Hi guys! I'm really sorry about the cliffhanger, but I invested quite a lot of time into this topic so far and I had a decision to make just to show what I have so far, or to wait another week and bring a video out, which would be 1h long or more. That would be definitely too long and cutting such long videos is quite exhausting as well. Unfortunately it doesn't matter where I'd cut, it would be always a cliffhanger, since I have plenty of interesting questions to answer. To be honest, some questions are still even unanswered for me, so I don't even know what exactly will land in the next part and what will still not be ready.
You're 586 I see.. ;) Yea, and that manuver with scsi + sound chips :D Lol.... I still keep EMUK10K base SB1024 Live! Beets current cards to the ground! Great project Sir!
Awesome explanation of switching voltage regulation. Modern motherboards use almost exactly this combination of components, if you pull the heat spreaders off your VRM section you can see the same types of components underneath. PWM controllers (high and low side), inductor coil/s and high end caps. Modern implementations are a fair bit more complex but the theory is the same.
Most interestingly, you even find synchronous step down converters on super socket 7 boards and later. I always thought these step down converters were introduced much later, since all other consumer electronics at the time still used transformers and linear regulators. The current consumption on 5V probably quickly became hard to handle though, which is why they've switched to 12V in the P4 era. Btw, ever tried to boot a shorted Athlon CPU? I've measured 50A+ current on 5V before I switched off the PSU after one or two seconds. The board always survived this torture so far ;-)
Did it take in 5v and give out what "main board" (bios) asked or more often jumpers😬 Very hard to remember. I got one "pc" with dual powers over 1000w total, 4 pentium pros on board (200 and those larger l2 cashes on each. 1 gb ram (8 sticks, half side can add 1gb more. It was "not cheap" when new. I know it works (when last tested some 10y ago), I have to start think what can it do? What if linux added and plugged to net, then sell cpu-power for some? *)
Man do I love it when TH-cam actually recommends an amazing video, along with an amazing channel I know nothing about! Great part 1 of this cool project, really enjoyed checking this out! I’ll be back for part 2, and to binge some of your old videos! Keep up the amazing work!
Video just ended with a cliffhanger. NO!! Very cool project and приятная музыка от русских людей . My lessions in russian back in schooldays are far long gone but i still can read it.
OK, I gotta be “that guy.” Bear with me. :-) When you’re saying that switching regs don’t get as hot because they spend time in the off state to cool down, that’s not actually.. well, not really at all how that works. If a transistor is turned all the way on (saturated) it tends to have very low resistance. In a FET, this might be the Rds On spec, and be a couple of ohms, or even fractions of an ohm. That is why transistors can switch large currents without getting terribly hot - not much energy is being lost through them. This is true whether the transistor’s on time is short or practically infinite, so on-time is not really a factor. If the transistor is not in saturation, and not off, it’s in the linear region, where its resistance is a function of the drive voltage (FET) or current (BJT), and this is where heat is generated. Even when used in a PWM circuit, which is cycling between off and saturated, it has to _transition_ through the linear region. This is what contributes to the majority of heat in switching circuits, and incidentally, in digital logic. In fact this is why higher frequency CPUs and GPUs get so hot - all those millions of transistors switching through the linear region for super short periods of time, several hundred million to billions of times a second. Lowering the voltage helps counter that (Ohms Law again), and that’s why Vcore keeps dropping as we’re able to manufacture more sensitive transistors. But I digress. So, a switching reg runs cooler than a linear reg because the switching regulator is minimizing the time spent in the linear region by turning full on or full off. This has the effect of overshooting and undershooting the voltage essentially to the rails, which is why switching regs require the inductor and capacitor to average it out. Without those passives, they would have to rely on parasitic inductance and capacitance to have any kind of slope at all, which is not particularly feasible except in niche applications (e.g., Class D audio amplifiers use the load - a speaker - as an inductive filter.) The linear reg, in contrast, lives _entirely_ in the linear region, turning that excess energy into heat by using a transistor as a variable resistor. That’s why they are hot. Linear regs are resistive, switching regs try not to be.
Hehe, funny to see how you were afraid :D You are absolutely right and this is a great explanation. As I was making this video I was sitting there and thinking how to explain all that in one sentence, but everything I thought of ended again in 10 more minutes of complicated explanation in an already a long video. So I took a wrong exit and just said, what is simply not right. Afterwards I think, that I just had to say, that the transistor gets hot in the moment of switch and cools down the rest of time, but as so often, the words we actually should have said come too late. Anyways, thank you for the explanation. I'm sure, if someone is interested, this is a nice contribution...
Yep and psu:s that time wanted give much more 5vamps. I think it it not gonna heat more if 12v used? Maybe even less, but I don't sure why I see it differently. But sure, that 12v can be used for just some exotic pvm-modules own needs. That is not fact, I just think it that way. And if I get glue why, I will tell.
I can't believe you just explained linear vs switching PSU's/Voltage Regulators in like 5 mins so simply and I understood after me researching so long to figure it out.
Well, I think, the question is not what we could do, but more of what we should do. I mean we can drive a car without the air in the tires, but that would damage the car in the long term. It's similar in this case. It is possible, but you shouldn't do it, since the CPU will not survive it too long, especially the one, which is as old.
@@necro_ware I think a better analogy would be driving your car with slightly over inflated tires :) Anyhow, these CPUs ran for years problem free in servers until they were decommissioned. But I agree, you shouldn't do it if you're not willing to take the risk.
Very good video sir, thank you for the spark I needed to get back to my electronic roots. I took classes at a local vocational center in high school where I learnt first the basics of electronics and then the 2nd year was more advanced electronics with pcb mapping and engineering our own circuits making small audio amplifiers. I enjoyed those classes because they challenged my brain and I got to see something that I made that most people have no clue how it do. I will be ordering me a new starter kit after I finish with this comment. You know, a good adjustable soldering apparatus, assortment of solder, rosin core and solid, a clamping station to hold current work piece and probably some capacitors for 3 motherboards and 2 video cards that I refuse to throw away because I know what's wrong with them. Have a nice day and thank you for reading my comment.
Really awesome project! I have an old 430FX board with a VRM slot that I'd love to try this in at some point. Really love your content. Thanks for sharing!
Awesome. This was hands down the simplest and best "casual" explanation of a switching VRM I have ever seen. Working on integrated CPU modules on mid 1990s, I can remember, when we did not have easy-to-handle integrated adjustable regulators, putting together a switching VRM with narrow voltage margins at high amps in a very small formfactor was quite a challange. Voltages skewing out of the hysteresis-window where commonplace and you where chasing software-"bugs" on the OS side for a crashing machine, all the while the VRM stability was the culprit really killed the days back then. (Schalthystererse in German, I am not so sure if I got the correct terms in english). Whatever, for an old fart like me this content is highly entertaining and enjoyable. Please keep up the good work - at least I enjoy it. May I ask, if you are a professional EE or "just" a dedicated hobbyist? Ah, forget that question, just heard your statement in the follow-up video :)
This was a very cool video and an awesome project. Your explanations were very easy to digest - you'd make an awesome teacher! Can't wait for part two!
Cliffhanger! Can't wait for part 2. Awesome work & design. You make SMD look so easy. I have a couple SMD kits to practice with before I get into serious SMD work and repairs. Cheers!
Thank you for your explanation on voltage regulation. I am an aero student who has recently been exposed to the arcane electrical arts, and your video helps me understand the fundamentals of regulators.
Wow, you really shouldn't see any of my content as educational. This is a purely entertaining channel for tinkerers and hobbyists. I am a software engineer and not an electronics engineer and I always tend to oversimplify things to make them easier to understand. Like in this case, I only end up explaining the basic ideas and some things which I said could be even not quite right. So, if you need educated explanation of the principles, please consider to watch more professional channels and always take what I'm saying with a grain of salt.
Electrical engineering is incredibly fascinating. You're very thorough and meticulous in your designing and testing before actually using your creations. The true sign of an expert! Can't wait to see part 2 of the video.
Thank you, but I'm actually not an expert in electronics at all. Despite, that I had quite a lot of electronics in the university many years ago, I'm actually a software engineer and I was teaching that quite a lot as well. On the other hand I am an absolute noob in electronics :D But to my defense, it's part of my character to stay thorough in everything I do, doesn't matter if it's electronics or cooking....
@@necro_ware now I'm envious! You're this good with electronics AND you're a software engineer? I'm struggling with becoming a software engineer and I can't even begin handling electronics the way you do. You're very talented.
Only way to learn is to dive in. I dabble in both software and hardware as well. (My profession is otherwise technical, and kind of a HW/SW blend.) You will soon learn, in both fields, that regardless how much you know, there is still far more to learn. But you have to start somewhere!
Hey so just felt like sharing this: I just got an old socket 7 pc that came with a P133 at 3.3V, and I placed a PMMX233 in there. It does work, and as far as I tested, it's pretty stable. It does run very hot and that makes me ~uneasy~, so I'll keep the 133 until I get to assemble one of those modules! Thank you very much for the project and the videos, tho! You inspired me to get into this era of computing and into modding my computer, and I'm having so much fun =D
This is crazy, kudos to you for reviving obscure/rare motherboard accessories like this and took it to the next level. Informative, easy to understand and fun to watch. You probably already know this but it would be a great idea to mention some models of Pentium OverDrive that feature on-package VRM for 3.3V motherboard support. With the amount of work you put into vintage computing, your channel deserves more subs!
I really look forward to these videos! It brings me back to a pc I had a lot of fun with, (hrs of Doom etc), my 486-DX50. I like seeing what they can do and could have done with some tweaking.
Man this really took me back. The first PC I kinda built with my own money was a Socket 7 using used parts and a Compaq I bought from a neighbor. I got a couple of sticks of RAM from a machine that wouldn't boot from a junk pile of PCs in the tech lab in high school, and two hard drives from a massive Compaq server tower that was being sold by a real estate company in the classified ads (data security was different in the 90's.) I remember swapping the CPU at some point, but I don't remember the model or frequency on it.
I like your soldering technique, and how you trim the through-hole leads before soldering to minimize any mechanical stress caused by doing it the other way. Also neat to see how you handle commenters. Subscribed!
I think it's great you made such an enhancement to an old MB: you brought it to its full potential after decades! With regard to the PSU module though, I reckon your design is an example of bad design practice which could lead to some serous issues. The main problem is that you are trying to get two independent feedback loops to regulate the same rail... even if they have, nominally, the same target there will be a difference in their sensing and such difference will also drift depending on many factors, principally the temperature of the ICs. This is likely to lead to an unbalance in the workload with one of the two taking on most of the burden. Another type of issue is likely to arise with the dynamic behaviour of the PSU: the regulators are trying to regulate each other and they will react to each other and not just to the change in demand by the CPU. In certain load conditions the system may become unstable. Finally the XL2596 runs at a pretty low frequency for the type of load you are asking it to regulate. Do the Intel specs say anything about the requirements on the frequency response of the regulator? The amount of power you are dealing with is substantial and I would look for synchronous (controlled MOSFET instead of diode) buck regulators (or controllers) in order to reach an higher efficiency. Typically you would also design an output stage with 2-phases but that would be a bit tricky and yes, even more expensive. Have you thought of reaching out to GreatScott! (th-cam.com/users/greatscottlab) who is also in Germany? This kind of project seems just up his street.
Hi Lorenzo! Thank you for the feedback. Yes, the design is not very good and since I'm actually a software developer, I do make such mistakes. But it is a good thing, because I'm learning from my mistakes and trying to improve. As you see in the video, this was the version 0.1 (the very first one). Meanwhile I'm at version 0.3, where I switched to a DC-DC Controller with an N-MOSFET instead. That should be a better solution in the near future, but this version 0.1 was a good start and as a prototype it did the trick. I learned a lot things, which I wouldn't, if I'd just take a schematics from somewhere. My main reason to do all that is to learn and not just to make another gadget :)
@@necro_ware, I hope I wasn't too harsh on you. I'm an hardware engineer and I've been designing switching power supplies for more then 20 years... I guess my "design review mode" kicked in! To be fair, if this is one of your first attempt to SMPS, you've taken on an objectively tricky task! Congratulations! If you've gone for a synchronous controller then things got serious! ;-) To keep things manageable I would have gone for a TI TPS56628 which is synchronous so the heat generated should be kept to a minimum. It's rated for maximum 6A, a bit tight, but not insufficient. The package is SO8, not too complicated to solder and with PowerPad which would allow you to use the ground plane around the IC as an heatsink. The switching frequency is 650kHz so the inductance can be relatively small and still keep the ripple low. With a good choice of capacitors on the output it should do the job. Have you given any thought to contacting Great Scott? Power electronics is his thing.
Someone else mentioned using ceramic caps in proximity to the IC, which is often a requirement for guaranteed stability. Another suggestion: Try to ensure your switchable voltage divider (the feedback Rs) don’t rely on that switch to run safely. By that I mean, make sure even if all the switches are open (or fail to contact) that your feedback doesn’t go open loop. Switch in parallel Rs, ideally on the top half, to make sure that, if all your switches fail open, then your output V is at its minimum. That’s much safer than having the worst case failure result in over-volting. Edit: Er, I probably meant bottom half of the divider. I dunno. It’s late and I always mix those things up when I’m talking off the cuff. You get what I mean. ;-)
No, your critics were absolutely top. Most of that I already discovered myself as well, that's why I designed the MOSFET solution with the controller instead. But it's currently only on paper, I still have to test it and so, every constructive input is a good input. I don't like comments like "...I'm shocked, that you don't know...", but every comment which helps me to improve is highly appreciated. As I said, I'm a software developer and where I'm quite ok in digital electronics, the analogue circuits and power supply is really a minefield for me yet. I try my best to gain my knowledge at that and I think, that I'm on the right way, but still there is a long way to go. Meanwhile I think, that I need the regulator for up to 15A, so it starts to be really tricky now. You will see the reason in the next part. But if you have more tips, you are welcome. As far as I can measure the regulator frequency doesn't seem to be a problem for the Pentium MMX. 150kHz seem to be alright. I think this would be a problem for CPUs for a much higher frequencies. I know the channel of Great Scott, yes, but I didn't contact him on that regards. I have some friends, who are hardware engineers and I could ask them as well, but it makes too much fund to understand such things on my own. It's always the same with me, I'm more into reading and trying to figure out it on my own, than to ask someone. The most valuable help one can give me is to give a source, where I can read about it ;)
@@necro_ware, I wish you'll make the new regulator to work. 15A is does call for a controller with discrete MOSFETs. Without getting into all the theory of feedback loop control, you should find plenty of very useful information and guidelines in the datasheets and applications notes of the most reputable manufacturers (TI, Analog Devices, Microchip, OnSemi, ST, ...). Good datasheets illustrate how to select the components to use with the controller. The buck topology is usually the most benign among switch-mode power supplies so, if you don't need to do anything particularly sophisticated or uncommon, it should be possible to put together a circuit which works well. I expect the inductor to be the most expensive component of the regulator. Unfortunately if the switching frequency is low, the value of inductance that leads to a certain ripple is high. This also means that the average output current below which the regulator has to operates in discontinuous mode is high (discontinuous mode is more "noisy"). Finding the right inductor may be quite tricky. With regard to the "bulk" output capacitor, you'll have to select a low-ESR electrolytic capacitor and then use one or more of them in parallel. The datasheets of the series of "Low-ESR" specify such parameter at 100 kHz (often it's "Impedance at 100 kHz). Most reputable Japanese capacitor manufacturers have one or more series of low-ESR caps in their lineup. In a series, capacitor with the same physical size have the same ESR and the ESR is smaller with larger caps. if the output ripple is high you'll have to adopt an output capacitor with higher current ripple rating so you'll have to keep an eye on this parameter too. The choice of MOSFETs is obviously critical. The only thing that I would write here is that you should aim for the right balance between RDSon and gate capacitance (and gate charge): generally a lower RDSon is associated with an higher gate capacitance which, in turn, requires the controller to work harder to switch the MOSFET on and off. If you stay at relatively low frequencies, around 100kHz, then the power lost should be low as well but it may still be worth comparing the dynamic characteristics of the MOSFETs you find have a low enough RDSon.
Nice and cool design for retro computing fans! I had same motherboard back in days.. It's now disassembled for parts or in the box in garage.. Can't remember.. How I recall, it wasn't working and I didn't bother to repair all old motherboards when I get box of them in different condition and repaired only best species.. Now I look to things different - I even repairing battery corroded ones because I love old architecture. Old boards, usually, is more reliable than modern ones.. 😂
:O New microphone! Awesome upgrade! I was thinking lately if an upgrade would harm the charm of your channel and the overall retro feeling but oh boy, no worries. This is great and I like it a lot! :)
Yes I was playing with a new microphone indeed. My room adds unfortunately a lot of reverb and I'll have to do something about that in the future.... Thank you very much.
I wish, I've had that module back in the days! It was during my time at university and was on on a budget as all typical students. This would have allowed me, to upgrade my P120 to the more powerful MMX-Pentiums.
My guess is that with this VRM you could use even an AMD K6-2/333AFR-66. At least the 2.2V vcore would have not been a problem. Clock multiplier set on 2 on the mainboard would tell a K6-2 CPU to set its own internal multiplier to 6, which would have been overclocking for a K6-2/333 on a 66 MHz bus. But then you have the opportunity to use 60 MHz FSB (less overclock) or 50 MHz FSB (underclocking).
Very slick work. At one time I had an ASUS TX chipset board maxed out with a Pentium 233 MMX CPU with 256 megs of RAM, a Creative AWE32 soundcard with 8meg of RAM and an Nvidia 5200 GPU. Good times.
Repair videos are nice, modding videos are always the best :) Ok maybe not so happy about the design options here* However I enjoyed this either way. Nice topic, video :) * It is impossible to cover everything and get into details here, however one thing you can easily fix (and is very important) is placing the diodes on the same side and as close to the chips as possible! The layout of the PCB on the output of these regulators is critical. Ideally you want everything on the top layer there, components and tracks.
Thank you very much! Yes I know all the restrictions, distances, trace widths etc. I was studying the documentation and datasheets quite long. And I was sitting at the design also very long, way too long, if you ask me. I probably wrote a book full of calculations to make it right :D For example the reason why the diodes are on the back side was actually the shortest distance (using direct vias between the pads) to the inductor, which was one of the serious requirements. I also designed the feedback to be out of the inductor flux. Also a hard restriction was to keep the resistors as near as possible to the converter, but it's not possible if you use two of them. So I tried to keep equal but short distances to the resistors from both ICs. I also made a lot of thoughts about heat dissipation transferring the energy using an array of vias onto the back side and many many more tiny but important things. It was quite hard, but I think that I made everything with as little compromises as possible. And what I see on the oscilloscope now looks very good....
I got a P55T2P4, the evolution of the P55TP4XEG with the intel VX. I was never able to make set the motherboard with i233MMX to run at 233. It runs at 200, very stable even with Windows Millenium. I solded the USB 1.0 port on it, but it did a mistake and 1 burnt... But a Nec 5 USB 2.0 port card, Voodoo 1, Nec PowerVR, S3 Virge DX 375, Sound Blaster 64 Awe, 512 ram cache stick, 96mb EDO Ram, IrDA module, 3Com Pci Ethernet, 2 HDD, 2 CDRW drives, 2x 3.5" floppy drives... with just a "noname" 220v/230W Psu
A very neat expansion for your board that I am pretty sure gonna work. my guess is that the VRM board will become pretty hot when loaded fully if that cpu really gonna need all amps as it stated. very good to step down the 5V rail instead as a major failure might give a smaller risk of injecting lethal voltage on the cpu than using the 12volt one. good job
@@necro_ware I can´t wait, I also have tons of vintage computers like this and I enjoy watching your videos about them, and maybe some can be used with my computers as well in the end :)
I love your hardware hacking add-on modules. You did a great job at explaining PWM for a beginner. I forget if the last time you spoke about switch mode power regulators you explained what purpose the Darlington transistor played. I think people watch your channel to learn a bit as well, so it might be something you actually want to dive into.
I didn't explain a darlington transistor yet, may be I mentioned it in the comments, but in a video I didn't have an occasion yet to explain it so far. May be one day.
There are many types of transistors. One class is called bipolar, and will be designated as NPN or PNP. Those require current to flow either from the base to the emitter (NPN) or emitter to base (PNP) to turn the transistor on. The transistor will then allow a proportional amount of current to flow between collector and emitter. The more current that flows through the B/E junction, the more can flow through the C/E junction. In that way a bipolar transistor is kind of like a current amplifier. The problem is, you need to provide that proportional current through the base to get the C/E to open. If you’re trying to drive that current from a low-power source, like the output from a digital logic chip, you may have trouble opening the spigot wide enough to avoid limiting the current you’re trying to control with the transistor. In that case you use a pair of transistors in a Darlington configuration: One as a first stage amplifier, which then controls the base of the second with an already-amplified current. The pair will take a very small current and boost it, twice, to a very large current. In that way, it acts like a cheater bar on a wrench, providing additional leverage. A single transistor might amplify the current from B/E to C/E by something like 100 times. In a Darlington configuration, the current in B/E of transistor 1 is then multiplied by that 100 times, and that is fed to transistor 2 which amplified it another 100 times for a total current gain of 10,000. Another type of transistor, called a FET, doesn’t have quite this same problem because the “base” (which, on a FET is called a gate), is controlled by voltage, not by current. So provided you can supply sufficient voltage to the gate, there’s very little (negligible) current required to control it. However it is still possible to find cascaded configurations with a driver transistor and a load-switching transistor due to things like .. needing higher voltages to turn on the FET (which may require 10V and need to be controlled by 3.3V or 5V logic), or how all gates have a certain amount of capacitance that, when switched at high frequencies, can put strain on whatever is driving it.
Nice! I have a Micronics board which had the VRM socket, but did not officially support MMX and a newer Micronics board with the VRM module in the socket. Fortunately swapping the VRM between these 2 was enough to enable the use of Pentium MMX CPU. But I have an Intel Advanced/EV board which I like a lot that has just the solder pads for the socket. After this video I will have to research if it also can be upgraded.
Highly Likely. Some later boards came with the VRM socket but only had a little dummy module installed that essentially just did the 4 jumper wire thing like was done on this board in hard-wired form for non-MMX Pentiums. Even some OEM PCs came like that (HP Vectra come to mind). Gave you the option to upgrade to a MMX CPU (good for marketing) but you had to pay for a VRM module (good for business) :P
Wow, that is some real hacking there - I really enjoy these vids, I understand the terms you use, but wow, your smarts are off the chart! I have a bunch of old motherboards - nailed to the wall as trophies (my first 386 through to dual PIII boards...) I recently got given back one of my older builds - a Rhino III and a Quantum bigfoot, that was found in the back of a garage.
Thank you for pleasant memories from childhood. I think nobody made 430FX boards for MMX Pentiums, because SDRAM support of 430TX dramatically increased overall performance.
That's true, but not only because of the memory. The later 430HX and VX could also cache a lot more memory and were equipped with faster cache. In the end TX was kind of flagman for Socket 7 from Intel. After that they moved to P2.
Voltages where so analog back then. Jumps in between of 0.100v, i.e 2.8V, 2.9V, 3.0V etc while these days due to the digital VRM implementation it can be so accurate till the last milivolt if required. Loved watching this vid. Did'nt know these Socket 7 boards could be modded all the way.
Not wanna be a smartass or give non-needed advice but you inspire many people to fiddle with retro hardware. Would be lovely to see that you ground yourself or advise your audience to do it so. Static electricity can really destroy the good mood :)
This gives me realy good memories. My first 'own' PC had an 133MHz CPU and I overclocked it just by jumperconfig. It runs not realy stabile on 200MHz, but without failures on 166MHz. And this with stock cooler an nothing more in the case, as the fan of the little PSU.
i reconize this MB my dad gave me the same one back in the day. and then i read the manuel and i found that i could also do this same thing :-) but we never realy got around to doing that because back then ics wer verry expensive and we had plenty pentium mbs that can take mmx cpus but my dad did put an resistor at the back side and mmx did work but not fully clockspeed. he didnt have a heat gun eather and the cost was just simply to much so we didnt do it. so nice to see you making it work after all these years :-) you are the best im verry ha[ppy to have found you . i tyhought i reconized it from somewhere lol such a long time ago this brings back nice memmories thanks yo so much
26:48 lol Yep, Back in the day I had a P166MMX overclocked to ~210 Mhz (undocumented 83x2.5 setting, if I'm not mistaken). I had it for about 5 years (since 1997 until around 2002 or 2003), since I was a poor student back then and had no money for an upgrade. All those years I didn't even bother to open the case, since everything just worked fine, sometimes 24/7, and since I didn't tinker with this config due to the lack of money. So when finally I made an upgrade, I opened the old case to salvage a sound card and perhaps some other parts from an old system and what I found out was that the CPU cooler wasn't even connected. lol I guess, the guy who assembled my system in the shop forgot to connect it and it worked all those years without any problems with just a small aluminium heatsink (btw also without any thermal paste between it and the CPU)! I was actually very impressed, since it was also overclocked and I played quite a lot those days, so God know what sort of temperatures that CPU sustained and for how long.
Very nice explanation that was easy to understand :) Time to make another retro computer news video if you don't mind me using a few seconds of yor video :)
I don't mind :) I like your news videos, but may be it's worth it to wait for the second part first. I'll try to answer some important questions there.
Hehe :D Well, I don't think, that you can find anyone, who would agree to love cliffhangers. But this one was for a reason, there is a chain of further issues, which would be each a cliffhanger for itself. Would I put them all into one video, it would probably last the next 1,5h. So I have to sort out first what is not so interesting and think about how I can put everything together. So I was standing in front of a decision to make a very long video (probably over 1h) in the next two weeks, or to make a cut now, reorder the materials and make two shorter videos, where I'd be able to release the first one now. The way was obvious for me, even if I have to tease you a little bit :D
@@necro_ware Out of curiosity.. do you think there's any way to adapt a CPU that has a lower IO voltage than 3.3V? I've got a pretty rare embedded Pentium MMX 266 kicking around that I'd love to use but I'm afraid of damaging unnecessarily. Reason is quite simple... it runs at just 1.9V VCore (the lowest any of my boards go is 2.0V, which I guess would be survivable, maybe even help it get to 300MHz), but in addition to that it runs at 2.5V IO which has turned into quite a headache... unsure what to do other than try and find a board designed for embedded Pentiums... ie needle in the haystack. Unless someone comes up with some other idea
Well done for the double buck regulator. However, your explanation is missing one key point, the schottky diodes. In a buck regulator, the diode provides current drawn from 0v and pumped by the coil each time the transistor is cut off. That's the only way to output MORE current than how much is actually injected at the input. That's what allows a better efficiency than a linear regulator. In your case, you need 5 amps, and to lower from 5v to 2.8v. 5*(5-2.8) = 11 watts and 5 amps + internal regulator needs drawn from the input. Considering a efficiency of 80% for your circuit, this buck VRM will ouput 2.8*5 = 14 watts of power, and we can assume it will use 17,5 watts with efficiency of 80%, so only requiring 3,5 amps from the input, and therefore pumping 1,5 amp through the schottky diode. The total dissipation will only be 3,5 watts...
i can scarcely imagine a time before sse or even mmx. a lot of what i use my cpmputers for use those types of instructions heavily and man does it make a massive difference
Hello Necroware, Keep in mind without load the big buffer caps could keep the higher voltage after you step down the regulated DC voltage, always measure in the opposite way, increased the voltage never decreased before measure! Regards, Mark
I have the same problem on my hand, a Socket 7 MB with Pentium MMX, and with the VRM socket, but without the VRM module itself. I've bought a bunch of cheap DC-DC switching converter modules from Aliexpress, and now I'm trying to imagine the wire gauge need for 6 A, and how I'm going to physically wire the module to the VRM pins. There's also a problem with the trim potentiometer being too sensitive and drifting off the desired voltage - need to replace it with a 20-turn pot.
I am only going back to Core 2 Quad Q6600, I had to live through the Pentium years, my first being a Pentium 60, and I don't miss those old processors at all.
It depends on what's your aim is. If s.o. is just a user, then a modern system is fine. I don't even see the point to use a Q6600 in such a case tbh. But if you want to learn about electronics, about how computers work etc, then the older the better. In modern systems everything is highly integrated and there's not a lot of space for experiments.
I'm surprised you didn't test your VRM under dummy load before subjecting the CPU to it. Actually I'm surprised you didn't put any load, as switching supplies and even some linears can be unstable without load. Also I would check how clean the output is with an oscilloscope on no/low and high load.
Hi guys! I'm really sorry about the cliffhanger, but I invested quite a lot of time into this topic so far and I had a decision to make just to show what I have so far, or to wait another week and bring a video out, which would be 1h long or more. That would be definitely too long and cutting such long videos is quite exhausting as well. Unfortunately it doesn't matter where I'd cut, it would be always a cliffhanger, since I have plenty of interesting questions to answer. To be honest, some questions are still even unanswered for me, so I don't even know what exactly will land in the next part and what will still not be ready.
Perfect! I love to binge watch as much as the next guy, but I'm enjoying the anticipation in this case
If the Marvel and Star Wars universes can end episodes or movies with a cliffhanger, I suppose you can too. :)
My disappoint is immeasurable and my day is ruined.
Thanks for any Time you dedicate towards projecting light. Your output is regulated by the hearts capacity for love.
Well Received
God Bless.
You're 586 I see.. ;) Yea, and that manuver with scsi + sound chips :D Lol.... I still keep EMUK10K base SB1024 Live! Beets current cards to the ground!
Great project Sir!
Awesome explanation of switching voltage regulation. Modern motherboards use almost exactly this combination of components, if you pull the heat spreaders off your VRM section you can see the same types of components underneath. PWM controllers (high and low side), inductor coil/s and high end caps. Modern implementations are a fair bit more complex but the theory is the same.
Most interestingly, you even find synchronous step down converters on super socket 7 boards and later. I always thought these step down converters were introduced much later, since all other consumer electronics at the time still used transformers and linear regulators. The current consumption on 5V probably quickly became hard to handle though, which is why they've switched to 12V in the P4 era. Btw, ever tried to boot a shorted Athlon CPU? I've measured 50A+ current on 5V before I switched off the PSU after one or two seconds. The board always survived this torture so far ;-)
PWM does not work like that at all
Did it take in 5v and give out what "main board" (bios) asked or more often jumpers😬
Very hard to remember. I got one "pc" with dual powers over 1000w total, 4 pentium pros on board (200 and those larger l2 cashes on each. 1 gb ram (8 sticks, half side can add 1gb more. It was "not cheap" when new. I know it works (when last tested some 10y ago), I have to start think what can it do? What if linux added and plugged to net, then sell cpu-power for some? *)
That machine have 4 those vrm-mods for one for each cpu.
I saved at the time few spare ones too and many other stuff too x)
Man do I love it when TH-cam actually recommends an amazing video, along with an amazing channel I know nothing about!
Great part 1 of this cool project, really enjoyed checking this out! I’ll be back for part 2, and to binge some of your old videos!
Keep up the amazing work!
Классная, необычная ретро-работа. С пониманием и объяснением. Спасибо Вам за отличный контент.
Video just ended with a cliffhanger. NO!! Very cool project and приятная музыка от русских людей . My lessions in russian back in schooldays are far long gone but i still can read it.
OK, I gotta be “that guy.” Bear with me. :-)
When you’re saying that switching regs don’t get as hot because they spend time in the off state to cool down, that’s not actually.. well, not really at all how that works.
If a transistor is turned all the way on (saturated) it tends to have very low resistance. In a FET, this might be the Rds On spec, and be a couple of ohms, or even fractions of an ohm. That is why transistors can switch large currents without getting terribly hot - not much energy is being lost through them. This is true whether the transistor’s on time is short or practically infinite, so on-time is not really a factor.
If the transistor is not in saturation, and not off, it’s in the linear region, where its resistance is a function of the drive voltage (FET) or current (BJT), and this is where heat is generated. Even when used in a PWM circuit, which is cycling between off and saturated, it has to _transition_ through the linear region. This is what contributes to the majority of heat in switching circuits, and incidentally, in digital logic. In fact this is why higher frequency CPUs and GPUs get so hot - all those millions of transistors switching through the linear region for super short periods of time, several hundred million to billions of times a second. Lowering the voltage helps counter that (Ohms Law again), and that’s why Vcore keeps dropping as we’re able to manufacture more sensitive transistors. But I digress.
So, a switching reg runs cooler than a linear reg because the switching regulator is minimizing the time spent in the linear region by turning full on or full off. This has the effect of overshooting and undershooting the voltage essentially to the rails, which is why switching regs require the inductor and capacitor to average it out. Without those passives, they would have to rely on parasitic inductance and capacitance to have any kind of slope at all, which is not particularly feasible except in niche applications (e.g., Class D audio amplifiers use the load - a speaker - as an inductive filter.)
The linear reg, in contrast, lives _entirely_ in the linear region, turning that excess energy into heat by using a transistor as a variable resistor. That’s why they are hot. Linear regs are resistive, switching regs try not to be.
Hehe, funny to see how you were afraid :D
You are absolutely right and this is a great explanation. As I was making this video I was sitting there and thinking how to explain all that in one sentence, but everything I thought of ended again in 10 more minutes of complicated explanation in an already a long video. So I took a wrong exit and just said, what is simply not right. Afterwards I think, that I just had to say, that the transistor gets hot in the moment of switch and cools down the rest of time, but as so often, the words we actually should have said come too late.
Anyways, thank you for the explanation. I'm sure, if someone is interested, this is a nice contribution...
I think the heat is about stepvoltage (0.2-0.6v) and multiplied by amps what goes true douped silica chip (tyristor, transistor and fet's usually?)
That cpu isn't "hungry" with power what I 🤔 "was hungry" hmm
I am 😐
Yep and psu:s that time wanted give much more 5vamps.
I think it it not gonna heat more if 12v used? Maybe even less, but I don't sure why I see it differently. But sure, that 12v can be used for just some exotic pvm-modules own needs. That is not fact, I just think it that way. And if I get glue why, I will tell.
I can't believe you just explained linear vs switching PSU's/Voltage Regulators in like 5 mins so simply and I understood after me researching so long to figure it out.
Fascinating. PWM controllers are also useful in variable speed drives for electric motors. On to Part Two!
Found myself rewatching your videos, still one of the most interesting retro videos out there :) Love it!
I used to run those 2.8v MMX CPU's on the 3.3v boards at 3.3v with no problems back in the day. Actually made it more overclockable :)
Well, I think, the question is not what we could do, but more of what we should do. I mean we can drive a car without the air in the tires, but that would damage the car in the long term. It's similar in this case. It is possible, but you shouldn't do it, since the CPU will not survive it too long, especially the one, which is as old.
@@necro_ware I think a better analogy would be driving your car with slightly over inflated tires :) Anyhow, these CPUs ran for years problem free in servers until they were decommissioned. But I agree, you shouldn't do it if you're not willing to take the risk.
Overclocked my 166mmx at 262 3.2v, woked until it was retired but it needed decent cooling
I've loved the various explanations about voltage regulation.
Very good video sir, thank you for the spark I needed to get back to my electronic roots. I took classes at a local vocational center in high school where I learnt first the basics of electronics and then the 2nd year was more advanced electronics with pcb mapping and engineering our own circuits making small audio amplifiers. I enjoyed those classes because they challenged my brain and I got to see something that I made that most people have no clue how it do. I will be ordering me a new starter kit after I finish with this comment. You know, a good adjustable soldering apparatus, assortment of solder, rosin core and solid, a clamping station to hold current work piece and probably some capacitors for 3 motherboards and 2 video cards that I refuse to throw away because I know what's wrong with them. Have a nice day and thank you for reading my comment.
Really awesome project! I have an old 430FX board with a VRM slot that I'd love to try this in at some point. Really love your content. Thanks for sharing!
Awesome. This was hands down the simplest and best "casual" explanation of a switching VRM I have ever seen. Working on integrated CPU modules on mid 1990s, I can remember, when we did not have easy-to-handle integrated adjustable regulators, putting together a switching VRM with narrow voltage margins at high amps in a very small formfactor was quite a challange. Voltages skewing out of the hysteresis-window where commonplace and you where chasing software-"bugs" on the OS side for a crashing machine, all the while the VRM stability was the culprit really killed the days back then. (Schalthystererse in German, I am not so sure if I got the correct terms in english).
Whatever, for an old fart like me this content is highly entertaining and enjoyable. Please keep up the good work - at least I enjoy it. May I ask, if you are a professional EE or "just" a dedicated hobbyist? Ah, forget that question, just heard your statement in the follow-up video :)
This was a very cool video and an awesome project. Your explanations were very easy to digest - you'd make an awesome teacher! Can't wait for part two!
Cliffhanger! Can't wait for part 2. Awesome work & design. You make SMD look so easy. I have a couple SMD kits to practice with before I get into serious SMD work and repairs. Cheers!
Thank you for your explanation on voltage regulation. I am an aero student who has recently been exposed to the arcane electrical arts, and your video helps me understand the fundamentals of regulators.
Wow, you really shouldn't see any of my content as educational. This is a purely entertaining channel for tinkerers and hobbyists. I am a software engineer and not an electronics engineer and I always tend to oversimplify things to make them easier to understand. Like in this case, I only end up explaining the basic ideas and some things which I said could be even not quite right. So, if you need educated explanation of the principles, please consider to watch more professional channels and always take what I'm saying with a grain of salt.
Electrical engineering is incredibly fascinating. You're very thorough and meticulous in your designing and testing before actually using your creations. The true sign of an expert! Can't wait to see part 2 of the video.
Thank you, but I'm actually not an expert in electronics at all. Despite, that I had quite a lot of electronics in the university many years ago, I'm actually a software engineer and I was teaching that quite a lot as well. On the other hand I am an absolute noob in electronics :D But to my defense, it's part of my character to stay thorough in everything I do, doesn't matter if it's electronics or cooking....
@@necro_ware now I'm envious! You're this good with electronics AND you're a software engineer? I'm struggling with becoming a software engineer and I can't even begin handling electronics the way you do. You're very talented.
Only way to learn is to dive in. I dabble in both software and hardware as well. (My profession is otherwise technical, and kind of a HW/SW blend.) You will soon learn, in both fields, that regardless how much you know, there is still far more to learn. But you have to start somewhere!
@@nickwallette6201 Absolutely
Hey so just felt like sharing this: I just got an old socket 7 pc that came with a P133 at 3.3V, and I placed a PMMX233 in there. It does work, and as far as I tested, it's pretty stable. It does run very hot and that makes me ~uneasy~, so I'll keep the 133 until I get to assemble one of those modules!
Thank you very much for the project and the videos, tho! You inspired me to get into this era of computing and into modding my computer, and I'm having so much fun =D
This is crazy, kudos to you for reviving obscure/rare motherboard accessories like this and took it to the next level. Informative, easy to understand and fun to watch. You probably already know this but it would be a great idea to mention some models of Pentium OverDrive that feature on-package VRM for 3.3V motherboard support.
With the amount of work you put into vintage computing, your channel deserves more subs!
Услышал родную музыку. Улыбнулся, подписался. Так держать!
I really look forward to these videos! It brings me back to a pc I had a lot of fun with, (hrs of Doom etc), my 486-DX50. I like seeing what they can do and could have done with some tweaking.
Great video! Liking your choice of music as well.
Man this really took me back. The first PC I kinda built with my own money was a Socket 7 using used parts and a Compaq I bought from a neighbor. I got a couple of sticks of RAM from a machine that wouldn't boot from a junk pile of PCs in the tech lab in high school, and two hard drives from a massive Compaq server tower that was being sold by a real estate company in the classified ads (data security was different in the 90's.) I remember swapping the CPU at some point, but I don't remember the model or frequency on it.
I like your soldering technique, and how you trim the through-hole leads before soldering to minimize any mechanical stress caused by doing it the other way. Also neat to see how you handle commenters. Subscribed!
I think it's great you made such an enhancement to an old MB: you brought it to its full potential after decades!
With regard to the PSU module though, I reckon your design is an example of bad design practice which could lead to some serous issues.
The main problem is that you are trying to get two independent feedback loops to regulate the same rail... even if they have, nominally, the same target there will be a difference in their sensing and such difference will also drift depending on many factors, principally the temperature of the ICs. This is likely to lead to an unbalance in the workload with one of the two taking on most of the burden.
Another type of issue is likely to arise with the dynamic behaviour of the PSU: the regulators are trying to regulate each other and they will react to each other and not just to the change in demand by the CPU. In certain load conditions the system may become unstable.
Finally the XL2596 runs at a pretty low frequency for the type of load you are asking it to regulate. Do the Intel specs say anything about the requirements on the frequency response of the regulator?
The amount of power you are dealing with is substantial and I would look for synchronous (controlled MOSFET instead of diode) buck regulators (or controllers) in order to reach an higher efficiency. Typically you would also design an output stage with 2-phases but that would be a bit tricky and yes, even more expensive.
Have you thought of reaching out to GreatScott! (th-cam.com/users/greatscottlab) who is also in Germany? This kind of project seems just up his street.
Hi Lorenzo! Thank you for the feedback. Yes, the design is not very good and since I'm actually a software developer, I do make such mistakes. But it is a good thing, because I'm learning from my mistakes and trying to improve. As you see in the video, this was the version 0.1 (the very first one). Meanwhile I'm at version 0.3, where I switched to a DC-DC Controller with an N-MOSFET instead. That should be a better solution in the near future, but this version 0.1 was a good start and as a prototype it did the trick. I learned a lot things, which I wouldn't, if I'd just take a schematics from somewhere. My main reason to do all that is to learn and not just to make another gadget :)
@@necro_ware,
I hope I wasn't too harsh on you. I'm an hardware engineer and I've been designing switching power supplies for more then 20 years... I guess my "design review mode" kicked in!
To be fair, if this is one of your first attempt to SMPS, you've taken on an objectively tricky task! Congratulations! If you've gone for a synchronous controller then things got serious! ;-)
To keep things manageable I would have gone for a TI TPS56628 which is synchronous so the heat generated should be kept to a minimum. It's rated for maximum 6A, a bit tight, but not insufficient. The package is SO8, not too complicated to solder and with PowerPad which would allow you to use the ground plane around the IC as an heatsink. The switching frequency is 650kHz so the inductance can be relatively small and still keep the ripple low. With a good choice of capacitors on the output it should do the job.
Have you given any thought to contacting Great Scott? Power electronics is his thing.
Someone else mentioned using ceramic caps in proximity to the IC, which is often a requirement for guaranteed stability.
Another suggestion: Try to ensure your switchable voltage divider (the feedback Rs) don’t rely on that switch to run safely. By that I mean, make sure even if all the switches are open (or fail to contact) that your feedback doesn’t go open loop. Switch in parallel Rs, ideally on the top half, to make sure that, if all your switches fail open, then your output V is at its minimum. That’s much safer than having the worst case failure result in over-volting.
Edit: Er, I probably meant bottom half of the divider. I dunno. It’s late and I always mix those things up when I’m talking off the cuff. You get what I mean. ;-)
No, your critics were absolutely top. Most of that I already discovered myself as well, that's why I designed the MOSFET solution with the controller instead. But it's currently only on paper, I still have to test it and so, every constructive input is a good input. I don't like comments like "...I'm shocked, that you don't know...", but every comment which helps me to improve is highly appreciated.
As I said, I'm a software developer and where I'm quite ok in digital electronics, the analogue circuits and power supply is really a minefield for me yet. I try my best to gain my knowledge at that and I think, that I'm on the right way, but still there is a long way to go. Meanwhile I think, that I need the regulator for up to 15A, so it starts to be really tricky now. You will see the reason in the next part. But if you have more tips, you are welcome.
As far as I can measure the regulator frequency doesn't seem to be a problem for the Pentium MMX. 150kHz seem to be alright. I think this would be a problem for CPUs for a much higher frequencies.
I know the channel of Great Scott, yes, but I didn't contact him on that regards. I have some friends, who are hardware engineers and I could ask them as well, but it makes too much fund to understand such things on my own. It's always the same with me, I'm more into reading and trying to figure out it on my own, than to ask someone. The most valuable help one can give me is to give a source, where I can read about it ;)
@@necro_ware, I wish you'll make the new regulator to work.
15A is does call for a controller with discrete MOSFETs. Without getting into all the theory of feedback loop control, you should find plenty of very useful information and guidelines in the datasheets and applications notes of the most reputable manufacturers (TI, Analog Devices, Microchip, OnSemi, ST, ...). Good datasheets illustrate how to select the components to use with the controller. The buck topology is usually the most benign among switch-mode power supplies so, if you don't need to do anything particularly sophisticated or uncommon, it should be possible to put together a circuit which works well.
I expect the inductor to be the most expensive component of the regulator. Unfortunately if the switching frequency is low, the value of inductance that leads to a certain ripple is high. This also means that the average output current below which the regulator has to operates in discontinuous mode is high (discontinuous mode is more "noisy"). Finding the right inductor may be quite tricky.
With regard to the "bulk" output capacitor, you'll have to select a low-ESR electrolytic capacitor and then use one or more of them in parallel. The datasheets of the series of "Low-ESR" specify such parameter at 100 kHz (often it's "Impedance at 100 kHz). Most reputable Japanese capacitor manufacturers have one or more series of low-ESR caps in their lineup. In a series, capacitor with the same physical size have the same ESR and the ESR is smaller with larger caps. if the output ripple is high you'll have to adopt an output capacitor with higher current ripple rating so you'll have to keep an eye on this parameter too.
The choice of MOSFETs is obviously critical. The only thing that I would write here is that you should aim for the right balance between RDSon and gate capacitance (and gate charge): generally a lower RDSon is associated with an higher gate capacitance which, in turn, requires the controller to work harder to switch the MOSFET on and off. If you stay at relatively low frequencies, around 100kHz, then the power lost should be low as well but it may still be worth comparing the dynamic characteristics of the MOSFETs you find have a low enough RDSon.
Nice and cool design for retro computing fans!
I had same motherboard back in days.. It's now disassembled for parts or in the box in garage.. Can't remember..
How I recall, it wasn't working and I didn't bother to repair all old motherboards when I get box of them in different condition and repaired only best species.. Now I look to things different - I even repairing battery corroded ones because I love old architecture.
Old boards, usually, is more reliable than modern ones.. 😂
:O New microphone! Awesome upgrade! I was thinking lately if an upgrade would harm the charm of your channel and the overall retro feeling but oh boy, no worries. This is great and I like it a lot! :)
Yes I was playing with a new microphone indeed. My room adds unfortunately a lot of reverb and I'll have to do something about that in the future.... Thank you very much.
@@necro_ware I have a bit of experience on microphone sound optimization and to me it sounds absolutely fine in this video
When you turned the video black in the end I thought it caught fire :O
What a cliffhanger
LOL :D May be I'm still fighting the fire, who knows?
I wish, I've had that module back in the days!
It was during my time at university and was on on a budget as all typical students. This would have allowed me, to upgrade my P120 to the more powerful MMX-Pentiums.
i actually used to overclock the cpu and run it at 3v or something i think it was 166 overclocked to over 200mhz dont remember its been a long time.
My guess is that with this VRM you could use even an AMD K6-2/333AFR-66. At least the 2.2V vcore would have not been a problem. Clock multiplier set on 2 on the mainboard would tell a K6-2 CPU to set its own internal multiplier to 6, which would have been overclocking for a K6-2/333 on a 66 MHz bus. But then you have the opportunity to use 60 MHz FSB (less overclock) or 50 MHz FSB (underclocking).
Very slick work. At one time I had an ASUS TX chipset board maxed out with a Pentium 233 MMX CPU with 256 megs of RAM, a Creative AWE32 soundcard with 8meg of RAM and an Nvidia 5200 GPU. Good times.
Repair videos are nice, modding videos are always the best :)
Ok maybe not so happy about the design options here*
However I enjoyed this either way.
Nice topic, video :)
* It is impossible to cover everything and get into details here, however one thing you can easily fix (and is very important) is placing the diodes on the same side and as close to the chips as possible! The layout of the PCB on the output of these regulators is critical. Ideally you want everything on the top layer there, components and tracks.
Thank you very much! Yes I know all the restrictions, distances, trace widths etc. I was studying the documentation and datasheets quite long. And I was sitting at the design also very long, way too long, if you ask me. I probably wrote a book full of calculations to make it right :D For example the reason why the diodes are on the back side was actually the shortest distance (using direct vias between the pads) to the inductor, which was one of the serious requirements. I also designed the feedback to be out of the inductor flux. Also a hard restriction was to keep the resistors as near as possible to the converter, but it's not possible if you use two of them. So I tried to keep equal but short distances to the resistors from both ICs. I also made a lot of thoughts about heat dissipation transferring the energy using an array of vias onto the back side and many many more tiny but important things. It was quite hard, but I think that I made everything with as little compromises as possible. And what I see on the oscilloscope now looks very good....
That rtc module is pretty sweet
I got a P55T2P4, the evolution of the P55TP4XEG with the intel VX. I was never able to make set the motherboard with i233MMX to run at 233. It runs at 200, very stable even with Windows Millenium. I solded the USB 1.0 port on it, but it did a mistake and 1 burnt... But a Nec 5 USB 2.0 port card, Voodoo 1, Nec PowerVR, S3 Virge DX 375, Sound Blaster 64 Awe, 512 ram cache stick, 96mb EDO Ram, IrDA module, 3Com Pci Ethernet, 2 HDD, 2 CDRW drives, 2x 3.5" floppy drives... with just a "noname" 220v/230W Psu
You have to set the multiplier to 1.5x, since 233MMX interprets that as 3.5x. At 66MHz FSB that would end up at 233 MHz of core frequency.
A very neat expansion for your board that I am pretty sure gonna work. my guess is that the VRM board will become pretty hot when loaded fully if that cpu really gonna need all amps as it stated. very good to step down the 5V rail instead as a major failure might give a smaller risk of injecting lethal voltage on the cpu than using the 12volt one. good job
Please stay tuned, may be some of your questions will be answered in the second part ;)
@@necro_ware I can´t wait, I also have tons of vintage computers like this and I enjoy watching your videos about them, and maybe some can be used with my computers as well in the end :)
I love your hardware hacking add-on modules. You did a great job at explaining PWM for a beginner. I forget if the last time you spoke about switch mode power regulators you explained what purpose the Darlington transistor played. I think people watch your channel to learn a bit as well, so it might be something you actually want to dive into.
I didn't explain a darlington transistor yet, may be I mentioned it in the comments, but in a video I didn't have an occasion yet to explain it so far. May be one day.
There are many types of transistors. One class is called bipolar, and will be designated as NPN or PNP. Those require current to flow either from the base to the emitter (NPN) or emitter to base (PNP) to turn the transistor on. The transistor will then allow a proportional amount of current to flow between collector and emitter. The more current that flows through the B/E junction, the more can flow through the C/E junction. In that way a bipolar transistor is kind of like a current amplifier.
The problem is, you need to provide that proportional current through the base to get the C/E to open. If you’re trying to drive that current from a low-power source, like the output from a digital logic chip, you may have trouble opening the spigot wide enough to avoid limiting the current you’re trying to control with the transistor.
In that case you use a pair of transistors in a Darlington configuration: One as a first stage amplifier, which then controls the base of the second with an already-amplified current. The pair will take a very small current and boost it, twice, to a very large current. In that way, it acts like a cheater bar on a wrench, providing additional leverage.
A single transistor might amplify the current from B/E to C/E by something like 100 times. In a Darlington configuration, the current in B/E of transistor 1 is then multiplied by that 100 times, and that is fed to transistor 2 which amplified it another 100 times for a total current gain of 10,000.
Another type of transistor, called a FET, doesn’t have quite this same problem because the “base” (which, on a FET is called a gate), is controlled by voltage, not by current. So provided you can supply sufficient voltage to the gate, there’s very little (negligible) current required to control it. However it is still possible to find cascaded configurations with a driver transistor and a load-switching transistor due to things like .. needing higher voltages to turn on the FET (which may require 10V and need to be controlled by 3.3V or 5V logic), or how all gates have a certain amount of capacitance that, when switched at high frequencies, can put strain on whatever is driving it.
Subscribed, this is awesome engineering keeping the best of these kinds alive!
Looking forward to part 2.
Nice! I have a Micronics board which had the VRM socket, but did not officially support MMX and a newer Micronics board with the VRM module in the socket. Fortunately swapping the VRM between these 2 was enough to enable the use of Pentium MMX CPU. But I have an Intel Advanced/EV board which I like a lot that has just the solder pads for the socket. After this video I will have to research if it also can be upgraded.
Highly Likely. Some later boards came with the VRM socket but only had a little dummy module installed that essentially just did the 4 jumper wire thing like was done on this board in hard-wired form for non-MMX Pentiums. Even some OEM PCs came like that (HP Vectra come to mind). Gave you the option to upgrade to a MMX CPU (good for marketing) but you had to pay for a VRM module (good for business) :P
This video is such an inspiration. Thank you!
This cliff hanger was better than any one in the last three seasons of Game of Thrones. I'm eager/angry.
lol :D
lmao that's a hell of an expansion card, wow!
Wow, that is some real hacking there - I really enjoy these vids, I understand the terms you use, but wow, your smarts are off the chart! I have a bunch of old motherboards - nailed to the wall as trophies (my first 386 through to dual PIII boards...)
I recently got given back one of my older builds - a Rhino III and a Quantum bigfoot, that was found in the back of a garage.
Thank you for pleasant memories from childhood.
I think nobody made 430FX boards for MMX Pentiums, because SDRAM support of 430TX dramatically increased overall performance.
That's true, but not only because of the memory. The later 430HX and VX could also cache a lot more memory and were equipped with faster cache. In the end TX was kind of flagman for Socket 7 from Intel. After that they moved to P2.
Voltages where so analog back then. Jumps in between of 0.100v, i.e 2.8V, 2.9V, 3.0V etc while these days due to the digital VRM implementation it can be so accurate till the last milivolt if required. Loved watching this vid. Did'nt know these Socket 7 boards could be modded all the way.
Thank you, that was just an example, please don't nail me down to every word
Not wanna be a smartass or give non-needed advice but you inspire many people to fiddle with retro hardware. Would be lovely to see that you ground yourself or advise your audience to do it so. Static electricity can really destroy the good mood :)
I have a strap, just use it really seldom. I always touch a grounded wire before I start working...
I love watching people do things never expected and new!
This gives me realy good memories. My first 'own' PC had an 133MHz CPU and I overclocked it just by jumperconfig. It runs not realy stabile on 200MHz, but without failures on 166MHz. And this with stock cooler an nothing more in the case, as the fan of the little PSU.
Excellent video! I also noticed that the audio quality is much better than in your previous videos.
Thank you. And yes, I was playing with a new microphone, but my room adds a huge reverb and I'll have to do something about it.
@@necro_ware Sound absorbing foam.
Super cool! I love seeing people add solder in mods to commercial products. Keep em coming!
Wow, that expansion card is so cool. Very nice find
Evil Cliffhanger! I look forward to part 2!
I would Love some more DIY and upgrade videos! Excellent!
i reconize this MB my dad gave me the same one back in the day. and then i read the manuel and i found that i could also do this same thing :-) but we never realy got around to doing that because back then ics wer verry expensive and we had plenty pentium mbs that can take mmx cpus but my dad did put an resistor at the back side and mmx did work but not fully clockspeed. he didnt have a heat gun eather and the cost was just simply to much so we didnt do it. so nice to see you making it work after all these years :-) you are the best im verry ha[ppy to have found you . i tyhought i reconized it from somewhere lol such a long time ago this brings back nice memmories thanks yo so much
26:48 lol Yep, Back in the day I had a P166MMX overclocked to ~210 Mhz (undocumented 83x2.5 setting, if I'm not mistaken). I had it for about 5 years (since 1997 until around 2002 or 2003), since I was a poor student back then and had no money for an upgrade. All those years I didn't even bother to open the case, since everything just worked fine, sometimes 24/7, and since I didn't tinker with this config due to the lack of money. So when finally I made an upgrade, I opened the old case to salvage a sound card and perhaps some other parts from an old system and what I found out was that the CPU cooler wasn't even connected. lol I guess, the guy who assembled my system in the shop forgot to connect it and it worked all those years without any problems with just a small aluminium heatsink (btw also without any thermal paste between it and the CPU)! I was actually very impressed, since it was also overclocked and I played quite a lot those days, so God know what sort of temperatures that CPU sustained and for how long.
Nice hardware from back in my day, I think I've got a socket 7 MB in the loft somewhere.
Very nice explanation that was easy to understand :) Time to make another retro computer news video if you don't mind me using a few seconds of yor video :)
I don't mind :) I like your news videos, but may be it's worth it to wait for the second part first. I'll try to answer some important questions there.
@@necro_ware ok thanks I will wait :)
..remembering my 486dx2 days.. can remember playing quake 2.. for YEARS using it.
Quake 2? 486DX-2? Are you sure? ;)
This is cool beyond description! Thanks!
I predict you will have a lot of subscribers in the future. Good job.
Yep, this is how the older boards had you overclock. None of these fancy bios controls, we did everything with jumpers.
I loved the video! great job! Waiting for the next one, greetings from Italy!
Well don't take too long to upload part 2 :) I might miss it
Loved this video, great work hacking this motherboard.
Thank You for great and interesting material.
I finally understood how switching regulators work.
This is already a success! :D Glad I could help.
Good job for making your own VRM, I knew someone had to make one at some point ! I'm not a great fan of cliffhangers, but it's good otherwise !
Hehe :D Well, I don't think, that you can find anyone, who would agree to love cliffhangers. But this one was for a reason, there is a chain of further issues, which would be each a cliffhanger for itself. Would I put them all into one video, it would probably last the next 1,5h. So I have to sort out first what is not so interesting and think about how I can put everything together. So I was standing in front of a decision to make a very long video (probably over 1h) in the next two weeks, or to make a cut now, reorder the materials and make two shorter videos, where I'd be able to release the first one now. The way was obvious for me, even if I have to tease you a little bit :D
@@necro_ware No problem :)
@@necro_ware Out of curiosity.. do you think there's any way to adapt a CPU that has a lower IO voltage than 3.3V? I've got a pretty rare embedded Pentium MMX 266 kicking around that I'd love to use but I'm afraid of damaging unnecessarily. Reason is quite simple... it runs at just 1.9V VCore (the lowest any of my boards go is 2.0V, which I guess would be survivable, maybe even help it get to 300MHz), but in addition to that it runs at 2.5V IO which has turned into quite a headache... unsure what to do other than try and find a board designed for embedded Pentiums... ie needle in the haystack. Unless someone comes up with some other idea
@@Knaeckebrotsaege well if I understand the pinout properly, it might be possible since there are io voltage pins
Excelente trabajo. Esperamos ver segunda parte😀
No esperes más, se publicó hace mucho tiempo ;)
Good video. I was always a fan of the Asus TX97 and HX97 boards.
Thank you, this board however is from the time where TX and HX were still the future....
Great Video ! Reminds me of the old days .
Excellent!
Wow, your audio is much better on this one.. Just wanted to comment on that!
Yes, I was playing with a new mic, but I have a bad reverb in my workshop and will have to fight that somehow.
Thanks for help me to finally understand PWM!
Well done for the double buck regulator. However, your explanation is missing one key point, the schottky diodes. In a buck regulator, the diode provides current drawn from 0v and pumped by the coil each time the transistor is cut off. That's the only way to output MORE current than how much is actually injected at the input. That's what allows a better efficiency than a linear regulator. In your case, you need 5 amps, and to lower from 5v to 2.8v. 5*(5-2.8) = 11 watts and 5 amps + internal regulator needs drawn from the input. Considering a efficiency of 80% for your circuit, this buck VRM will ouput 2.8*5 = 14 watts of power, and we can assume it will use 17,5 watts with efficiency of 80%, so only requiring 3,5 amps from the input, and therefore pumping 1,5 amp through the schottky diode. The total dissipation will only be 3,5 watts...
Was very surprised to hear a russian song in the background)
as always, very good job!
Да и ответы в коментах, как бэ намекают
That’s an evil cliffhanger 👿. But I remember I have that mainbord back in the day with a pentium 1 120mhz and an asus video audio combined card.
i can scarcely imagine a time before sse or even mmx. a lot of what i use my cpmputers for use those types of instructions heavily and man does it make a massive difference
Лайк сразу. Жду 2 часть )
New sub for me. Very interesting. Also nice to find model povedeniya.
Can you please tell me the name of the song, I can not find it. Thank you.
Cliffhanger..aaargh! :D lol, fantastic video!
Nice video, keep it up, thank you :)
Liked the music.
Hello Necroware,
Keep in mind without load the big buffer caps could keep the higher voltage after you step down the regulated DC voltage, always measure in the opposite way, increased the voltage never decreased before measure!
Regards, Mark
I have the same problem on my hand, a Socket 7 MB with Pentium MMX, and with the VRM socket, but without the VRM module itself. I've bought a bunch of cheap DC-DC switching converter modules from Aliexpress, and now I'm trying to imagine the wire gauge need for 6 A, and how I'm going to physically wire the module to the VRM pins. There's also a problem with the trim potentiometer being too sensitive and drifting off the desired voltage - need to replace it with a 20-turn pot.
Also a way to solve the problem :)
16:20 Real man designs his own VRM!
Man, you are some true pcb wizzard!
I think I have this very same motherboard sitting around in a stack somewhere.. might be amusing to try and toss a VRM on it some day.
That was an awesome video.. thanks so much for all your work..
Fantastic video as always :)
20:00 this calm hand....i thougt my pc was hanging up ^^
:D LOL
I thought I was watching a Socket 7 upgrade video but suddenly it turned out to be ElectroBoom digging into computers.
w8 for part 2
Amazing enhancement ! Thank you !
and that's why it's a good idea to test your projects before mounting them and feeding 5v into something that expets 2.8v :D
Double voltage = double speed, trust me
Fantastic video.... But the cut is verrryyyyyy evil ;-)
i wish i have skills at this level! Good job
I am only going back to Core 2 Quad Q6600, I had to live through the Pentium years, my first being a Pentium 60, and I don't miss those old processors at all.
It depends on what's your aim is. If s.o. is just a user, then a modern system is fine. I don't even see the point to use a Q6600 in such a case tbh. But if you want to learn about electronics, about how computers work etc, then the older the better. In modern systems everything is highly integrated and there's not a lot of space for experiments.
I'm surprised you didn't test your VRM under dummy load before subjecting the CPU to it.
Actually I'm surprised you didn't put any load, as switching supplies and even some linears can be unstable without load.
Also I would check how clean the output is with an oscilloscope on no/low and high load.
Sure I did test it with a dummy load as well, I just didn't show it, because it was boring. Everything else... well I just didn't show it (yet).
I had that board at one point in time.
Seeing 64mb of SIM ram brings a tear to my eyes. People used to be like you have 64mb of ram I only have 16mb
That's 128MB, but back in the days I had only 8MB on my Pentium 90.
Necroware - you are totally evil! Thanks for all the theory you explained it very well.
You are lucky, the second part was released already. Others had to wait a week ;)