Let's upgrade an early Socket 7 board beyond the limit (Part 2): VRM, MMX, Testing
hello and welcome in my last video i showed a painting mainboard from the pre-mmx times which i wanted to upgrade to a painting mmx this main board doesn't support dual voltage required by a pentium mmx out of the box so i decided to design a voltage regulator module which would make it possible to use that cpu on that main board if you didn't see that last part i suggest to watch that one first and then come back to this video first of all a small remark about the explanation of a switching voltage regulator which i made in the last part unfortunately it was not right about one detail i said that the transistor cools down when it is off and warms up when it is on however the transistor actually warms up only in the moment of switching and as soon as it is completely on or off it can cool down again the explanation for that would be a bit too much for this video that's why i made that shortcut in the last part which obviously was wrong and was instantly spotted by many mindful viewers and i would really like to thank you all for the corrections so where did i stop oh yes i want you to test the pentium mmx200 cpu in this main board for the first time so let's do it and that's it thank you for watching just kidding as you probably heard there was a post beep which doesn't actually sound too good it's kind of choppy but at least the cpu doesn't get too hot and the vrm didn't explode that's a good sign i guess i mean exploding vrm is usually a bad sign don't you think okay i will need a second hand to make some measurements so let's mount a cpu cooler a graphics card and a pulsed analyzer let's give it another try at least some numbers are running on the post analyzer so the whole thing is not dead um oh and i already see an image on the display but let me check the voltage first yeah almost 2.8 volts and stable very good and here is the display as you see we have an image the cursor is blinking so it doesn't totally hang but it also doesn't continue so far what's odd is the detected cpu a pentium 180 megahertz that is definitely wrong but maybe we can change it using a newer bios version and where can we get the latest bios exactly the ultimate retro project former known under the name uh-19 it has recently moved to a new domain ultimateretro.net and got a lot of new and exciting improvements since i introduced it last time this is the best place on the internet if you are searching for images documentation bios dumps and more for the retro main boards let's search for the p slash i-p-55 here is the list of matches and we need the xeg version here we go the images the documentation and here is the latest bios from 1997. perfect by
the way if you have a main board which is not yet in the database or which misses pictures bios doms or something else please consider to contribute all the data you can provide to this project it is made by the community for the community and some really nice guys are working hard in their spare time to make it even better so they need all the support they can get and i wrote the bios using my tl866 ii plus already let's replace the old one and see if something changes by the way i always try to use another p-roms and keep the original untouched at least as long as i'm not sure that the newer bios works as it should afterwards i still can override the original piro so let's see if it helped oh the post beep does already sound better and is not as choppy anymore and would you look at that the new bios helped indeed and there is penthume mmx message laughing at us from the display however the 166 megahertz clock is not right the jumpers are actually set to fsb of 66 megahertz and multiplier of three so the overall speed should be 200 megahertz so i tested off camera a little bit and it turned out that it makes no difference how i set the jumpers for the multiplier the cpu always remains at two and a half and 166 megahertz however the frontside bus seems to be right at 66 megahertz let's try the pentium 133 once again and see if the multiplier jumpers work at all this cpu needs single 3.3 volt voltage so i have to remove the vrm and here is the first good news this modification is completely revertable all i have to do is to remove the module and put the jumpers across the pins as they were set originally and as i showed it in the last video so now the cpu should run again normally at 3.3 volts just as before so both multiplied jumpers are currently on set and due to this table here we should get a multiplier of 1.5 with a frontside bus of 66 megahertz we should now see 100 megahertz pentium detected
and here we go 100 megahertz indeed so let's set the jumpers to open closed which should mean two times 66 megahertz or 133 megahertz core frequency and we are getting 133 megahertz indeed so the jumpers are absolutely okay but why don't they work with the pentium mmx well after some investigation i found following first of all a well-known fact that the pentium cpu has two pins bf0 and bf1 which are used to configure the core clock these pins can be set either to 0 or 1 and so different front side bus multipliers can be configured so far so good but here is what i found in the intel paint you motherboard design guidelines on the painting processor the bf1 and bf0 pins have internal pull up resistors but on the pentium processor with mmx technology the bf1 pin has an internal pull up resistor and the bf0 pin has an internal pull down resistor the jumpers are pulling the bf0 and bf1 pins low and that works only on the non-mmx pentium cpus because they have internal pull ups on both pins and since the pentium with mmx has an internal pull down on the bf0 it is already always low no matter how i set the jumpers so all i have to do is to add an external pull up resistor on the bf0 pin to be able to set it to high if needed apparently the documentation also says that the value of the external pull up resistors used on the bf0 and bf1 pins should be 2.2 kilo ohm or less okay let's put back the vrm module the cpu is already in place let's set the jumper to the multiplier 3 and let's give it another try and we have a 200 megahertz pentium mmx a cpu which was released a year after this main board shows some signs of life in it let's put back the cooler and i will also insert this compact flash to ide adapter with a card where i have ms-dos pre-installed with some tools and games and we are in dos let's start something for example doom to see if it is able to finalize the benchmark at all meanwhile let's test the voltage the voltage regulator is barely warm now but it didn't run too long yet the voltage is set to 2.8 volts and we are getting about 2.77 volts that's good enough but meanwhile i recalculated the resistor values and the next revision of the regulator should deliver even better values anyway 2.77 volts is absolutely within this pack so nothing to worry about and we got about 1 000 real ticks however the performance values are unimportant now all i currently hope to see is some kind of stability the species detects pentium mmx200 as well and the memory throughput test uses the mmx instructions too i started now pc player benchmark which i'm going to let running for a little while let's see how stable it can work over a longer period of time this linear voltage regulator is usually used to power the cpu and it gets very warm and often fails because of that however since we have now a dedicated vrm for the cpu core voltage this on-board regulator is only used for the i o voltage and therefore remains stone cold currently it is around the room temperature about 19 degrees celsius what i would also like to test is the output ripple it is important to have it as low as possible for a stable work and as you see the ripple is about 50 millivolts not super good for 2.8 volts but also not bad i think painting mmx is designed to stand up to 100 millivolts so 50 millivolts is an acceptable value my oscilloscope seems not to be able to catch the frequency exactly but it seems to jump around 150 kilohertz which absolutely corresponds to the fixed frequency of the dc dc converters which i'm using here let's leave the system running a little bit and see how it behaves so about one hour passed the pc player benchmark is still looping on the screen nothing bad happened so far let's see how hot the regulator is yeah almost 50 degrees celsius although absolutely in limits still it is quite hard to be honest the cpu seems to be cooler and it blows the air directly onto the voltage regulator so it gets cooled by the fan as well but still i think this was a successful first run i would say the whole thing didn't explode it is not too hot the voltage is okay the ripple is also within limits and yeah the pc player benchmark runs also since an hour without any issues so the pentium mmx 200 megahertz is running with 66mhz bus and multiplier 3 cool and stable sole farm this is already a huge upgrade over the 133 megahertz pentium which was in this board before and the mmx is even beyond the original specs of this main board but the question is how far can we get can we push the limits of this board even further of course we can initial idea of this project was to get the voltage below 3 volts so this main board will be able to run with cpus other than the non-mmx pentium versions and this seems to work so far however the voltage regulator can provide voltages down to 2.2 volts and which cpus do run with such voltage i am
happy to present to you an amd k62 i have this cpu in different frequencies and this one is a 366 megahertz but the speed doesn't matter now interesting detail is that this cpu runs with 2.2 volts the k62 cpus were specified to run with up to 100 megahertz frontside bus but our board can only 66 megahertz so let's start low and see what happens if we just switch the intel pentium mmx200 by this amd k62 the multiplier jumpers will remain at 3 and i just have to set the switches on the voltage regulator to 2.2 volts this is when all three switches are on everything else will remain in the same as it was with the pentium mx200 before and would you look at that the cpu starts at 200 megahertz it says amd k6 and not k62 but that doesn't matter the bios just doesn't know the cpu id because it is simply too new let's check the voltage first yeah we are at around 2.2 volts and that is fine and we can continue the test as you see in dos the software detects the cpu correctly even if the buyers didn't know that it is actually an amd k62 the cpu id tool says that it is an amd k62 running at 200 megahertz a multiplier 3 and 66 megahertz front side bus everything is right let's see what speed says tells us yeah and also here we see k6 to 200 megahertz it even reports active 3dnr feature the benchmark values don't mean a thing now since i didn't tweak anything in bios whatsoever this is something what we can check out later let's run some pc player benchmark and test the vrm temperature so far so after about 10 minutes of pc player benchmark we have 47 degrees celsius this is not nearly a critical value i still can easily touch the regulator okay let's take a look at the k62 datasheet in particular at these three pins pf0 bf1 and bf2 now compare this pin out to the one from the intel pentium where we see bf0 and bf1 but instead of bf2 there is an unconnected pin well as you probably remember those bf0 and bf1 pins were used on the pentium cpu to select the multiplier and the bfz european is the one where i soldered the pull up resistor previously to make it work within intel pentium mmx so amd k62 has obviously three of those pins vf0 bf1 and bf2 you see k62 was primarily made for the super socket 7 main boards which usually were prepared for much wider cpu clock ranges so two pins were simply not enough to control the multiplier up to six so compared to the original intel pentium amd introduced the third bf2 pin in addition to the ones which were used on the intel pentium previously and in this table you see the three bits which represent the bf0 bf1 and bf2 pins on the cpu in little indian manner so the most tried bit is bf0 and the most left is bf2 the cpu pins bf0 bf1 and bf2 on the cpu were connected to the three jumpers on the later socket 7 main boards and so the user could configure the multiplier unlike on our asus mainboard here which due to its age knew only the bf0 and bf1 there was simply no jumper for the bf2 to remain compatible amd equipped the bf2 pin internally with a pull up resistor so on main board which have no ability or no jumper to set the bf2p into low this pin was always pulled high so looking at this table we have to concentrate on the first four entries where the first bit bf2 is set to high or is one with other words other setting would be only available if we would modify the main board and add an additional jumper to the bf2 pin on the cpu i'd prefer not to do it now that's why let's ignore all multipliers in the second half of this table for now we already know the multiply 2.5 which would set the cpu clock at 166 megahertz and 3 which would
set it to 200 megahertz this is what currently the jumpers are set to 0 1 by the way 0 means jumper is set and one means jumper is not set let's take a look at the next possibility the multiplier 3.5 it expects bits 1 1 so both jumpers are not set let's remove the jumpers and give it a try and as you see the board reports amd k6 233 means 66 megahertz front side bus and multiplier 3.5 that is correct even if the k6 is actually a k62 but i explained already why it is shown wrong here and again in the cpu id tool we see k62 at 233 megahertz detected with the multiplier 3.5 and the front side bars unchanged at 66 megahertz and let's play the same game here again i'll fire up the pc player benchmark to put some load onto the cpu and test the temperatures after some minutes and again about 15 minutes passed and we have 55 degrees celsius an absolutely acceptable value the vrm feels very warm but is far from burning and the onboard linear regulator for the 3.3 io voltage is still stone cold so there is no stress on it at all very very nice let's go back to the mdk 6 data sheet once again i promised to go beyond the limits of that asus mainboard and i would say we already stroked that aim however we can do even much better take a look at this entry for the multiplier 2 the datasheet says here that two or six with a remark down here the ratio selected is dependent on the stepping of the model eight the two times ratio is supported on the model 8 7 to 0 whereas the 6 ratio is supported on the model 8 f through 8. well the model 8 refers to all versions of the amd k62 but as far as i believe the 7 through 8 stepping versions were the cpus with the id 26050 and due to the datasheet they supported the multiplier of 2 but the versions with the id 26351 which i have belonged to the f through 8 stepping and supported a multiplier of 6 instead of 2. so what we have to do is to set up the multiplier of two
on the board and the cpu should actually run with a multiplier of six let's first set the jumper but that was not everything what we have to think about the multiplier of 6 and 66 megahertz front side bars would mean that the cpu would run at 400 megahertz and with rising frequency the cpu gets a lot more power hungry take a look at this table the amd k62 which i have will draw up to 11.25 amps at 400 megahertz and that is a real bummer since i designed this vrm initially for the pentium mmx and not for such a cpu which would run at 400 megahertz with this design the regulator is specified for up to 6 amps only and 11.25 amps would be almost twice as much as it can deliver but who would i be if i wouldn't try it though i built multiple of these modules and i think that i can sacrifice one for the signs however i expect it to get very hot and i have to make some preparations i have here some small heatsinks which hopefully can cool down the ics on the module a little bit so i glued a pair of that heatsinks already on one of the modules i already tested this module behind the scene with 2.8 volts and it works flawlessly what i need to do is set it to 2.2 volts again this is where all three switches are on
so the second thing is an additional cooler which will blow over the regulator i will connect it to the 12 volt fan header of the cpu so the cpu fan will not work now but i hope that this bigger fan will be strong enough to keep the voltage regulator and the cpu cool for some time at this point i would really like to make an important disclaimer not only about this project but about everything i do on this channel because in the last time i read more often that people assume that i'm an expert in electronics that is absolutely not the case and can be even a dangerous assumption i am quite an experienced software developer but i am in no way an experienced hardware developer all you see on my channel the diy projects the repairs everything is just my hobby i had a lot of theoretical electronics in the university many years ago but only as much as a computer scientist needs to know i never collected any practical hardware development experience in my job and all you see me doing here is a result of my self-development i just like to learn new things and as everybody who is learning i also do a lot of mistakes i always try to do my best but sometimes i go the wrong way i think it's normal and i try to understand and to explain many things as simple as possible because i think that simplification is the key for the solution of complicated problems but i also tend to oversimplify things sometimes and drop too many details so always take it with a grain of salt and please never see this channel as a tutorial i'm just glad to be able to share my hobby with you and i always hope for a nice experience exchange in the comments so everything is ready let's see if it explodes okay we have a pulse signal and on the screen we see amd k6300 that is definitely not right but again maybe only because the bios is now extremely confused and here we see 375 megahertz which is again not what i was expecting to see but let's continue nothing exploded so far i think that i don't have a lot of time but the cpu id tool clearly reports an amd k62 400 megahertz with a 66 megahertz front side bus and a multiplier of six on this board which was specified for a non-mmx cpus with up to 200 megahertz only this is amazing let's see what speed c says also here amd k6 400 and down here look at that we are almost at the performance of a pentium 2 400 okay now i go with all in and start the pc player benchmark let's let it run and test the temperature of the voltage regulator 72.8 degrees celsius that is quite hot already if you consider that it was cooled directly by a strong fan okay it is very hard to touch as well and as you see the benchmark is still running i am really surprised and a little bit disappointed that nothing exploded but i think i'll stop the experiment now because i would like that this system survives since i have further plans with it okay first important note is that as i started to work on this project my main aim was to get the pentium mmx200 running on that main board and i would say that i reached their target however pentium mmx pulls only up to 5.7 amps and i designed the voltage regulator for that current in the beginning i unfortunately didn't think that i also could try to run a k62 cpu on this mainboard i thought about that first as i was searching for the bios update which would be able to recognize the pentium mmx and as you saw it would work theoretically shooting the system performance to the moon but the voltage regulator was just not made to handle up to 11 amps and so i currently can't go that far with this revision second possible issue with this design is actually due to my lack of experience i used 2 switching regulators each specified for up to 3 amps to be able to handle up to 6 amps in total my concern was from the beginning that trying to control the same output from two regulators could end up in a fight between the regulators and although i didn't encounter any issues during my tests i still have a bad feeling about it and some viewers have the same opinion in the comments to the last video so i think that i have to do something about it as well as you see this revision 0.1 is the very
first one and fortunately i already designed the next one which should be able to stand the current of up to 15 amps and do everything with one dc dc controller and a beefy mosfet but that one currently exists only on paper and i'm going to order the new pcbs and parts in the next days so i am glad to announce that there will be another video to this topic in the near future where i hopefully will get as much out of this system as i can i'm also going to upgrade the cache and run a lot of benchmarks so if you are curious don't forget to subscribe there were also some questions in the comments to the last part if i am going to open source this project and the answer is yes just as always i open source all of my projects and this one will be not an exclusion however i decided to not open it yet since i would like to provide you guys the better revision first which hopefully will also work for the higher current so please stay patient a little bit longer and hopefully in the end i can provide something what is even better as i already said i really like to learn new things and for me every day where i didn't learn anything is a wasted day so if you have some good information source where i can read and learn something about one particular topic please consider to share it with me otherwise i just hope you enjoyed this video and would be glad to have you on my channel again thank you and goodbye
2021-11-29 08:38
