Let's upgrade an early Socket 7 board beyond the limit (Part 2): VRM, MMX, Testing
HTML-код
- Опубликовано: 6 сен 2024
- I have an interesting Socket 7 mainboard Asus P/I-P55TP4XEG, which I wanted to use for an amazing retro PC. In that mainboard a huge potential is hidden, so let's see how far we can get. This video is the second part, where I would like to test the VRM module for the dual voltage CPUs and see how far we can get.
Music by Model Povedeniya
modelp.bandcam...
Patreon:
/ necroware
(motherboard with incompatible voltages)
Random youtuber: «Oh well, let's find another board on eBay»
Necroware: Goes «Beast Mode», builds several VRMs from scratch, pulls official datasheets out of a cavern, defines multipliers manually with resistors, works flawlessly.
Also Necroware: «It didn't explode, I'm disappointed»
Absolute legend. XD
ROFL :D
In part 4 we will take a look at a simple tweak that will allow us to run overclocked Ryzen 5 5600X on this board, right? :D Great video, thanks!
Thank you for the idea :D
@@necro_ware ... I- I- think that was a joke but, go ahead :D
@@jojoosinga can u imagine how big a bord back then would have to be since they weren't trying to inigrate alot of chips to multiple tasks wow it would take up a room 🤣
@@trashtronics1700 Sure they were. This is a board with integrated serial, parallel, floppy, and hard disk controllers, and most of it is in two chips along with the DRAM controller and interrupt controller and ald that fun stuff.
@@CptJistuce imagine having to design a bord back then that would support something like a Ryzen it would be at minimum about a billion chips lol
"Every day where I didn't learn anything is a wasted day." - ... and they say the best way to reach a deeper understanding of what you've learned is to teach it to others. Thanks for teaching us what you've learned.
A video from CPU Galaxy, Adrian's Digital Basement, Sucra, and Necroware on the same day.
It was a very good day.
I love old Super 7 times. You had one CPU slot and multiple CPU's to buy, from different companies: intel, via, cyrix, amd etc....
Thx for the vid. Can't wait for part 3 :)
I agree, super socket 7 was so versatile, super socket 7 created by AMD (used by all ) was the turning point forced INTEL to create its very own socket the slot 1 (AMD answered back with SLOT A)
@@michaelperugini4199 And that's how the war started between AMD and Intel.
I miss the lga775 era...
If we could only have a modern universal socket and more company's that produce cpus I missed out on the socket 7 era started computers around the slot 1 slot a era it must have been a blast using all those different CPUs on one system
@@rikka0_059 Only Intel on it, though
Oh, man. I had _actually_ forgot that you used to have to set jumpers to select bus speed.
Thanks for the memories.
Socket 7 is -one of- my favourite. I owned a K6-2 and you reminded me that I did - Awesome chip. Along with the venerable Cyrix 333. I loved this time in computing so much. It was amazing. So many memories.
You wanted to walk a mile, but ran a marathon! This was a great 2-part series to watch.
I personally identify heavily with everything you said about a software engineer trying out hardware hacking as a hobby and making mistakes while at it (and I did see the buck converter comment in the last video). I'll only say, please don't take it too hard on yourself. Especially if someone types a long but factually correct comment in your videos, then it's a positive contribution & they mean well.
As a typer of a long and (I hope?) factually correct comment, I definitely mean well. Mistakes and misunderstandings are inevitable. I think it is important to identify and correct those mistakes so others aren’t led astray, but if you accept and own up to your fumbles, all is well! Because, let’s face it, we _all_ screw up. It is a part of learning. Nothing wrong with that at all. :-D
Holy crap, the thumbnail is SO COOL!
This "Pimp my Motherboard" series is really interesting. Stage 1 and Stage 2 were cool. Looking forward Stage 3
The Larry David bit at the beginning was hilarious!
;)
Why is it that I love these old machines do much?
Very interesting video.
I'm not an expert in electronics and discovered this channel today, but gosh, it was a trip.
I just witnessed someone modifying a motherboard as old as me for it to support a more recent CPU.
Then, managing to run the CPU of the opposite company on it, and then overclocking it.
That was something
Great work! I can learn always something from your videos. 😍👍🏻. Thanks for sharing.
Hi Peter! I learn a lot of interesting things from you and I'm very happy if I can give something back ;) If that works in both directions it's absolutely great!
and by the way, I love your way of fun you brought into this video at the beginning. 💥, thanks for watching, no kiddin… hahaha, I was cracking here 😂👍🏻
You say you're not a hardware developer, yet you are developing hardware. The VRM work alone is amazing!
thank you very much for the video, its really nostalgic for me . i love theese times where the users good to be know how computers work, set jumpers playing fantastic DOS games etc. iam already your fan now. Cheers from Hungary :)
Amazing video. My favourite retro channel. Thank you!!!
"...simplification is the key for the solution for many problem..." Yes! This!
Thank you for your insight sir, and I've really enjoyed watching your videos.
The versatility of the socket 7 platform always surprises me! Really enjoyed this video! Just incredible how far you were able to stretch an old 430FX board. Can't wait to see the next revision and may give it a shot on my 430FX board once the designs are available.
As always nice video.
24:14 Because your videos are good and detailed thats why people assume that. I think that is a compliment :)
Thank you very much! I really don't mind if people like my videos and as long as they are seen as entertainment, I'm really happy. Important is, that nobody takes my channel as a tutorial, that could be dangerous for the health and hardware of the others ;)
Regardless of what you do, if you don’t learn something new every day, you are missing out.
Education doesn’t stop in high school or college or university. Always try to reach further!
Thank you for this fascinating experiment!
Great project! You really took this motherboard far beyond its original limitations.
These videos are so awesome!
Amazing work, great to see nostalgic hardware pushed beyond believe.
Hanging on a cliff for a week was surely worth it. Nice to see how far you can push that mainbord. I still miss my old pentium 1 120mhz and asus p55 main board. If I ever find one, I really want to mod it like you did. Entertaining and honest video👍
This is the first time I've seen your videos, and what a great series! Learned a whole lot about something I never thought would be possible. I found the subscribe button was very easy to click.
Very impressive… it’s been years since I had a Socket 7 board so it brings me back to when I was 12.
Nice experiments.
I didn't know that AMD K6 can run in same mainboard socket as Intel P1.. I even never compared pinout.. 😂
Yeah, it's great to learn something new every day. 👍
All Sockel 7. Sockel 5 CPU will work, too
Congratulations! This is most impressive. I have the P/1-P55TP4N which is slightly different and has support for the P55C-MMX, but does not have the external VRM traces that your board has. So far, I am afraid to test a powerhungry CPU in the integrated VRM.
I love this video series, keep it up!!!
Great Video 😁
Thank you for sharing your projects, I look forward to learning from your mistakes as much as your successful DIY attempts and 'am grateful you include them as you theorize.
Cool videos! I'd like to see it more often. Thank you.
Amazing, this is what i loved about old pc modding.
I'm appling some tests with candidates in front of me, and then I see that explosion... You don't know how much I laught...
LOL :D
I don't know how I haven't found your channel before, but this stuff is RIGHT up my alley. I love repairing retro PC hardware, and the way you're approaching this project is exactly the type of stuff I've done before, myself. Thank you for detailing not only what you're doing, but why and how. Digging into datasheets and unlocking what's possible is a really awesome topic. Instant sub.
your recent projects are very similar to mine. It's actually kinda weird. I am not even using google to search for information about my project. So, when I saw a bunch of these vids reccomended I was surprised and though "wow the google spies finally got me"... Then I was absolutely shocked to see every time I go into google to listen to some music, you upload a video about what am about to do... "10-30 minutes ago"... Like we have some telepathic connection! :D :D :D
Subscribed! This was a great adventure and I'm looking forward to the next part. One thought about not open sourcing something because a better version will be coming: if you open source it, there's a chance someone will build upon what you've done and get you to the next step faster!
Out of experience, there is a bigger chance, that someone will build it and complain, that it gets too hot with a K6-2@400MHz. Unfortunately far too many people hear, read and understand only half of what is been said or written. Furthermore, my aim is the way, not the result and the best way to learn something is to try to make it on your own.
@@necro_ware fair enough, especially on the latter point of learning song the way :)
Watching this again as I picked up a random mix of forgotten computers today and this board was hiding in one of the cases.
I don't exactly know how you ended up on my recommended but I think what you are sharing is really cool so I'm def going to sub. Keep doing what you are doing!
This channel may not be a tutorial, but as someone with 30+ years exp with electronics, I can tell you it's a decent starting point. Always do your research, as Necroware did.
I just saw both videos and wow, this is a deep dive that really interests me!
If you're not yet willing to try the motherboard modification that adds the third 'jumper' for the proper 5.5x speed, and are worried about overclocking that 366, maybe try a 60MHz bus speed instead of 66? That combined with the 6x multiplier would give it an internal clock speed of 360MHz, close enough to its rated 366. Then again, that runs the PCI bus at 30MHz, slowing down all the peripherals, leading to slightly slower overall performance. Once you get that far, and are interested, I would be interested to see benchmarks between 6x60 and 5.5x66, see how much the bus speed really makes a difference. (If only this board could do 75MHz bus, and a 4.5x75 at 350 comparison to 5.5x66 at 366, see how much of a difference bus speed makes!)
Spectacular hack! 400% speed increase. If only I had this knowledge back in the days!
G'day mate I like how you admit you are not a electronics expert much like yourself we are qualified in different IT fields I am a network engineer/pc tech with ms certification and linux plus etc etc I myself do electrical engineering as a hobby people like yourself give me more knowledge for the future so I thank you for a new learning experience within my hobby with EE it taught me how to repair pcb's how to solder different types of chips etc but in the future I would like to try some things with a old lga 775 board and now you have the engineer in me to think what can I do I have to do more research on what I wish to acheive if at all it could be plausible please keep the videos coming they are very good enteraining well explained and informative thanks mate
I can hardly wait to see the benchmark improvements!
Amazing to see that the regulators behave so well at almost double the specs. Thanks for sharing.
Yes, I was also surprised. I actually thought, that it would explode.
@@necro_ware I'm pretty sure they do have some thermal protection so I would assume they simply shut down when overheated. .
Very nice video :) This will be a very unique pc :)
You did an outstanding job on explaining the FSB options. I'm looking forward to your new VRM design. I've never found a ready made and adjustable module in the 10A+ range that was small enough to mount on a mainboard. I think good availability and cost of components on a VRM design is a lot more important than efficiency. So even if a synchronous buck converter would run more efficient and cooler, paralleling readily available and cheap standard buck converter solutions may be a better for the retro community. Whatever you design, it is probably never going to get as hot as the original linear regulator on the board anyway. I rember having burned my fingers on those more than once in the 90s. I wish there would be a tiny module that could be mounted in place of the TO220 transistor that most boards have (and maybe reuse its heatsink). Btw: Are you sure the clock generator on that board can't do 75 or maybe even 83 Mhz? ;-)
Thank you, glad that you liked it. As I said, my new design is yet only on paper, but I'll order the parts for a prototype soon and we will see, if it can stand 11A+. On paper it does, but reality keeps often surprises fur us ;)
The FSB of 75 and 83 would be only possible with a mainboard mod and then all the buses would be overclocked. Out of curiosity it would be interesting, but because of stability I wouldn't do that though. If this board eventually runs cool at 400MHz with a K6-2, that would be already a great success for me. I think, I'll get that running and I'm very exciting to make some benchmarks then.
@@necro_ware A 75 MHz FSB is actually pretty stable in most circumstances. It runs the PCI bus at 37.5 MHz and ISA at 9.375 MHz, both of which are within the realm of possibilities. PCI 2.1 introduced the 66 MHz bus spec, so PCI 2.1 compliant cards from 1995 and later are generally fine with bus speeds above 33 MHz. But even cards designed for the bog standard 33 MHz bus were generally tolerant of 37.5 MHz. An 83 MHz bus on the other hand will give you a bad time. Cyrix showed us that.
ISA on the other hand never had a defined clock and ran from 4.77 MHz up to 20 MHz in badly designed boards with no clock divider for the bus. Early ISA cards designed in the IBM 5150 to the AT era often didn't tolerate clock speeds above 6-8 MHz, but in the mid to late 80s as the ISA bus speed increased, so did the prevalence of cards that could tolerate higher bus clocks. Late 80s and onwards ISA cards generally could run up to 10-12 MHz without issue, and some faster than that.
An AMD K6/2 with a 75 MHz FSB would be a great improvement over a 66 MHz bus. A 400 MHz K6/2 on a 66 MHz bus is just painful. I had such a system in the late 90s with one of those Evergreen Spectra upgrades, and while it was definitely faster than the 200 MHz Pentium MMX it replaced, it was still a pig when compared to a K6/2 on its native 100 MHz FSB.
A 3 pin connector to replace a To220 could be insert in the same PCB.
@@rath9303 There are TO-220 buck converter module replacements for linear regulators, but they tend to be noisy. It may be fine as a substitute for an existing hot linear regulator, but not if you plan on stuffing in a higher power CPU. Unlike the VRM socket where you'd have multiple pins for power delivery, the TO-220 just has one, and cranking 10+ amps through it isn't a great idea.
@@GGigabiteM i know, that with 3 pin this regulator can replace a damage linear regulator.
I'am sure that if i select 0,1V more, the CPU will get the right voltage.
If i replace the CPU capacitors and connect the to220 pin direkt to them, 10A or more are not problem
Maybe the new VRM should provide 3,3V, too as this saves power compared to the linear onboard Regulator.
I think 3.3V is provided by the PSU
@@bene5431 Not on the old AT power supply, only by new ATX power supply
@@bene5431 the 3,3V from the PSU has not enough Amperage and used only for PCI. There is a linear Voltage regulator next to the CPU to provide 3,3V for the CPU. And Linear is inefficient.
@@rubberduck4966 So they used a linear regulator to generate the same voltage again because the PSU doesn't have enough power on that rail? that's stupid
@@rubberduck4966 I watched part 1 again, and you're right, it seems to be on the heatsink below the CPU. It's still stupid
Wau, wau, wau - you are doing amazing job...I really love this channel.
I suspect that adding a small ceramic capacitor right between the output leg of the regulator and ground would dramatically improve your ripple situation. You could probably get under 5mV of ripple with a 20uF MLC capacitor, costs $0.10. You could probably even add it to the board you have already, just bodge it in there and test.
Also, if you want high currents, look for a switching controller, rather than a switching regulator. They have transistor outputs that you use to drive high-power output transistors. They make the board a little more complex, but you can get essentially whatever current you like with them and pick switching transistors that support that current.
Also look into synchronous controllers. It's one more transistor instead of a diode, but they're dramatically more efficient, meaning less waste heat for you to deal with.
I tried different ceramic caps on the output behind the scene, but I couldn't get any significant difference.
Another concern here might be that the motherboard traces themselves maybe aren’t designed to handle 11A of current… could it damage the board?
Yes, it is a possible problem, but on this board it looks good IMHO.
@@necro_ware nonetheless it would be wise to at least add a floppy molex (or larger) to the PCB. The fact that this board probably has enough beef to withstand the power, doesn’t speak for every board.
Anyhow, great and awesome content!!!
@@Inject0r Yes, I'll keep that in mind.
IMO, look at the VRM pins. Those pins are good for, I dunno, 1-2A apiece? How many Vcore pins are there? If there are six or more, I wouldn’t worry about the motherboard. If there are only the two fed from those jumpers, then that might be pushing things.
The V+ pins are probably fed from one of the inner planes. Which means 1) you won’t be able to see the traces, so you won’t be able to gauge it yourself, and 2) if it’s a power plane, you’re more than likely going to be JUST FINE. ;-)
@@nickwallette6201 true. Though that doesn’t count for every board. Some boards don’t comply to standards. It’ll be very wise for “us enthousiasts” to add the floppy connector in the PCB design. It doesn’t have to be populated if it’s not needed. 🙂
Very nice video, had this board back in the until it broke down, If I only knew what I know now I would have replaced the voltage regulator and tried to repair it. Maybe you can also try to get an k6-2+ or 3+ getting rare but that could be the max for this board…
Thanks, unfortunately I have no K6-2+/3+, so for the target is a normal K6-2 for now....
You are my new hero! For many reasons I can't describe now
That POST analyzer card is extremely useful for retro PC's. I had one 486 where the reset switch got stuck, and it was impossible to see. I thought the board died, as that button got pressed, while I was swapping processors. I mistakenly thought I killed the board, as even the old CPU wouldn't boot up after that. I looked at the analyzer, and saw the system clock wasn't running, which was odd. The crystal wasn't sending the pulse. I checked the reset switch and yep. It was stuck in the on state.
"I think that I don't have a lot of time" :D Amazing video!
Just wanted to say, nice channel name, good play on word ! hope you get more famous soon !
A French Baked Sweet.
one of the first "nice" computers I built as a kid was a pentium mmx 200. that's around the time I started to figure out the internet and piracy. I must have reloaded windows on that box thirty times.
thanks for doing this stuff it's great to see someone have fun like this.
Very nice indeed, Great to see how good your board could handle the load before it got hot and impressive how fast this motherboard can run at the same time. what you could do if you plan to make a newer revision of the VRM is to put an external power connection on it for more current capability (kinda like the P4 power connector) since the mainboard itself may be too weak for that amount of current. but if you put a power connector on your power board you can chose to input 5 or 12 volts to the regulators as you please and have plenty of juice for you AMD. a part 3 is called for here! Good job Necroware!
A good idea, with the external connector. I'll double check the traces on the board and see if it is necessary. I'll keep this idea in mind.
@@necro_ware I think that the external conector is realy necessary
You are featured in Hackaday!Congrats my friend!
I love projects like this. Nicely done! Looking forward to the next one!
Wow, I had no idea these AMD K2 CPUs were THAT power hungry. I have a 550MHz variant and no wonder it has problems with my SS7 mainboard. I bet the regulator circuit cannot provide enough power for it. Looking forward to the next part, which I'll probably use as a guide to completely replace stock voltage regulator circuit on my board :)
with great performance comes great power consumption...not any different than inhell of the time
If your 550MHz is a K6-2+ or K6-3+ it draws way less than the older models. Die shrinks were massive back then.
@@andreewert6576 Nope, just a plain old K6-2. "Plus" versions are very rare where I live, and I'm not quite ready to buy them on eBay for those prices (-:
Your work is truly impressive. This VRM module looks very neat, too👍
this is great stuff to see, and makes me wonder if you'll take what you learn with this to the 486 and make some 5 volt to ~3.3 volt adapter for that platform
I wanna see if he could break the 200mhz barrier on a 486!
I think you get an Overdrive processor.
Will not work on most 486er boards because they direct connect to 5V.
Maybe it is possible to cut the trace and give the CPU separate 3,3V
Some Boards have a design für VRM and use jumper or solder bridge to replace it.
Here you can use a very simple 3-5A LDO Linear, because they don't need so much power.
@@rath9303 there also existed adapters that would fit between the cpu and the socket to provide 3.3v
@@CookieJar6502 i know. But not easy to get and a little bit expensive. If i have the possibility to insert a LDO instead a extra socket it will choose the LDO.
i feel like this is a Massive success for a first run!
I have a BS in IT and a AA in electrical eng. I have been fooling with computers since 1979. I have never seen anything like this. I get it.. I'm with the simple camp. Love your attitude too. To easy. Awesome.
just from watching part one and then seeing that intro I had to sub thank you so much for the informative content
More of this hard core over locking please. Great.
Well this was actually not overclocking, but more of an upgrade :)
Excellent work as always, Scorp! Love how you made your own little VR boards!
I have a similar old 66MHz FSB socket 7 board. Mine's from Asus with an P200 originally but it's slightly newer than yours, supports MMX/3DNow by default and apparantly has a hidden undocumented feature to make the on board VR run at 2.2V, really cool! So I just set the undocumented jumper settings, popped in a 400MHz K6-2 (Chomper) and set the multiplier to 2x. The AMD makes it 6x internally so it's about twice as fast now than with the original 200MHz CPU :-) It's a super nice and stable system. Next: maybe 100MHz FSB and maybe some overclocking to 450MHz if possible.. Old school PC hot rodding FTW! ;-)
You need a working 4.5x multi to get 450MHz on 100 fsb. If you're stuck with the 6x but can tweak the fsb, 75x6=450 ;)
@@andreewert6576 Thanks, my Asus (late socket 7) 430TX mainboard is sort of "in between", so there are a lot of undocumented features. It runs a 66MHz FSB now but I believe it's possible to tweak this up to 75 indeed. I don't think 100 is possible (chipset is too old and I may fry the RAM as well).
The K6-2 handles a 2x multiplier setting as 6x internally so I should be good. Might also have to tweak the voltage settings a little bit as well though...
@@pipschannel1222 in my experience, the K6-2 don't respond that well to additional voltage. You shouldn't need to go past 2.4V.
As for the RAM: you won't fry it, as long as you dont try to give it more voltage with a mod. You might need to run it at looser timings though, or use 133MHz sticks.
@@andreewert6576 I know, they're not the most overclockable CPUs... It's at 2.1V now so I might have to go to 2.4V to make it run stable at +50MHz but that's the max..
They run pretty hot already so I won't go beyond 450MHz and 2.4V.
I'll have to read the manual but the FSB may also support 100MHz indeed.. Still have to look into that when I've got the time ;-)
I like to watch these videos. Thank you.
My understanding from my old EE lectures is that running two independent VRMs like you do means that you can easily end up with a lot of the power running through one or the other (IE not balancing properly), or in worst case even back-feed one of them which will blow it. But I'm not an expert EITHER!
Modern CPU VRMs often use "teamed" power stages which might sound similar but those actually share controller and PWM signal (and require some care during design to work) so it's very different from what you're doing but might be why some think it's safe. Or the fact that as you've discovered it does tends to work most of the time.
I think your idea of a single 15A regulator is better and simpler, I would probably add an auxiliar molex power plug on the board though (unclear if that is already in your plans).
If you want better regulation/less input/output ripple than that gives you I think the next step up after that is probably a multi-channel VRM controller with built-in drivers and then use the design in the spec-sheet. The built-in driver part means the design is much simpler and needs less parts, IIRC AHOC mentioned that there are standard VRM controllers with up to 4 channels with built-in drivers, though an 2 channel controller is likely much cheaper (and needs much less pins which means a nicer form factor) and is still a big improvement over a single-channel design.
Yes, the two regulators was probably a bad idea indeed, as I mentioned it in the video. I had no problems during my tests, but you never know. So I already designed the new revision - now it's a synchronous voltage regulator with one DC-DC controller and two N-MOSFETs. It's not tested yet, but afterwards I would say, that it was everything else, but simpler. It was by far much more complicated to calculate everything properly. I also had to read and talk to my electronics engineers friends a lot to understand some really nasty small things, which I didn't have to think about in the previous revision. The devil is as always in details.
I recently looked at getting some retro hardware due to the fact a lot and I mean a lot of industrial stuff is based on similar hardware. Shit used to be dirt cheap, then came retro builders and people buying up old stock to mine gold. I remember buying a stack of 486 main boards, with CPU and some assorted cards, new old stock for $3 each back in the early 2000's. At the time those boards were maybe 8 years old.
Hey, great video! I built my first computer system in 1998 using a 430vx board made by company FullYes. I kept the system for many many years. About 10 years ago, I also found out about the k6/2 trick with the multiplier and got it to run at 400MHz. My only problem was bios recognised it as a 486 but it functioned correctly in all tests. I didn’t realise the power requirements for the k6/2 were so high! Not sure how I didn’t kill the board, but it worked in that configuration for years until I sold it all
This vid got me interested to tinker again :)
Back in 2001 I had a Biostar MB-8500TVX motherboard running an Evergreen branded IDT Winchip 200 MHz CPU. It was in my spare pile when my BIOS update on my Abit KT7A went wrong, so I pulled it out and installed a K6-2 450 I had so I would have something to use while waiting for a new BIOS chip to arrive. I knew nothing about voltage regulators, but I had used an Asus P55T2P4 with a K6-2+ @ 500 MHz (inspired by Oldie Tuning on Tom's Hardware) so I thought that running my 450 MHz CPU at 400 on this board would be fine. One hot summer night, I was playing Jane's WW-2 Fighters on it when I heard a loud pop, and it died. I didn't know what had failed, though I later found that not only the motherboard died, it had also taken out the CPU. I now have no doubt that it was the CPU's voltage regulator that failed as I now have a Biostar TX chipset AT board that I ran for a while with a K6-2+ at 450 MHz; this requires you to have a small fan blowing over the voltage regulator to keep it from dangerously overheating. I replaced the Biostar with an Asus TXP4, which is a superior motherboard to the Biostar, and my assumption is that they used a switching regulator rather than a linear one since the Asus board has no issue with the voltage regulator overheating even when running a K6-3+ CPU at 500 MHz.
Looking forward to your next video on this board!
I would like to add on the point of correction that the graph is also not really correct, it's not "waiting" for the tension to go up/down around the desired tension, it's full off/full on every time, what happens is that the inductance/capacitor that are part of the VRM are going to average the tension from the MOSFET doing the commutation, hence giving the final voltage that aspect.
I was pulling my hair out (what little hair I have left) trying to find out why a Pentium 200MHz MMX chip was ignoring the 3x multiplier and remembered this video! In my case it was a much newer board though, based on the Intel 430VX chipset. It's an early revision 1a Gigabyte GA586ATV, but appears to support MMX chips out of the box with dual regulators for 2.8/3.3V chips and the appropriate jumpers to switch between. It has a similar pair of jumpers for setting the multiplier. I assume they probably fixed this in later revisions. I'll be giving the resistor mod a try soon! Cheers! 👍
Just curious - can it be reversible? And how do they implement that in the motherboards supporting both non-MMX and MMX CPUs?
You should also consider the inductor current rating; many switching supplies on mainboards use multi-phase designs to reduce the current through any one inductor and to create a more stable output.
This brought me back to my old job where we had about 30 socket 7 and super socket 7 boards when I started in 2000. Once I found out about the k6-2 6x multiplier we bought enough k6-2 cpus to upgrade all the p54 and p55 cpus. Luckily all the random motherboards(almost all were pc chips boards) supported 2.5v or lower already and just needed bios updates to bring them to 400 or 450(using a 75mhz bus)mhz. I also remember I had one board that had a 133mhz fsb that I was able to boot at 400mhz with a 3x multiplier. What a fun time it was to get extra speed for around $40 per cpu at the time, too bad those days are long gone.
I'd also be interested to see if a k6-3 or k6-3+ would boot on this board with the new regulator.
Unfortunately I neither do I have k6-3 nor k6-3+ at hand.
@@necro_ware Yeah I thought about that after I wrote it that they were kind of rare. I was writing that at 3 in the morning when I couldn't sleep. Looks like they are going for a small fortune on ebay too.
Ridiculous and awesome. Thank you for sharing this with us
Great video & awesome making the vrm 😊L. Thanks Necroware
I had a similar mainboard to what your using .
However a 200mhz intel MMX was the chosen CPU I used .
At which the Main Board could run a 200Mhz CPU natively with jumper configs on the board according to the provided jumper maps stamped on the board.
I had 2 different styles of Intel CPU's A ceramic like you have in these videos & A Black colored Intel CPU both with MMX . Also had a non MMX CPU clocked in at 200 MHZ.
Here's the weird thing.
If the Jumpers config was set per CPU to run at 200 MHZ .
Each of the CPU's were overclocked to 233 MHZ . This overclocking was not only shown on the POST screen but in software.
With messing around with the jumper configs I got it to run a 200 MHZ cpu at 266Mhz but it wasnt stable.
However the sweet spot for stability was 233 MHZ.
This was also an ASUS Board , I no long have this machine and dont remember what model the board was.
I remember the time that I a AMD K6-2 450Mhz or 500Mhz processor into a Socket 7 motherboard (a FIC PT-2200) with a custom bios that allowed larger hard drives up to 128Gb that had a linear regulator based power supply and set the jumpers for the voltage as low as I could at 2.5V and set the multiplier jumper to 2x /6x position and had the FSB at 66Mhz and the system worked but the regulator got to 150C!!! and I am surprised that it did not explode!
Yes, eventually you got an overvolted CPU and overheated regulator.... Sound like a start of an exciting adventure :D
Brilliant! :D And you could also mod the BIOS to include the updated CPUID code! :)
Most of the heat from those voltage regulator ICs is dissipated through the ground/belly pad. If you add more vias you could put the heatsink on the other side of the board and It will get airflow from the CPU fan. If you wanted to go one step further you could solder copper heat sinks directly to the board for the best thermal transfer.
i know you guys grew up with this stuff but the fact you dropped an AMD cpu into the same socket as the pentium is mind blowing . Amd and intel sockets are so different from one another these days.
Did you ever get back to this? It'd be amazing to see this motherboard running so far beyond its spec in a stable fashion.
I love the K6-2. That was my first PC and I over clocked it to the moon (600MHz) I miss that machine
Aw yes part 2!
Have almost the same motherboard, little bit newer revision. Better support for fsb and cpu voltages. Little bit differently modded. no cache slot and newer cache chips. Modified bios to support bigger hdd's. Running a k62-500. Still needs the dallas mod, now its just a lots of tape to hold bios settings. Fun mobo! :)
Ok, that was really kewl. I am just completely floored by this, great Job! God Bless my friend. I am a new Sub at this point!
Excellent video!!!. Waiting for version 0.2 here :-)
Cool experiment, keep them coming!
I am not an electronics engineer myself too, but I would be concerned about the 5V rails for the beefier VRMs as 15A on 2.2V is 33W and if I do not take account the losses it is a 6.5A load on the 5V rail, and maybe it worth to check, if the motherboard is able to handle this load on the 5V rails
Maybe it's worthwhile to bring a + and -5V straight from the PSU to the VRM to spare the traces.
@@schlbus Or just re-solder the traces to add more flesh to them
So cool! This is absolutely brilliant work and as someone trying to really learn more about these things, your work is invaluable. I loved this one. Whenever the final product is ready, I can't wait to try it out on an AOpen AP5C/P that has great sentimental value for me. 430FX is a great chipset too, so there's a lot to be had with this upgrade!
Thanks for all your brilliant work, your contributions, and for another awesome video!
As usual very interesting topic for me ! On the subject of VRMs, I've already tried connecting a cheap buck converter based on LM2695S (single IC - ready-made boards intended for DYI constructions) to Shuttle HOT553 with P233MMX. In the short run it works, but of course 3A is not enough for 233MMX and I was thinking of using those with XL4016 onboard (with big heatsinks for XL and diodes) that promise 8A current. But now I am looking forward to your 15A project ! Thank you ! Keep it up !
I can't see doing anything like this to my only pre AGP Socket 7 board. I have 3 Super 7 boards that are fast enough. Two are FiC VA-503+ Rev 1.2. Those boards work nice aside from not playing nice with some ISA/ PCI network and sound cards at the same time. If the video drivers are to new then that will also cause AGP card problems and may cause the network to not work, or the audio to half work. I found version 4 era Nvidia drivers to play nice. We all know some older boards can have quirks to get working. The other one is a Jetway with Alladin chipset with XP on it that just works. I find it fun to compare DVD playback using graphics cards ranging from early CGA, EGA, and VGA technologies. Those include a Diamond Viper V330(Riva128), Diamond Speed star 64 4mb, Cirrus Logic, S3 variants, Ati Rage pro, Ati Radeon R7000/9600pro and from a Nvidia Vanta Lt up to a Geforce 2 MX400. I can verify that most video cards were held back by the CPU speed back in that era. Even the 2D cards did a good job with a faster CPU. Not so long ago Diamond/ATI used to be partners with Nvidia.
Wow! You have my respect. If I need once a tip how to come from one topic to another in 2 seconds, I'll contact you :D
In the late 90's, the Evergreen Spectrum processor upgrades for socket 5/7 machines bundled an AMD K62 333 mhz CPU with a voltage regular, a fan and a CD full of patched BIOSes for various motherboards of the time to allow similar upgrades without modifying the board hardware itself. These are relatively rare and expensive (think $400 ish) so this is a great alternative approach for sure!
Pretty cool project!
Make a "left hand" and a "right hand" VRM design so that the components can be aimed in the best direction for air flow to cool them. With the components on the opposite side of your 0.1 revision the airflow from the CPU cooler would help. Back in the 486 and Socket 7 times I filed a lot of CPU cooler bottoms flat to get the best contact so very little compound was needed. Clamp a flat file in a bench vise and slide the cooler across it, rotating it to different angles to even out the metal removal. Most of those extruded aluminum heat sinks were far from flat on the CPU contact surface. An easier method would be to use a milling machine with the spindle perfectly perpendicular to the table, and take very light and slow passes with a large end mill until the surface is flat.
Another thing I did with many heat sinks was glass bead blasting to remove the anodizing. The roughened surface should have a bit larger surface area, better for conducting heat to the air. It would be interesting to do a series of thermal performance tests on a Socket 7 cooler. 1. stock out of the box. 2. file the bottom flat. 3. glass bead blast all of the heatsink except the bottom.
Another comparison could be different colors of the same heat sink. Ideally, black ought to work the best since dark surfaces are typically better at radiating heat, but in these applications only the outside surfaces around the edges have a free path for radiant energy. Inside the heat sink, any energy radiating from one fin goes directly into the next fin.
This is interesting! I'm early 80s kid, so I started playing with 80186 machines way back and those intel sx33 and intel dx66 times were really great! I had my best time playing Doom 1/2 with zero modem cables. I would love to do that again, but i'm not willing to see the pain collecting those machines. Luckily there is a youtube channel to memorize the good old times that might not be that great after collecting all that hardware again.
Well in the end it's just a hobby :) There is no need, that everybody, who was hit by a nostalgia have to start to tinker on his/her own, sometimes a channel like mine is sufficient to be taken back for a while. I'm glad you like it and have a small remark, you probably didn't play with the 80186, that is a microcontroller, which wasn't used in a PC, but in some powerful automation systems and things like that. For the PC there was 8086, 8088 (also known as XT), 286 (also known as AT) and 386, 486, Pentium etc...
@@necro_ware I know that i had nokia mikro mikko, and it had massive motherboard. Nokia mikromikko II was a 80186 machine but i might have had the version three. In any case it had only sx version and all the chips were easily changeable dip or zif sockets. Also i have HP LX200 pc that is at least 80186 compatible. I might be wrong about the arch, because it says compatible..
In any case, it was interesting time, that there wasn't bios prompt. but bios config was done with jumpers.
If you can flip the IC’s to the cpu side you’d get better cooling on the ICs from the heatsink fan.
Nice project, nice video. thanx