Let's upgrade an early Socket 7 board beyond the limit (Part 3): S7-VRM progress and new revision

[@darthtripedacus1]

Thanks for the video and I am sorry to hear of your tragic circumstances. We appreciate you.

[@alberthofmann420]

What were those circumstances if I may ask?

[@aprilliac]

@albert, i heard someone in his family was stricken with nunya or possibly ligma.

[@dieSpinnt]

@@alberthofmann420 They were tragic. Period!

[@Shmbler]

Now that's a proper VRM ;-) Would be nice to see the actual current draw of the CPU in test so people can see why a linear/async buck regulator won't do.

[@necro_ware]

I think I showed it in the second part already, where I had to blow a lot of air to keep such a buck converter below 70°C, in the winter, in a cold room :) But in the next part I'll address it a little bit.

[@pavelfara9333]

This VRM is better than the original design on many boards of that time 👍. I have an Octek board which is from the "we want support the MMX but Intel not provided the righ specs yet" period. I have installed a Pentium 233MMx in to it. It was running but the VRM got hot as a toaster. 🙈

[@Guillermo_XT]

Immer wieder ein Genuss sich diese Videos anzuschauen selbst ein Laie kann den Erklärungen folgen 🙂👍

[@necro_ware]

Vielen Dank 👍

[@krzbrew]

So glad that you are back. I have discovered your channel while you were absent and have just begun missing new content.

[@Damicske]

Now you have more pcb real estate, at a multi turn trim pot on the feedback signal so you can really dial the voltage ;)

[@rkurbatov]

That's simply amazing. I love the way how enthusiasts like you give a new life to the old piles of interesting hardware, improving everything they touch :) And covering the most interesting decade of retro hardware - 1990-2000.

[@borisjevic6338]

Btw, love this series for the socket 7.

[@tomasz89g]

It's encouraging to see us software-types can bow in to the hardware world too, thanks, I'm gaining confidence in tinkering myself from watching your learning journey.

[@villefilho]

I´ve never seen an actual VRM module for desktop class mobos... that´s really interesting, thanks for the video!

[@into-db3oe]

Cool, I'm waiting for the next version of VRM.

[@minombredepila1580]

Thank you very much for providing the community with such nice & useful implementations. I still use your Dallas replacement !!!!.

[@daw7563]

Cool project. It would be interessting to compare powerdraw (total system power). Probably just a couple of watts but still interessting.

[@necro_ware]

Yeah, probably it will draw less power, because a switching voltage regulator is much more effective, than a linear one. However, I would have to compare some low speed Pentium, since the on board LDO is not able to power anything else. And at that speed, the difference would be peanuts.

[@goeland4585]

Yes! perfect start to the weekend :)

[@osamely_varan]

this is kind of higher level than usual. but still love it. and i even did understood every third sentence at the start :D

[@BrassicGamer]

What a great project. I admire your dedication to learning.

[@VShuricK]

fantastic job. One question - why dont use polymer capacitors? They have two-three times more ripple current and better ESR, what is very good to use in DC/DC and cheap enough today.

[@LaserFur]

some regulators (like the simple switcher family) rely on the output ESR to be stable. On one design I added a 10 inch trace between the regulator and it's output cap and the rest of the circuit. On a recent design I did use a polymer cap on the output, but then I had to follow it with a ferrite bead so that the capacitance of the rest of the circuit would not lower the ESR too much.

[@Alex.Adametz]

Designing such things is a real challenge. Great job!!!

[@angelamcmahon]

Love your videos. I hope whatever that tragic thing is is getting better now.

[@AndyHippoR]

Looking forward to seeing this project completed. I've been tinkering with something similar myself, but using an off the shelf adjustable DC/DC converter and some prototype board. Would be nice to try one of these on my GA-586ATM/P

[@felixokeefe]

You could make slots in the VRM module so that the hooks of the module socket hold the module in place. I'll try and find a picture of an original module for reference.

[@felixokeefe]

Or this one which is more typical for a mmx VRM module pcrebuilding.altervista.org/images/resized_my_pcrebuilding/5a0d50eeec4be_p1220367_2.jpg

[@ruben_balea]

The Semtech MP54C and MP55C datasheets are available on popular datasheet websites and include the module dimensions

[@necro_ware]

Yes, I thought about it already, but to make it properly not only the holes have to be made, but also the PCB has to go slightly below the connector. I measured it and decided not to bother yet, since it is sitting quite tightly already. May be in the future, should it be really needed.

[@snakezdewiggle6084]

A great video, thank you ! Socket 7 and socket 8 AMD run perfectly Under Volted and Over Clocked, with about 6% increase in temperature. ;)

[@TheFalschspieler]

Happy Easter! Thank you for the Video!

[@RetroTinkerer]

Hi, this is so cool, thank you so much for sharing your design. I have at least 3 boards that would benefit from this even when running 3.3V CPUs it seems like a good idea to offload the old and stressed VRM!

[@ddognine]

Way to go! A lot of work to develop a VRM with such a wide range versus splitting the range in half and having a version A and B.

[@TheOldGraff]

Awesome work!

[@georgeaiaskaridis5812]

Incredible work! Looking forward for the next video but, can you explain what issues you had because of the vias? How did they affect ripple?

[@necro_ware]

The DC-DC controller is extremely sensitive and even some parts around it should sit as tightly as possible. There are some very strict layout considerations. If you start to jump over the layers back and forward, this can amplify some parasitic effects. It is suggested by the datasheets to avoid it if possible.

[@georgeaiaskaridis5812]

@@necro_ware Thanks! :)

[@SeanBZA]

@@necro_ware Also for high current you need lots of filled vias, no soldermask over them, and solder paste placed on them, so it will reflow and fill them. Larger vias also have better current capacity, so non soldermasked and untented vias of around 0.8mm will work better, and keep 1mm between them, and stitch as many in as you can. Also go to a 4 layer board, and devote the one inner layer to GND and the other to Vcc, leaving more room on the top and bottom for current carry ability. At the connectors put vias to short traces to connect them in addition to the inner layer connections, and leave a good clearance around all vias that do not connect to either plane, but make all vias non blind. More layout room, and better thermals, especially for the upper mosfet that will have a lot of heat, and the lower one can still couple thermally through the board to the inner layers, even though there are vias connecting it to the bottom layer, you will still spread the heat better with more copper.

[@craigkarsten6719]

Thanks for your work and for sharing this project. I have a socket 7 board I'd like to try this out on.

[@DnaX]

Compliments for the new design! I think it's really well done and useful.

[@robbiesz]

This project would have been the perfect collab with Great Scott. His and your expertise put together would be something out of the "retro world"! :-)

[@marcinkrupinski3782]

Yes, my day is complete.

[@nachtrave5172]

I was literally just experimenting with a 486 DX Overdrive PSU circuit not too long ago. Really cool to see some similar designs and such. =)

[@ninedogs2418]

Great job. But I would place the mosfets on the other side of the PCB so that the cpu cooler blows them.

[@necro_ware]

Most of the heat on that SMD parts goes into the PCB, which contains via grids on this VRM to transfer the heat to the other side. Furthermore, heat shouldn't be an issue on this VRM, so far even at 10A it remains stone cold.

[@Metalliferous]

Great video as always, can't wait for this to be released!

[@mach1stang]

Very awesome. Looking forward to getting one of these for my old IBM Aptiva. It has a 120 mhz p1 I'd like to put in mmx 200 in it. This would do the trick.

[@fft2020]

Another awesome video as always !

[@JohnSmith-iu8cj]

Awesome work you do for the retro community!

[@smallmoneysalvia]

I would absolutely love a video series describing switching dc power supplies and design considerations

[@vermilionvoyager2470]

Inspiring, keep up the good work sir!

[@rs-qq3os]

Great project , keep going !!

[@JosepsGSX]

I love these videos. The series is fantastic, and every time the a new setup boots, a smile goes to my face. It is pure joy watching that dot of history been alive again. Also, the jump into the VRM design and the results you are achieving are astonishing. I will love to see the next part. Absolutely. Thank you!

[@vswitchzero]

Awesome project! This is way more impressive than any onboard regulators I’ve seen on socket 7 boards, that’s for sure! I’m actually testing an Evergreen Spectra 400 at the moment, which is a K6-2 400 with on-chip regulation for upgrading older socket5/7 boards. The regulator bits aren’t very large but I’d be curious what level of ripple it has and what you thought of the design 🙂 .. great video as always. Looking forward to constructing one of these for an old 430FX board I have with an empty regulator socket 👍

[@cann0nf0der]

Awesome project I love it!! Thanks for sharing this with the community 👍

[@xXBeefyDjXx]

This is amazing progress, I recall the original builds and love your work, a true love letter to retro tech!

[@PardusRain]

I love this "neo-retro" modding, it is very interesting to see how far things can be pushed with older parts.

[@TheWizardHW]

Amazing work!! I love this channel!!

[@peacefrog5099]

I'm very excited 🎉

[@tony359]

Such a cool project, thanks for sharing! Very interesting that the MSI and Asus have the connector in different places but thankfully you always end up with the components away from the CPU! What would happen if the VRM is mounted the other way round? Just thinking if there would be a way to implement some safety features. Sorry to hear things haven't been great for you, I hope things are getting better.

[@thomasmurphy3927]

Hope your good man. Keep it up

[@TekRaiderz-qh5ic]

Great work. thanks for sharing

[@mogwaay]

Great work on the VRM it looks great. As another software engineer turned electronics hobbyist analogue power stuff really confuses me too so awesome job.

[@SeanBZA]

Output side adding in a LC filter will help, simple to add by putting the high current inductor, with a damping resistor, probably 1R, across the inductor. Inductor added between the first capacitor and the output capacitor. Also add in some 10uF 25V ceramic capacitors, likely in 1205 format or larger, across the existing filter capacitors. Yes they will probably only have 2uF capacitance, but lower ESR than the big capacitors.

[@ale6242]

Very cool stuff!! Happy to see more content from you and hope you are doing well!

[@2dfx]

Maybe not wider, but a TALLER board may facilitate the use of through-hole parts for your passives that way.

[@zarkeh3013]

I could be wrong, but those clips engage with holes in the VRM's PCB which are missing on these. It's been a long while since ... but I do remember there being Holes.

[@BenState]

Amazing mate, amazing!

[@PROSTO4Tabal]

amazing project, keep those videos comming

[@JasonsLabVideos]

OHH this stuff brings back memories !!

[@jonatas464]

TKS a lot. Regards from Brazil. Could you explain how some circuits know to auto detect voltage?

[@necro_ware]

Basically the CPU says to the mainboard, what it needs. Pentium II for example has hard wired pins, which are connected directly to a DC-DC controller, so the CPU knows what it needs and controls the voltage. Some earlier systems, even 486 had also sophisticated solutions, where the regulator raised the voltage slowly and measured when the CPU starts, but I don't really like such solutions, since they are unreliable and hard to repair.

[@natr0n]

very nice

[@BenState]

Sorry to hear you had tragic circumstances mate

[@sdjhgfkshfswdfhskljh3360]

Maybe I'm wrong (because I'm noob in electronics), but I think that it is also important to measure output waveform with proper load (CPU in this case). Maybe CPU is not much different from resistor, but with capacitive and inductive loads voltage stablilizers sometimes becomes crazy - and instead of DC start producing AC.

[@necro_ware]

I did of course measure the ripple with a CPU already as well. In the next part I'll talk about it a bit.

[@dabombinablemi6188]

Some board manufacturers in the slot 1 days ended up using huge 3300uf capacitors (that interfere with the back of my PIII 650's heatsink), with a couple of smaller polymer caps to get a similar effect to what Asus achieved on their P2 slot 1 boards with 14 (or more) 1000uf capacitors. Same board manufactures (eg. Intel with their SE440BX-2 models) though would cost down the boards, removing or downgrading features and not using polymer caps.

[@rallyscoot]

I hope that the bigger capacitors can be a little bit lowered with other size, so they wont stick out to much.

[@necro_ware]

Well, yes, it is possible to take smaller capacitors, but they should be at least 2mF each and that's already quite tall.

[@awd42]

@@necro_ware Maybe use axial capacitors instead of radial? Might need a taller board, but it could be thinner.

[@peteregan9750]

interesting videos - keep it up !

[@SianaGearz]

These things aren't so simple to design and i don't think too many EEs bang these out all that smoothly and easily, especially not with such wide output requirements. Usually it's a pretty specialised field. The MOSFETs actually cool through the PCB better than head on. Because the die is near the PCB side and basically sits on that thermal pad that you have vias to. On top there's plastic, a lot of plastic which is not all that thermally conductive.

[@Damicske]

That's true, more via's on thru the board and not coated on the backside is better for cooling. Make it so you can put a small heatsink on it (search for gpu memory heatsinks)

[@necro_ware]

Not necessary, I'll explain why in the next part.

[@viciious1234]

Not sure whether the following suggestion makes any sense, but would adding a fuse on the output of the VRM make any sense?

[@tristankordek]

At the output more low-capacity smd capacitors, for example ceramic, and they should help with ripple. On the back of the PCB you have a lot of free space, these capacitors also increase the heat radiating surface.

[@necro_ware]

Of course there are already multiple ceramic caps on the input and output to remove the noise.

[@EvilTurkeySlices]

Would be cool to have one of these that goes in-between the processor and the board for motherboards with no VRM header.

[@dawee5345]

Thanks for the exciting video. When do you planning your next repair marathon video? I really interested int those footages to see the skills needod for this kind of repairments

[@ArtemAleksashkin]

О, наконец-то продолжение! Спасибо 11:07 да спс за ответ

[@flecom5309]

with a lot of modern power supplies not having much in the way of current in the 5V rails would it be possible to add some jumper headers to select the 5v incoming from the motherboard or an external connector that you could use to feed 12V (usually abundant in modern power supplies) into the VRM?

[@menotyou8369]

More vias would help to dissipate the heat through board and out through the copper on the back side of the board.

[@creopard]

Seems all this newly gathered knowledge paves the road for a complete "Necroware Retro AT(X)-Mainboard"?👍😉

[@necro_ware]

lol, that's still a long road to walk, if we want to make it from scratch with all new parts.

[@Ll3maFN]

whats the song in this video?

[@necro_ware]

It's custom made for the channel. May be it will be released on bandcamp one day together with other music, which was made.

[@Ll3maFN]

@@necro_ware cool

[@Space_Reptile]

little idea for the connector: if there is a redunant/unconnected pin, plank it off and use it as a key, that way its impossible to reverse the module and destroy the board

[@volodymyrzakolodyazhny]

Author may try use of a polymer caps instead of 'usual' Low-ESR caps, at least at output of a regulator. Polymer ones have smaller ESR and can handle much large ripple currents. And they can have smaller size too.

[@ruben_balea]

Version 1.0 is going to be a Very Refined Module 😉

[@necro_ware]

I'm trying to deliver something reliable. I don't want, that people around the world start to destroy their mainboards and CPUs by using something what I badly designed.

[@ruben_balea]

@@necro_ware I know, but I couldn't resist making a pun 😅

[@tomekrv942]

I saw the first part of this. It was long time ago. I needed rrecently something like this for my fic 486vip-io2 so I used some cheap converter and it is working but first tried to build my own and I know that it is not easy.

[@tdsangel]

i simply love your content! even in german. ;D

[@dolphhandcreme]

Make some cutouts on the sides so the lever-hooks can grab into the module!

[@husbyugur77]

ma man ... you are just awesome ty sm

[@volodymyrzakolodyazhny]

Yes, as electronics engineer I can confirm that designing a DC-DC regulator is very easy. I usually make one or two before my breakfast just while cleaning my teeth.

[@necro_ware]

:)

[@NightSprinter]

Would you suppose this could let me run an MMX233 on my AST that has a P133? Would love to make this a DOS/Win95 machine for a Voodoo1.

[@Miltiadis_Vouzounaras]

Great video (as always)!!! How hard it would be for the community to make a new open source hardware motherboard with new and better parts (VRM, cache, BIOS, audio, onboard CF Flash etc)?

[@necro_ware]

It wouldn't be easy, if we really want to use new parts and not rely on old chipsets and all that stuff. Otherwise it is still easier to find a Socket 7 mainboard. They aren't very expensive (yet).

[@JohnSmith-iu8cj]

I would like to have a look at your gerber files. The routing looks like it could be improved

[@Ikkepop]

I wish you would do that series about theory, that would be very interesting

[@sinecosine5421]

Thank you for what you do. As a developer myself with a HW hobby, I truly appreciate your videos. Is there a reason you did not use lower profile capacitors?

[@necro_ware]

Thank you. I used, what I had at hand, however those caps have also quite high capacity and you don't get small caps in that range easily. It's possible to use smaller caps, but at costs of voltage stability.

[@SudosFTW]

You should just call it the S7 VROOM! at this point. also, high-quality aluminum polymer caps are what you likely need for the VRM in the end, they're a seriously decent choice over electrolytic when talking about power delivery applications. on most Dell Pentium II, III and early Pentium 4 boards, Sanyo OSCON poly capacitors were used in the CPU power delivery circuits due to their ability to handle switching voltages very quickly. it's a bit more of an expense, but it's worth it, even if the only capacitors you can get in your budget are cheaper brands like CapXon, which I recently learned have had a pretty decent line of poly caps for the past 5 years or so, much better in fact than their electrolytic counterparts.

[@SchkuenteQoostewin]

I have a PC Chips M520 board with a VRM port. I would like to have one of these built. I would build it myself but like you I am a software engineer but one with only one hand working. Soldering and such is a bit harder with this challenge, is there anyone to whom you can suggest to have this built?

[@Narwaro]

“Voltage ripple […] millivolts” The linear voltnut: “You mean microvolts, right? RIGHT?!?!” xD

[@mykolapliashechnykov8701]

Curious about the brown extension near a PCI slot on the Asus board. Looks too small for the PCI64 though. What could it be? Never saw anything like this before.

[@necro_ware]

The answer is in the first part ;)

[@mykolapliashechnykov8701]

@@necro_ware Oh nice, just Asus being Asus. Thank you!

[@Qyngali]

Suggestion, put the dip switches facing upwards so they can be adjusted easily without removing the card.

[@necro_ware]

Thank you for your suggestion, I'll put this to the list of possible changes in the future.

[@SeanBZA]

@@necro_ware You get DIP switches with pre aligned 90 degree mountings just for this use.

[@binxyde]

Nice! Do you still have the protective film on your oscilloscop? Looks so blurry 😯

[@necro_ware]

Yes, I removed it first, but the screen was so glossy, that the camera always started to refocus all the time, so I added a new film, which now makes it slightly blurry, but still better, than before.

[@Stefan_Payne]

Have you measured Ripple of that VRM Module? Also: why didn't you go with Polymer Caps?

[@Ironclad17]

A multiphase vrm on socket 7! Just how far can these old cpus be pushed?

[@schmatzler]

Nice! I got one of these Asus MediaBus boards a while ago. I'm currently missing some parts to fully assemble a PC and decided to just "wing it" and maybe run a Pentium MMX at 3.3V speed with a fan. But this seems to be a much nicer solution than that.

[@necro_ware]

If you run a Pentium MMX at 233MHz and 3.3V, it will beat the shit out of the on board linear voltage regulator. Cooling the CPU will be not you main problem ;)

[@schmatzler]

Alright, I'll wait for you to finish this project and release it instead of killing my board instantly. :D

[@clintthompson4100]

Hello Necroware. I am loving this newest revision and can't wait to see the next improvement as well. I have a similar board that is made by PC CHIPS and its has the same place and pins for a VRM and only supports to the original Pentium 200. It was a Cyrix PR 166+ and with a bios update it will support the 75Mhz FSB X 3 multiplier and I want to keep it a Cyrix system (MII CPU support with bios update) and would love to know how much power wattage that CPU draws and if this can also be supported. Have a good one and hope you see this.

[@necro_ware]

Hi, I don't know how much current MII pulls, but I'm quite sure not much more, than a K6-2 400. So long story short, yes, this VRM should be able to power the MII as well. I don't have it, so I can't test.

[@clintthompson4100]

​@@necro_ware Hello Necroware I saw that the Cyrix PR 300 M II draws up 22 watts max depending on if its the 66Mhz or 75Mhz FSB and the K6 2 400 is 16.9 watts. Hoping this well in specs of what your VRM can handel. Have s good one. Thank you again for this project.

[@matthewday7565]

Maybe you need to make space for a series inductor in the output filtering, for greater ripple reduction than capacitors alone - or maybe change one of the capacitors to 100uF or 220uF, close to self resonant at the switching frequency

[@mattparker9726]

YUS~!!!!

[@VetallRX]

Do not touch the working converter board with your fingers. Sometimes this can knock down its operating mode and create voltages at the output that will damage the processor. I encountered this in electronics.

[@necro_ware]

No risk, no fun ;)