Overclocking IBM PS/1 486. Cooling and Power experiments (Fake buck converters!)

[@Rafal-G86]

For a while I had a dream of Christopher Walken making vintage computer videos. Apparently dreams do come true.

[@lelandclayton5462]

I work with a lot of voltage regulators and the best one that comes to mind is the D30V30F3 from Pololu. 3.3V out with 3.7A maximum current and uses input voltage 3.3V - 45V. I used the 5V and 12V variants for my home brew S100 computer.

[@Epictronics1]

Thanks. I think we need 3.5-4v for this project

[@lelandclayton5462]

@@Epictronics1 In that case a variable voltage regulator. They have some that are also variable. The D30V33MAL can do 1.4V to 7V with 3.8A max.

[@Epictronics1]

@@lelandclayton5462 I'll take a look, thanks!

[@g4z-kb7ct]

If you are putting a loose cooler on top of a CPU, instead of thermal paste use a thermal heat pad. These are available in sheets any size you want. I have some 0.5mm, 1mm, 2mm and 3mm in 20cm x 10cm sheet which can be cut to any size you want. These squash down and will make better contact with any hot chip. You should also put thermal pads on anything that gets hot including modern motherboard chipsets, NVME drives and modern TV's as they are made to fail and don't have enough thermal heatsinking to last any more than the warranty period. I have many old large LCD TV's here that I fixed and added thermal pads to all of them and they are still working fine after being used for many years.

[@SockyNoob]

Agreed

[@m4rgin4l]

I love the design language of PS/1s

[@MoonBeamie318]

Another great video! It was fun chatting with all of you. Can't wait til next week! ❤

[@Epictronics1]

Thanks!

[@sinsaru]

Nothing compares to relaxing after a long day with a new Epictronics video. What can I say? Some people relax with boring tv shows. I watch 30+ year old computers get overclocked 😂

[@atrdigital5935]

Same. It’s just something about old technology that makes me happy.

[@Epictronics1]

You're in the right place :)

[@MikesRigs]

In the 90's we cooled the dx4/100 OD by sticking a 40mm fan on the top with some double sided tape, ran very cool but could never get it to hold 120 reliably... Im inspired to revisit this now!

[@jasmijndekkers]

Great job you did with overclocking that IBM PS/1 486. Greetings from Steven from the Netherlands

[@Epictronics1]

Thanks :)

[@magicsmokeescapist4069]

As someone wrote aboute Downhill mountainbiking. If everything seams under control, you're not going fast enough. 😂

[@bollux78]

It fits, it boots, it's epictronics! Great video, I think the 486 era was the funniest and more elastic to overclock!

[@Epictronics1]

Thanks :)

[@lm4278]

Dude. You really deserve way more than 16k subs. Great content. Love your stuff.

[@Epictronics1]

Thank you :)

[@psilimit]

"And we'll just put a tower cooler on top..." I'd subscribe, but you already got me on the other video where you're planning on speeding up that 286 with the 486 upgrade ; )

[@DStageGarage]

The LM350 is a nice 3A adjustable linear regulator very mych like a LM317 on steroids. The only problem is as far as I know it only comes in TO-220 and TO-3. TO-220 can be bodged for SMD mount though. As for regulating that cheap one you could try desoldering the potentiometer and putting it outside feeding the kynar wires between the pins of the socket (under the plastic part I mean).

[@Epictronics1]

It's tight between the legs of the pin headers, but I'll check to see if it's possible.

[@g4z-kb7ct]

24:00 With 12V input and only 3.3V output you are supposed to heatsink the reg if you require 3A from it. Also if you connect a huge cap to the input and output like for example a 2200uF on input AND output it'll probably work and have stable output if you put a heatsink on the chip too. Even the little SOIC8 chip can drive 3A with a heatsink but it's stated in the datasheet it must be heatsinked for 3A otherwise the thermal protection will kick in and it'll shut down.

[@Epictronics1]

We need to find a chip that doesn't require a heatsink. Otherwise, it won't fit inside the interposer.

[@g4z-kb7ct]

@@Epictronics1 So just put it on the outside. A good linear reg is LM1084-33. See the datasheet. It's old but easy to get. This is 3A and might work without a heatsink. I have some of them in an old original Commodore 64 PSU (with legs twisted) and they work fine without a heatsink but of course the C64 only draws about 800mA. The pinout is like LM317 so it's not plug and play for a LM78xx series and legs need to be twisted but the LM1084 is a more modern device that runs cooler. You probably need to find out exactly how much current the CPU draws. just design something so the reg is on the outside as putting it on the board is too limiting.

[@Epictronics1]

@@g4z-kb7ct That's a good plan B. However, the original interposer from the 90s doesn't have any known issues with heat. So, I'm gonna try to find something that fits inside first.

[@g4z-kb7ct]

@@Epictronics1 If you look I'm sure there's another type of 5A linear reg available in the same type of package. Would only need to re-design the board to take the different part. If you want to remove all the original interposer parts and scan it both sides I can make a 1:1 copy using a modern 5A reg.

[@Epictronics1]

@@g4z-kb7ct Thanks, I'll see what I can find and do some tests

[@josteinkallevik]

What I did on my socketblaster was just not intstaling the vreg on the board, but i used an external regulator. Then I can adjust the vcore without removing the cpu everytime. I used the same cheap one you have there in the end (the blue board). And the 5x86 is stable at 160Mhz 3.45V. I need to buy me a load tester just like you have there, that looks usefull to have on many projects:) Thank you for the video, perfect relaxing sunday morning watching this.

[@Epictronics1]

Thanks. I'm gonna do the next test with external power too. Ideally, however, we will see well-working internal VRMs in this project eventually

[@josteinkallevik]

@@Epictronics1 Looking forward to your next update on this :)

[@Epictronics1]

@@josteinkallevik Thanks

[@psilimit]

A link to get a 486 Socket Blaster? You're my hero!! I bought you a Coffie (never used that service before). Also my first time actually using PCB Way. I'm excited for...whatever is coming in the mail. A kit, a part, a fully assembled unit? No matter. It's all better that being completely unfindable other than on Wikipedia. I'm excited to add this to my retro rig!

[@Epictronics1]

Thanks :)

[@BigBadBench]

That was a fun experiment! I’ve always wanted to do some specific testing on those cheap converters. Scary stuff!

[@Epictronics1]

I was expecting the switching frequency to be off and minor noise at 3A, but apparently, it's way worse than that!

[@altbeetle]

16:50 nice multimedia case, I really wanted something like this once. I do not understand why such cases have not become popular, because it is much more convenient than additional speakers on the table

[@Epictronics1]

Yeah, I need to find one of those. That case is as quirky as the mobo :)

[@mikejones-vd3fg]

Wow! *wipes tear from eye* i may have told this story but i had this IBM as a preteen and was in love with doom2, i actually didnt notice the bad fps until i saw it on my friend pentium a few years later, I just shrunk the screen down and I was fine. But it was definately lacking, especially for games like C&C and duke which were near unplayable. So i looked up in the local classified and found a DX50mhz chip, great ill just swap it, it even has an upgrade slot! Little did iknow it only accepted upgrade processors. With the help of my dad i bought it anyway and proceeded to bend the pins trying to install it, it just wouldnt fit. Then i read the manual and realized my mistake. To see this computer outperform what i could ever imagine, well done! Bravo!! *wipes tear from eye* Bravo!

[@Epictronics1]

Thanks! I had a similar situation back then. I had the "un-upgradable" IBM PS/2 Model 50. It feels really great every time I see this underdog score a few more points :) It already outperforms many of my 486 systems and we're not done yet :)

[@thecorruptedbit5585]

I have one of these with a proper 486 socket. It's a nice little system, especially with a DX2-66 - now I can see just how well it would fly with the overdrive in it!

[@Epictronics1]

Do you have the board with two sockets and onboard ET4k graphics?

[@thecorruptedbit5585]

@Epictronics1 I do not. Single socket, and I believe the graphics are a Cirrus chip (I can't find a straight answer about PS1 2168s). Same beautiful case though.

[@Epictronics1]

@@thecorruptedbit5585 Oh, it's a 2168. I've got one of those too. My 2168 has an identical board to the one we used in this video

[@avidrandomer]

If you want a good little regulator board to try, go with the ones sold as "mini560". They use joulewatt high efficiency regulators, with high switching speed. They are good for around 30W.

[@Epictronics1]

I'll check it out. Thanks

[@rasz]

@@Epictronics1 uses 8A JW5068A, only $1 for the chip but modules are $1.5 so no point getting chips :)

[@g4z-kb7ct]

Your fluctuating power issue is easily solved. Don't use a buckreg as they are a switching device and will change based on load. What you need is a simple linear regulator like AMS1117-3.3. This can provide 1A. If the CPU uses less than 1A this one is fine. It it draws more than 1A then you need to find another suitable linear regulator.

[@Epictronics1]

I don't know for sure how much current it needs, but someone claimed it's 3.1A. The original interposer from the 90s has a 5A part

[@g4z-kb7ct]

@@Epictronics1 Holy cow 3.1A??? Wow! Well the only way you will do it with a SMPS is by having a lot of backup capacitance and that adapter has no room. Or maybe it can use lots of tantalum caps on both sides and make it just slightly larger with caps around the edges too. I think in most motherboards there is room around the CPU to make the board larger. Maybe I can re-design it. We can talk in email if you want to discuss it and take some motherboard measurements etc....

[@Epictronics1]

@@g4z-kb7ct Thanks, I appreciate the help. I'll test out a few parts first and find a suitable chip that will keep a stable 4v and stay cool enough. I'm thinking four tantalums inside the board, just like the vintage interposer.

[@g4z-kb7ct]

@@Epictronics1 could maybe add another interposer at the bottom with the power supply and tantalums etc. There's plenty of vertical room in most PC's

[@Epictronics1]

@@g4z-kb7ct Scrapcomputing just recently made a new version of this interposer about twice the hight. But he's having some issues with it. Two much resistant with the added connectors? picking up interference?

[@maxtornogood]

A lot of crashing in this clip, did you overclock the camera as well? 😜

[@Epictronics1]

I wish! My camera is 12 years old, It could need some overclocking :)

[@GraveUypo]

i think a tiny fan would do more to cool those cpus than that silly giant tower that isn't actually getting any of the heat from the small heatsink

[@Epictronics1]

I think you're right

[@bkims]

Getting a fan on those ODPRs would do way more for the cooling than attaching a heatsink to the poorly thermally interfaced heatsink cover on the stock heatsink XD. Also the thermal interface from the logic gates to the heatsink is probably so terrible that ~40-50c heatsink temps is probably MUCH lower than the gate temps. You should attach a peltier cooler to one of these one day for proper extreme retro overclocking :D.

[@Epictronics1]

I have actually tried to find a suitable peltier for this project but I haven't had much luck figuring out which one to get :)

[@Nukle0n]

@@Epictronics1 Peltiers are a bad idea, they create condensation, unless you wanna do it to learn the same thing people did back in the day :)

[@Epictronics1]

@@Nukle0n As a permanent solution, they probably suck, but for tests like this, they are probably great

[@vhfgamer]

@@Epictronics1 Couldn't you just scavenge one from one of those cheap little gimmick refrigerators they sell at the hardware store?

[@stonent]

@@vhfgamer The ones that actually use more power than a real fridge?

[@markg3506]

Nice contents, I love 486 CPUs

[@Epictronics1]

Thanks

[@Shand1982]

I see PS/1 in tower form, I click like. :)

[@Agoz8375]

This is fun 😊 thank you for that video

[@Epictronics1]

Thanks :)

[@ZiggyAndTheSpiderFromMars]

I recall getting the 300mhz to 360mhz Celerons to 500-550mhz consistently. That was my all time greatest overclock. Probably everyone's.

[@Epictronics1]

I actually made a video about those Celeron 300As! I've got two of them in the ABIT BP6 build. Awesome CPUs!

[@kasimirdenhertog3516]

To disable the 486SX, isn’t there an extra pin in the empty socket that does that? Instead of using the wire. In another video they installed a 487 copro in the empty socket and demonstrated it disables the other CPU, because the 487 is in fact a fully-fledged CPU (486DX).

[@Epictronics1]

That is correct. However, with the bodge wire, you can use any 486-class CPU

[@kasimirdenhertog3516]

@@Epictronics1 I see, thanks!

[@Mr_Meowingtons]

Add a resistor to the VCC pin and ground to add a load to ajust the voltage.

[@JohnSmith-iu8cj]

The problem with ceramic instead of electrolytic capacitors is that they have a lower ESR. That creates ringing at the resonance frequency. The solution is to use a small resistance somewhere around 10 ohms in series to increase the dampening.

[@Epictronics1]

Could that be the issue we saw with the small module?

[@rasz]

@@Epictronics1 small module definitely had insufficient filtering if the noise was >50-100mV

[@Epictronics1]

@@rasz I didn't bother checking noise and ripple because it was so obviously off :0

[@AnnaVannieuwenhuyse]

​@@Epictronics1Please do! Meters often don't accurately reflect the voltage when ripple is present.

[@enilenis]

I have one 486 I experiment on a lot. M601 motherboard, that is 1st gen, and doesn't support DX4 chips, but strangely, it not only runs fine on a 133MHz DX5, but overlocks it to 160MHz without issues, at which point it approaches P75 levels of performance (minus the bus speed). I migrated right pass a 486 in the 90's. None of my friends had one. I only assembled them as a modern vintage gamer, and was underwhelmed somewhat. They're too slow for 3D and too fast for early 90's stuff. Kind of stuck between generations that really mattered, like the 386 and Pentium.

[@helmargesel3972]

Thanks for sharing your knowledge

[@Epictronics1]

It's a pleasure :)

[@brotherdg1]

My 5x86 160 has been running on 5v for a very long time...

[@AnonyDave]

It's weird seeing that period correct one these days. Being able to use an LDO linear regulator for a desktop cpu is definitely not something that has been possible for a long time. But also I'd have to assume it was getting a little edge case even back in the days of the pentium.

[@Epictronics1]

I imagine it gets really toasty, squeezed in between the PCB and the CPU. But the Trinity seems to work fine despite the heat

[@axi0matic]

Perhaps the bad floppy drive wasn't detecting the disk? Worth checking the relevant switch / sensor?

[@Epictronics1]

It's quite possible. I'll check when we need it.

[@esc2dos]

In Scrapcomputing's last video about the redesigned " SocketBlaster", he suggested the issues he was having with stability were perhaps due to it's increased height, maybe you will solve his problem with a better VR. Looking forward to your follow up on this.

[@Epictronics1]

Yeah, I watched his video. Great experiment with the new module

[@tomekrv942]

I have FIC 486-VIP-IO2. This motherboard needs special VRM board which I don't have so I used one of these converters on a blue board (LM2596) and my 5x86 is working on 3x50MHz. It can complete all benchmarks but sometimes hangs up during normal work.

[@Epictronics1]

Cool. We are going to push this board to 50MHz too, eventually.

[@tomekrv942]

@@Epictronics1 it was always interesting for me which settings are better 4x40 or 3x50MHz? I am working far from home so I don't have much time to test it so I just set 3x50MHz because in my opinion it should be better but I am not sure. My motherboard have VLB and PCI slots. FSB 40MHz was the highest to work with VLB graphic cards. For higher FSB I had to use PCI card.

[@Epictronics1]

@@tomekrv942 I would guess, 150 is going to give us more FPS than 160. We'll find out

[@UpLateGeek]

The regulator on the old interposer module (not to mention the ODP CPUs) is a linear regulator, vs the switching regulator on the new one. The linear regulator gets away with a lot less capacitance because it can just pull as much current as needed (up to its limit), and only needs the capacitance for stability, whereas the switching regulator is literally switching on and off, so it needs more capacitance to maintain the voltage. As you saw, the clones of those chips often introduce a lot more switching noise, and possibly have other issues like thermal drift or switching frequency issues (not to mention lower max switching current). They could also be using the wrong value or current rated inductor, and really the value of inductor required is going to change based on the input and output voltage, and the switching frequency - again, this could be wrong in a counterfeit chip. Anyway, if you really want a stable power supply, you really need to get one that's designed and built for the specific task. Unfortunately it's not really as easy as following the example in the datasheet, hitting autoroute in your EDA software and pressing print on your favourite PCB fab, there are a few catches that the manufacturers of these cheap modules often get caught by. But this post is already long enough, so I won't go into all those details. However, there are plenty of linear regulators that might be suitable for this which would be much easier to implement. generally these only require a couple of resistors and capacitors, throw in a trimpot for varying the voltage and you're done. You just need to find an LDO model with a dropout current of 1V max at the maximum current for your CPU, which I guess is around 3A based on the one in the old interposer module. I did a quick search and found the LD29300 from ST, or the LP38501/LP38503 from TI. The amount of additional heat these generate may be an issue, but we'll burn that bridge when we get there.

[@Epictronics1]

Thanks. The original interposer seems to work well with a Linear regulator despite it possibly getting quite hot inside the interposer. However, it only runs the 5x86 @133. We are going for 160, possibly even more! so, I'm just gonna have to test both and see how they perform

[@clintthompson4100]

Great video and best of luck over clocking to the 160MHZ. Shame you have to deal with all the fake garbage chips and voltage regulators. Amazing that people will purposely screw other people over to make a buck but than again dishonesty is a part of human nature(most from my personal experience)and that is just sad money and greed is more important.

[@Epictronics1]

Thanks. We'll try out some genuine parts from Mouser next

[@clintthompson4100]

​@@Epictronics1 That is awesome

[@Grinder2112]

Silly, /but/ fun?! You've well shown silly *is* fun.

[@GGigabiteM]

I have an IBM PS/1 model 2133-C11. Outward appearances are virtually identical to yours, internally it has a different logic board. Mine *had* a 386SX-25 soldered on the logic board, I desoldered and replaced it with a TI486SXLC2-50. Getting the cache and clock doubling to work was a bit of a challenge, but I got it working. I overclocked the logic board by swapping out the 50 MHz oscillator responsible for the CPU to a 66 MHz oscillator. This made the machine hideously unstable, until a BIOS corruption issue lead me to trying a faster BIOS chip (70ns vs 120ns) and most of the stability issues went away. I still can't run Speedsys without it locking up the machine, but other benchmarks run fine. Two current issues I have to yet sort out is that: 1) it uses proprietary "16 bit" 72 pin SIMMs. IBM made a special 72 pin SIMM that is split into two 16 bit buses because of the 16 bit bus used on the 386SX. Using a regular 72 pin SIMM causes the machine to throw a ton of critical BIOS errors "system board failure" "memory failure" "configuration error". The board thankfully came with two of them (one 4 and one 2 MB module) and with the onboard 2 MB, I have just under 8 MB available with the system stealing a bit of memory for the system board. I need more RAM to be able to run some later DOS games or Windows 95, so I need to either find some of these unobtanium modules, or figure out how to reverse engineer and build some new modules. 2) The system won't load Option ROMs, which precludes the use of most ISA cards. There's no options in the very basic BIOS setup menu about Option ROMs, so I have no idea what to do about it. I really need the ISA slots to work because the onboard IDE controller is so old that it doesn't support LBA or even ATAPI. This means hard drives only, and really old ones at that. Thankfully I still have a couple of ancient 80-280 MB hard drives that still work to boot the system, but not having a CD-ROM sucks.

[@Epictronics1]

Awesome project :) Good luck, I hope you solve it

[@nickwallette6201]

Wait ... I thought ATAPI support was entirely within the protocol. I guess I've never actually tried a pre-ATAPI controller with an ATAPI device, but I had always assumed it would work. Albeit without other modern improvements, like DMA and a 32-bit (VLB/PCI/EISA) bus interface.

[@GGigabiteM]

@@nickwallette6201 ATAPI wasn't adopted as part of the ATA specification until 1998 with ATA-4. This is why there were so many different competing CD-ROM interfaces in the mid 90s, and why some sound cards had 3 or 4 different CD-ROM connectors on them. My machine was made in 1992, and it predates even the first ATA standard released in 1994. It is purely dumb IDE. It also suffers from "IBM is stupid" syndrome. The BIOS is so broken that ISA cards are basically unusable. I did a bit more research and discovered that almost all of the IRQs and common address space where option ROMs are supposed to load are all reserved by the IBM BIOS and can't be used for anything. Really regretting spending money on that TI CPU right about now. Unless someone can disassemble the BIOS and fix it.

[@nickwallette6201]

@@GGigabiteM No, no, I get all that. But I haven't looked at "how ATAPI works" enough, I guess. I always thought it was purely the act of sending packetized SCSI commands to the device via the ATA interface. Given that IDE is really just ISA-Extended-By-Cable, with a very narrow address bus, it didn't seem like there would be anything to prevent trying to talk to a CD-ROM on pre-ATA IDE controllers. Like, what does the IDE interface know is even the difference?

[@GGigabiteM]

​@@nickwallette6201 Early IDE being basically a buffered connection to the ISA bus is the problem. Since it is entirely driven by software (in this case BIOS INT13h disk routines), the BIOS routines being broken is the reason it doesn't work. I have a suspicion that IBM didn't put the disk controller in a standard location, because utilities that are used to locate the IDE controller don't find any disks, even though one is connected.

[@Nukle0n]

I probably don't have to tell you that sticking a big cooler like that on top of the existing heat-sink is only gonna make extremely minor contact. It would be interesting to see if it actually made a difference with the original heatsink pried off and replaced with that big silly modern one but I also kinda doubt it, it seems more like a limitation of the silicon than any kinda heating problem.

[@Epictronics1]

Yes, of course. But I don't want to ruin the ODPR :)

[@Nukle0n]

@@Epictronics1 yeah It would be a destructive process, at the very least it won't go back on in the same way

[@fra4455]

Great video

[@Epictronics1]

Thanks

[@fra4455]

@Epictronics1 thanks

[@HorstGBurkhardt]

20:23 I've actually looked at designing a decent voltage supply for the three-pin header on the 486SocketBlaster, due to being displeased with the available options. It's important to note here that not only is the Linfinity part itself obsolete, but it's using a linear regulator, which wastes more energy as heat. Running 5V -> 3.5V is actually right at the limits of what it needs for voltage regulation, so it's not super inefficient for the interposer application. We generally like to use switching regulators now, because they're highly compact and waste less energy as heat, but they have to be more complex to provide the same stability, and if they're not designed well, they'll cause problems with your CPU anyway. I think I accidentally tweeted at you thinking you were BuB some time back now about my design, at that time you hadn't done anything on 486SocketBlaster. But to design an interposer using linear voltage regulation, in the same way as the older interposer (at least for power supply) would be a trivial task and might be an improvement on cheap buck convertors. Or, there's my switching design based on LM2596S, which I was also looking at potentially using to replace the hot-running stepdown on Voodoo3 cards. I did some research suggesting that the worst case would have the CPU drawing 3.5A, which also tells you how much current handling capacity you need. The specifications do suggest that you might draw closer to 8A, but I had a lot of difficulty verifying that claim.

[@Epictronics1]

Thanks. We might actually need as much as 4v to run the 5x86 over 133MHz. I will be testing out a few components quite soon in a vid. I'll test for stability, noise/ripple, and heat.

[@vhfgamer]

Have you thought about investigating that strange SIP socket for the video memory? I'd love to know what chip is supposed to go there. I recall from a previous episode that you tried populating it, but nothing happened.

[@Epictronics1]

Yes, I probably had the wrong chip for these boards. Since it was so easy to upgrade with a chip on the mobo instead, I didn't continue to search for the right SIP chip

[@SidneyCritic]

The originals are linear, and buck is hard to clean-up output, ie, what's the AC like. so maybe a big/strong linear with a lot of GND plane/heatsink.

[@briangoldberg4439]

Linear Technologies is now owned by Analog Devices. They absorbed all of their designs, so they might have a more recent IC that is an equivalent. I would contact AD and see what they have to say.

[@Epictronics1]

Thanks, I'll check

[@datPinto]

I miss my old PS/1 Consultant... very similar specs.

[@Epictronics1]

I believe they are the same machine. It was just marketed as the consultant in some markets.

[@datPinto]

@@Epictronics1 yessir. Just a name difference. I remember upgrading the ram to 8 Megs and even adding a Creative multimedia kit, with a SB16 and double speed cd rom! Those were the days. Oh, and a new modem too... the jump from 2400 to 14400 was amazing! Ah, memories! If I could find one in decent enough shape, I think I'd pick it up, just for nostalgia sake. In the meantime, I have so many "retro" pentiums.

[@ScottyBrockway]

Fake chips are such a bane these days, and also a fire hazard.

[@Epictronics1]

You got a point there. Never considered the fire hazard

[@kokodin5895]

i wonder if necroware socket 7 vrm can be hacked for higher voltages currently it is from 1,6 to 3,5v jumper adjustable board you stick into special slot in some socket 7 boards, but it most likely worked as stand alone board , just give it power and pick up the output

[@Epictronics1]

3.5v might work

[@pocok95]

Previously I had very good results with DC-DC converters manufactured by Recom. They produce configurable output voltage types, but the ones rated for higher current can be quite pricy. But maybe worth a glance?

[@Epictronics1]

I'll check, thanks

[@g4z-kb7ct]

@@Epictronics1 I have some here, they are switchers, no good for your use-case.

[@charlesdorval394]

Even the cooler didn't want anything to do with that lol

[@Epictronics1]

lol

[@erickvond6825]

As fiddly as it is to repair a floppy drive of that era I'd really like to know what went wrong with it. Figuring that out might be seriously useful to others trying to diagnose the same fault.

[@Epictronics1]

Since it doesn't seek at all, finding the fault shouldn't be too hard. In one of the P70 videos, I very recently repaired an FDD that wouldn't spin up.

[@UFO_researcher]

Mini 360, all day son.

[@rasz]

@24:30 its the small diode that gets hot not the LM. LM is most likely genuine, fake wouldnt do 3A.

[@Lady_Zenith]

There are some AMD 5x86 chips that can tolerate 5 volts. Most of them do not even post at such voltage, but at the same time, they do not die. Some revision do post and run fine. Tho yes, they do heat a lot after that. I ran one like that at 160 for many years, on a board where the mosfet died and I just shorted the 5V input to output and ran 5V directly to the CPU and it ran and worked fine without any adapters. The only issues on this board without any adapters would be making the L1 cache run in write-back mode, cause if it runs as write-trough you give up some of the performance.

[@Epictronics1]

Haha, that's an awesome fix :)

[@Lady_Zenith]

@@Epictronics1 You can do the same with the DX4 overdrive you have btw, just short the mosfet on it. Iv not seen intel DX4 that could not handle 5V, but they need a lot of cooling afterwards. If you need something inbetween, cut/desolder the output leg of that mosfet and just use resistor to get the desired voltage at around 4V. I had no DX4 that could not do 120Mhz with a little bit of trying.

[@Epictronics1]

@@Lady_Zenith Awesome

[@brotherdg1]

I have 3 amd 5x86 processors - they all work fine on 5V with active cooling. Intel dx100 - does not work.

[@idahofur]

Ah yes, We had the pc market. We lost the pc market. Lets introduce the ps/1. LOL

[@Epictronics1]

The funny thing about these PS/1s is that they are clones of the clones.

[@idahofur]

@@Epictronics1 Yes, oh yes. I forgot about that saying.

[@jeremiefaucher-goulet3365]

Soon on the channel, the world's first water-cooled 486 😂

[@Epictronics1]

Thanks. That's not a bad idea :)

[@SianaGearz]

You could measure the off-resistance of the CPU outside the socket across the voltage rail - i think it should go lower in operation but not higher. Then you can build a substitution resistor and jam it into the socket as a test load and use that to tune the voltage regulator without it going all wonky and without occluding the middle of the adapter. If it's still wonky well that's not a good sign and you should peruse the datasheet to stabilise it, maybe it wants some extra capacitance either side or whatnot.

[@Epictronics1]

That is quite possible. I haven't checked the datasheet for the MP1584. Maybe, the caps are too small on these boards. That might explain the saw tough wave form instead of DC lol. At the moment I'm checking the LM1084. It seems to have the perfect specs for a 5x86

[@SianaGearz]

@@Epictronics1 Ehhh that'd be a heater under the CPU... Though if that CPU only draws let's say 5W, then losing 3W on a regulator is completely adequate and not really a problem.

[@SianaGearz]

If you suggest the CPU needs 3A at 3.45V that means the CPU burns 10W and from 5V the regulator would be burning 5W, yeah you can live with that i guess...

[@Epictronics1]

@@SianaGearz It's more likely 4v @160. We'll know for sure when I have received a genuine LM1084 from Mouser. Also, 3A is a theoretical max. We may not need to go that high. More tests need to be done

[@Knaeckebrotsaege]

wouldn't it be possible to have an "off board" voltage regulator? ie have the traces responsible for the power supply come out of the side of the interposer on a PCB extension, so you're not limited to the miniscule amount of space underneath the CPU, and also won't have issues with keeping the regulator cool. Might look a bit weird to have a "wart" coming out of the side of the interposer, but it would probably solve a bunch of problems at once edit: actually, if it's not stuck underneath the CPU, you could also feed the regulator power externally from a molex or berg connector or something, not relying on the power coming in through the board. Heck you could probably feed it from 12V from the PSU to make things a bit easier than trying to step down from 5V

[@Epictronics1]

Yes, that would be an easy solution. However, the voltage regulator seems to work fine in the original interposer, so I hope to find a modern alternative

[@tlrptg]

LM1084. needs 2 small caps and 2 rezistors. 5 amps and rock stable

[@Epictronics1]

I was reading the datasheet for the LM1084 when I saw your comment! It does indeed look perfect for a 5x86

[@awd42]

That's a linear regulator, so it would need a heat sink. Switching regulators like a (genuine) LM2576 (which is still currently made) are better. Interestingly, TI's product page for that links to a couple of their modern highly-integrated small form factor (e.g. 6 x 4 mm) buck converter modules, like the TLVM13630. Those look promising...

[@Epictronics1]

@@awd42 You're right. I'll have a look at the LM2576. Thanks

[@JohnSmith-iu8cj]

Most of these sps are around 3A. I don’t know but I guess you can use them in parallel. Isn’t that how modern vrms work? They have many „phases“ in parallel I believe.

[@Epictronics1]

I think 3A should be enough, but I have found a better 5A chip that I want to try next

[@AsbestosMuffins]

wasn't the bad cap era like the mid 1990s through 2000?

[@Epictronics1]

There were two bad eras. the early electronlytic SMD caps and the "stolen formula" era

[@g4z-kb7ct]

The regs on the cheap buck boards are unlikely to be fake.... no one is going to fake a part that costs just a few cents. The complete boards are mass-produced in the hundreds of thousands and would likely only cost 50c to make. I have some of the same bought for $2 and I designed it into a Commodore 64 PSU. I simply bought the boards with the parts since it's cheaper to do that then I removed the parts and fitted onto my PSU board which also has a crap-load of solid capacitors on it. The LM2596 can provide 3A very easily. But when using any switching regulator you must provide a lot of capacitance on the input and output (1000uF should be enough on both sides) and it will work like magic. But as stated in another comment you should be using a linear reg. You may want to consider putting the linear reg separately somewhere else, maybe strap it to some other part that has a heatsink or mount it to the PCB somewhere using a 2mm thick thermal pad and some epoxy glue to hold everything in place. Or you could redesign the board so the reg and adjustment pot hangs off the side on a separate heatsink.

[@broderperdurabo]

' Same thing om my Toshiba T3200 floppy. 3 small 10 Micro Smounted caps.

[@james5583]

Härlig kylfläns 😄

[@Epictronics1]

Was it large enough?

[@james5583]

@Epictronics1 yes, but was it the biggest you have on the shelf? 😁 Great videos btw :)

[@Epictronics1]

@@james5583 Thanks. Yep, I'm still looking for something bigger ;)

[@jaredbrown691]

Would you sell me a flashed bios chip? The PS/1 was my first family PC and it had 25mhz sx w/ 4 mb. Upgraded to 50 mhz overdrive but would love to get more L1 cache.

[@Epictronics1]

The files are available to download. Check with you local computer club and ask someone who has an EPROM programmer. They will flash your BIOS for you

[@antonioesposito585]

The best cpu I tried on my PS1 is a kingston turbochip, very stable and cool @160mhz. Cyrix or IBM 5x86 crashed, Pentium overdrive doesn't work.

[@Epictronics1]

If I ever find one. I'll give it a try

[@TheVdub1980]

Where's the first video

[@Epictronics1]

kzbin.info/www/bejne/oanFnJ6mmpuqnK8si=b6bkU9jLrhCGvQsR

[@wonk1976]

i found the same thing with a intel 486 dx4 100 16kb WB cpu it was faster clock for clock at 120mhz than my 5x86 133mhz in the same tests and the intel was almost as fast at 120 as the stock amd at 133

[@Epictronics1]

Oh, I should have tried a few DX4s for comparison

[@SianaGearz]

The LM2596 modules are all fake! The real component on there is LM2576 (or identical clone). It's only supposed to do 3A best case, and difference is the switching frequency at 52KHz as opposed to 150KHz, which means the inductor value is wrong! XL6019 modules seem to work really well, but i don't have the filthiest ones, mine are sort of on the legit side, bought from JoyIT/Conrad. Apparently XLSemi switched around the pinout on that chip to deter fakes, since they were having that issue with their prior chips.

[@Epictronics1]

50khz is actually quite spot on for these fake LM2596 :) Way off from the original 150

[@ingenfestbrems]

Are you from Sweden? Norway here 🇳🇴👋

[@Epictronics1]

Yes :)

[@TerroMin]

There is much better modules with real chips, like Mini560. They are is slightly warm under 10 Watt load

[@djdoo]

Wow! so many unstable China junk... Should you put a linear voltage regulator engraved inside a cpu heatsink and work both as a regulator and a heatsink for the CPU and throw to trash those trash ones? Great video as always a bit crashy but nice, cheers from Greece Jim

[@Epictronics1]

Thanks :)

[@yoppindia]

make correct selection of regulator cap and coil otherwise you will have toaster under x586 CPU, killing the CPU.

[@Epictronics1]

For sure. The recommended components in the datasheet are on the way with the next chip we'll try

[@ViegasSilva]

25:35 - 25:43 🔞