a big inductor nearby if you see one  that's probably a switching regulator a linear   regulator on the other hand would probably have a  bigger hit spreader instead also the 3 volt output   is just an example since the pwm controller can be  certainly configured for different output voltages   like 2.8 volts which we need for the pentium mmx  so far the theory and with all that information   i designed a pcb which should fit into this vrm  connector like that so let's build and test it   since the pentium mmx draws almost 6 amps  of current i found it complicated to find   inexpensive parts which would stand that load  so i decided to use a pair of dc dc converters   each designed for up to 3 amps and  since the cpu draws maximum 5.7 amps   we should be well below the limit this decision  cuts the cost at least in half probably even more a is for the connector i have the simple 2x15  female header here but i will not solder it now   instead i'll solder to wires for a basic test  first where i can connect a power supply and   see if the regulator does what it should at all so  let's use a workbench power supply for the initial   test i'll set the input voltage at 5 volts and  limit the current to 500 milliamps let's connect okay nothing exploded and  the parts remain stone cold   the voltage on the power supply remains  at 5 volts and the current is only at 38   milliamps that means that we have no short so  far that is a good sign let's test the voltage okay 4.8 volts that is totally wrong i  was expecting something around 3 volts i made some jumpers up here to set other voltages   let's switch it a bit and see if  something changes okay still 4.8 volts and again 4.8 volts unfortunately these switches don't change  anything at least the ics remain cold and   there is something on the output  that's already a step forward and   i would like to quote a good friend of mine if  something works on the first try you probably   overlook something very important so i'm  kind of glad that it doesn't work yet so i found the issue and the culprit was  just as so often sitting on the chair   unfortunately i ordered the wrong dc dc  converters i installed those xl2596s-12e1 which is a converter for 12 volt output but i  needed these xl2596s dash adj1 where adj stands   for adjustable since i want to adjust the output  using the switches and as you see i reordered the   right parts and prepared the vrm module for  replacement now let's install the right ics okay the adjustable converters are now  in place let's give it a second try   the power supply is again connected  and we seem not to have any shorts   the module is now set to 3 volts let's see  if we get something around that on the output okay that looks much better we now have  3.1 volts that is a bit too high but   since there is currently no load maybe it  is slightly off let's set it to 2.9 volts and here we have 2.9 volts indeed  let's continue with 2.8 volts   and it is on point what's next 2.6 volts and we are at 2.6 well almost  2.7 actually now 2.5 volts yeah almost 2.5 volts as well 2.4 looks also quite good and 2.3 volts is okay-ish and last but  not least to point to waltz well it's almost at 2.3 volts  but still doesn't look too bad   well those values can vary a little bit as soon as  the module is in the main board and is under load   but i think they are close enough for a prototype  and they are not too easy to get exactly i made   such a table with formula for the resistors and  voltages to calculate the right values which are   currently installed on the module i guess these  values can be adjusted in the future if needed   for now i can see that the voltage is close  enough so we can move on with the tests now the leads for the workbench  power supply can be removed again   and the connector can be finally soldered to the  module back to the main board as i said it has   no header to connect the voltage regulator module  so i will simply use two rows of such pin headers   and solder them into the main board the board has  big copper ground and power planes in the area   around the vrm so i'm heating up the board with a  hot air in one hand and use this soldering station   in the other to get the solder out of the holes  i'll insert the pins directly into the module   so i easily can cut the length and align the  connector pay attention that there is no key so   the direction is important pin 1 on the vrm is  ground just like in the documentation by intel   plugging the module wrong way around  could damage the module the cpu   and the main board so it is very important to  double check the orientation and the alignment and here it is ready for the next test  i would like not to burn the cpu now so   i'll add a wire into this cpu socket  on the vcc pin let's see what we get the main board is on nothing exploded so far and  we have the required 2.8 volts on the vcc pin   that gives some hope   and i think it's time to get brave and insert the  cpu for the real test i'll also add a pc speaker   to be able to hear some post beeps if there will  be any the pentium cpus didn't get too hot very   fast and could even work couple of minutes without  a cooler completely i'll keep my thumb on the cpu   to feel if it gets hot too fast so maybe i will  be able to turn it off quickly fingers crossed and i'm very sorry for being so evil but this  video already became quite long and i'm tired   so i thought why not making a cut here i hope  i could wake up your interest just a little bit   if you are curious to see if  the cpu survived this treatment   and if the module could handle the load  please tune into the second part of this   video and if you like this one believe  me you will love the next one as well   don't forget to subscribe and leave your  feedback below and so far thank you and goodbye

Channel: Necroware

Views: 104,766

Keywords: retro, hardware, soldering, repair, review, nerd

Id: CMiGVQbMC5U

Length: 27min 47sec (1667 seconds)

Published: Sat Nov 20 2021