I’ve bought a bunch of 5x86 for doing overclock builds. I’ve found that 90% of chips run fine at 160mhz. 10% at 180mhz. 5% or less at 200mhz. Often the system crashes are due to the cache and/or ram not keeping up with the faster bus speeds. Also, I run the chips at 4V for 180mhz and 200mhz.
I think many of us suffer from a desire to fix what is broken, regardless of the merits or logic, the satisfaction gained is often worth more than what was actually being repaired.
Maybe another addition to the Socketblaster would be not only a clock generator, but also some jumpers to adjust the voltage regulator so you can under/overvolt the CPU? Might be able to get it up to 150-200MHz with a little kick in voltage and some more cooling.
I'm gonna try with overvoltage and better cooling next time. Jumpers for fixed voltages would require quite a lot of work I think. Maybe something like the mc34063 with different resistors could work, but it's fairly easy to adjust the voltages with the trim pot.
Measuring 40MHz with a multimeter is a little bit "how your doing" from the start I think. It is not so much only the Meter (I wouldn't have thought it goes up to even MHz range anyway), but the test leads: Those have pretty meaningfull inductances and capacitances at that frequency (dealing with RF here). Try to twist the two leads around each other maybe to get better results? But you should really be using a frequency counter with an oscilloscope style probe for these frequencies as the reflections on the unterminated multimeter leads might even break something in a bad case
Entirely depends on the multimeter as to how much "how you doin'" it is. I have a GS Pro-50 DMM that can accurately go up to 30 MHz, and it's so sensitive that it can pick up mains frequency from my bench light with the probe just floating in the air. Also got one of those AliExpress specials, a "Mestek" DMM, and it goes up to 100 MHz surprisingly. I've probed 386/486/Pentium bus clocks with it, and it's fairly accurate. On the other hand, I'm sure one of those Walmart or Harbor Freight $10 specials with a Hz setting will give you a bad time lol.
I am not entirely surprised that replacing good Nichicon capacitors has not resulted in a noticeable improvement with the PSU. The Japanese companies have excellent materials science, and I learned to trust their capacitors.
It may already be in the comments but.... Use enamel coated fine wire to do bodges like that. it is way thinner and looks very neat. I use it for all my track repairs. you can strip it with a blade in the areas you want to solder and the insulation will not melt easily with the iron when soldering. this will allow for many repairs to fine tracks without bulky insulation.
@@Epictronics1 I believe you are the only one hacking the PS/1 board like you are doing. I haven't seen anyone else on TH-cam doing it or read about it on the internet forums. Bravo! Pioneering for us to learn from and do some fun projects.
I just use the oscilloscope to measure frequency for anything over a couple of MHz. And if possible I'll use the spring attachment to minimise the ground loop, although that's probably unnecessary if you're only looking at the frequency and not the shape of the waveform itself. Also be careful inserting pins into the socket, it's possible to push the contact inside at the wrong angle and bend it so it no longer makes a good connection. It's a shame you didn't fix the power supply, but it's good you've got a spare. I'm a little surprised that it actually turns on. Usually that kind of ticking noise is the power supply failing to start because there's some issue with the bootstrap circuit. There's usually an auxiliary winding on the transformer which charges a capacitor to run the controller chip, and if that capacitor never charges enough to run the controller chip, it will just constantly pulse the transformer at that low frequency, causing that ticking sound. So I would usually suspect either the capacitor or the diode on the auxiliary winding when I hear that ticking noise. But yours actually starts, and is only making the ticking noise when it's off, which is really weird because it means the bootstrap circuit is fine.
Thanks for sharing. I haven't given up on the PSU. Since it started to work after we replaced those six caps, I think the logical thing to do is to replace some more or all the electrolytics. I'll order that diod too in case it's marginal
There is a small standby supply in the PSU so it can be turned on with the push of a button, it's probably where the noise is coming from. By the way, I always replace the small capacitors on the mains side of a SMPS and it often fixes the problem. It's quicker to eliminate them before looking for other problems.
@@looks-suspicious At the very end of the AT era, there were some weird soft start AT motherboards. I've worked on two myself, one being an FM Towns PC and the other an IBM tower like in this video. You can actually adapt an ATX power supply to work with these soft start boards, I did such a conversion for a customer of mine a few years ago. He had a fully decked out FM Towns Pentium that was overloading the original 200W AT supply. I heavily modified an Antec SL350 to replace the ATX connector with the AT P8/P9 connectors and the 3 pin header for the soft start. One wire goes to +5vsb and the other to the PS_ON, the last being ground. Machine worked great and no longer shut off after a few minutes.
I have a few ATX supplies that make noise in standby, too. AFAICT, it's normal. Probably just a really low duty cycle, or the controller changes the topology under low-load conditions.
What a coincidence: same ESR meter and same multimeter. It sure is not that difficult, but that makes me think I might have not choosen the worst ones. (edit: oh! I might have responded too early. I just watched the rest of the video. Lucky me!) I see by the Uni-T display you have it stock. I hacked mine to disable the Data Output, which enables the sleep timer. I left a mechanical switch hidden under one of the unused data port holes where no IR led is placed, so I can enable the output if an unlikely to happen day I want to log some data. The sleep mode is just a blast to me, as I forget it on more often than not. After my second battery went down after very little use I did the mod it and I cannot remember how long it has been running with the same battery. Totally recommend doing it for anyone even slightly absent-minded.
I'll measure again without it in the socket. Some people may not agree with putting so much work on an ESS card, but I have only good things to say about those cards
@@Epictronics1 Having a way to monitor Cpu voltage might be useful as you don't know how much it drops under load, especially if you overclock the cpu, could well be the cause of the instability. If i recall Cpu Galaxy had somewhere around 4 volts when he got 180 and 200 mhz clock speeds.
@@austinmaxi We have checked the voltages in a previous video. The buck gets stable under load at the intended voltage. In the next attempt, I'm gonna try 4v. thanks
As others have said a heatsink with some weight and a fan is good when using these 586’s. In my personal experience they don’t like doing 150-180mhz without extra voltage. To stabilise 160mhz I have always needed additional voltage, they scale well with volts and this will likely fix your stability issue :)
@@Epictronics1 try 3.4 all the way through to as high as 4v for very high clocks provided you use good cooling. I found I needed 3.5-3.7v to get these stable at higher clocks.
adjust your voltage regulator on the interposer to provide 4V. this will overvolt the 5x86 but almost guarantee at least one of them will be stable at 160MHz with the overvolt
@@Epictronics1 well really it just needs several pads added for resistors that set the voltage divider and another solder pad to link to each pad then you jump the voltage you want. Basically the same thing that is on most motherboards of that era. Then you can try 3.3 as well as 3.4, 3.5, 3.6, 3.7 etc etc. If you need help redesigning the board let me know and I can make it happen :-)
21:55, I think the 486 SX 20 would take the crown of the slowest on my first PC. Doom was a slideshow on it😅 Great to see what the platform could do in this video.
As usual, it was an absolutely awesome vid! :D Quick hint, to get off the hot glue easily, spray it with IPA and it will all disconnect from the rest super easily :)
I noticed the socket blaster that it is using a common inexpensive buck converter. I have heard and seen that those produce quite a bit of voltage ripple, might be worth trying a linear regulator and/or adding filtering to see if you can push the overclock further.
There is nothing wrong with those original 200V capacitors. The measured ESR and capacitance is actually better than the new ones. The PSU is a half bridge design and those 2 caps are in series. Rectified 240V AC is about 340V DC. They each have about 170V DC on them. 200V is fine.
@@felixokeefe In this case we don't need a tester to know that the large caps are _not_ excessively leaky. If they were electrically leaky enough to matter they would be self-heating and likely bulging. Plus leaky capacitors tend to read high when tested with a multimeter which we did not see here.
Seems like those boards are very FSB dependent (or more likely the Local Bus graphics subsystem is), since with a 40MHzx3 you were getting better results than the 33MHzx4.
for a completely dead PSU, you should check the smallest caps in the primary area (everything before the secondary heatsink/rectification). Especially Caps right next to heatsinks or between heatsinks and coils... Caps with 6,3mm Diameter and lower like to dry out and go bad that way.
Just saw that your issue is the noise when the PSU is turned off. And since its weird that the PSU seem to have a 5V Standby CIrcuit for some reason (0,02A, +5VS), I think that might not be an issue... @@Epictronics1
@@Epictronics1 Love what you do, please keep it up! I recently got an IBM PS2 Model 55LS that I guess isn't that common - no floppy! It's dead so I am going to start with the PSU. I may need some assistance if you are willing :)
The 5x86 that almost booted dos (nr. 3 was it?), reminds me that when I overclocked those CPUs, I had same results (booted dos ok but it would no run anything, it freezed at minutes after boot), until I used a properly mounted heatsink with fan; even then, sometimes it liked to hang and freeze, until I used thermal paste (it was the first time in my life to use thermal paste on a ceramic CPU) and only then I could run Win95. It seems to me that your CPU can run at 160MHz but you really need to cool it down. (or play a bit with its voltage?). The coil in your PSU may be singing just because the ferrite core vibrates; it may had became a bit loose, try to wiggle it to see, that may be the case. Adding a bit of silicon will fix the noise. Many coils I saw had a bit of glue or silicon to prevent them to vibrate (especially those working at audible frequecy), that noise seems to be the mains frequency. Electrically and functional it works ok.
That's very useful info. I'll make a new attempt with proper cooling. Yes, the PSU seems to work properly after I had replaced those caps. I would have replaced that coil but I can't find any part numbers or specs. The numbers written on it don't show any results on Google. Thanks!
Could be. The BIOS doesn't have an option for changing wait states. Unfortunately, I am not aware of a way to check wait states without having the option in the BIOS
I do recall back in the day, my Cyrix 80Mhz could consistently outclass faster CPUs w/ 33Mhz bus... in certain workloads, it seems to matter a lot more than I would expect. I suppose Doom uses a lot of bus time, so it adds up.
Most multimeters don't measure frequency. Those that do are usually cheap chinese ones and usually only measure up to 10MHz. You will need to buy either a dedicated frequency meter or just use the scope. You're not measuring frequency that often and you already have the scope so just use that. No need to buy another multimeter.
What was that sound card circuit board designed in the shape it was? I don't think I've seen another computer expansion card who's board is cut at what I assume is a 45 degree angle, except I think in an 8 bit keys video.
I have an IBM PS1 tower with this identical board in it. I made a comment about it on your first IBM PS1 modding video. Mine was configured different than yours... came stock with a 486sx 33mhz. It also came with the full 1mb video ram populated, among other things.
Oh, yeah, I was meant to ask you if your MB supports hard drives larger than 512MB. Apparently, there was an improved BIOS for this board but it was referred to as 07H1001.
@@Epictronics1 That's a good question. I hadn't officially tried it. Right now I'm running a 512mb compact flash solution. Buuuuut... it might go higher. The original owner had upgraded the computer to a Conner CFS1275A hard drive, which my research says is 1275mb. It was mostly dead when I got it, but it did briefly boot an install of windows 95 before dying for good. So it might be worth experimenting with. Also, perhaps there some sort of reader I could get which would allow me to rip a copy of the bios and archive it? What kind of reader should I buy?
@@vhfgamer Great! I use a TL866 Plus to rip eproms but I think the new version Xgecu T48 is about the same price. If it can do 1257MB HDDs it's definitely a newer BIOS!
@@Epictronics1 Ouch, those are a lot more money than I was expecting. Are there any basic cheaper versions that would read the chip? I'm not as interested in burning anything, I just want to help out.
1. Buy a (used) frequency counter instead of a new DVM, as otherwise recommended. 2. look for a 133ADZ (instead of 133ADW) type CPU which is supposed to be better suited for overclocking (no guarantee however).
Hey Epic, I just came got my hands on a ps/1 consultant (the 1992 2133-g43 variant with 2 sockests and 2 bios chips). I just bought a DX4ODP75 hoping to bump up the specs a bit. I also have a shuttle HOT-419 I think it will work in if it doesn't work out in the consultant. Do you know if the changes you were hoping to see have been added to any socketblaster projects yet?
@@theALFEST I read somewhere that slower CPUs are just factory rejects of faster CPUs. If that is true, it totally makes sense that most 133 will run happily at 150. I really hope someone will add 50MHz to the interposer. That would be a dream solution for this IBM
@@Epictronics1 Yeah, but a way to add some wait states is definitely needed for 50 mhz. Also good specimens go to 180 or even 200 but these cpus are harder to find and you need different MB for sure.
Not true, but the ADWs do run a little hotter so you need a heat sink and fan. I’ve found both ADZ and ADW are the best overclockers. The later 5x86s don’t overclock as well.
Honestly getting a multimeter to read 25mhz is already kind of impressive.. As elektro-peter mentioned. You'd be better off with a handheld oscilloscope for that kind of thing. (Hantek makes a decent couple, and there's a few other reputable brands making "affordable" ones too [USD-$200+ depending on quality+features])
I was under the impression that the x87 socket could be used as an upgrade socket by any OverDrive chip marked with "ODPR" which has a pin that disables the main CPU negating the need for your ODP bodge hack.
Not quite. that socket is specified for a 487 Math co. The ODP has that extra pin you mentioned. the ODPR is a regular overdrive. So, without the hack, you could get to 75MHz with an ODP100. In other words, you can "only" get to 25.88 FPS without hacking the board
The clock chips won't allow me to run it at 50 unfortunately. Hopefully, someone will redesign the interposer to give us all that option, on any board!
In general the faster the bus speed the better you will be, quicker ram access and such, of course your other hardware needs to support such speeds. I have been looking into PCI overclocking to 66MHz for a crazy fast 486 system, apparently some of the later PCI chipsets allowed for 66MHz bus as an option. I am just at the research stage on that though.
I got lucky with your hack. My PS1 2133 571 works perfectly with both kingston turbochip and amd x5-133ADW via interposer @160Mhz. I made 1798 realtics in DOOM. I'm desperately looking for a bios upgrade, specifically IBM P/N 07H1001 to work with Pentium overdrive and support more than 528MB hard disk.
Wonder if the instability of the overclock could be cause by the butt converter, is it possible to move the pots outside the enclosed area to get better measurements while pc is running and beable to adjust as needed while pc is running? Another option is setup and external clean power source for testing like battery pack or power brick.
Back when they were still relevant I never ran across a 5x86 that wouldn't run at 160mhz. While I'm sure there are some out there that won't, it seems odd that all 3 failed at that speed. I wonder if there is something else going on with the board. I also observed back in the day that my 3x50Mhz was faster than the same chip at 4x40Mhz. Bus speed seems to play a big part with 486 class CPUs. For what's it's worth this was on a vesa board with a Diamond Stealth 32 1mb graphics card.
It probably was mentioned already, but you're not quite 'maxing out DOOM' because it is the average FPS. When the scene is simple you probably get 50, while in busy areas it might still be 20. Of course in 1994 that would've been the equivalent of 240fps.. :-)
Yeah, could be electrically leaky or it was that small cap with 20ohm ESR. Which was the cap I would expect to be the cause before even seeing the footage, because several other TH-camrs have stated it's the one that "always" goes in these SMPS.... It's close to the heat sink so it gets cooked and has to handle the high-frequency switching noise which is much harder on it than the 120/100 Hz the usually much colder bulk capacitor needs to handle. Mr Carlson's tester would work but you can build a simple leak tester on a bread board with just a bench power supply, the tested capacitor, a resistor and a LED! It won't catch as much as his, but it'd very likely enough for this. IE charge the capacitor with the led/resistor as load over it using the bench supply (at a voltage that's safe with the resistor/LED combo). If the capacitor is really bad the bench supply's load limit may kick and the LED might not light up at all. If it does light up, the actual testing consist of turning of the power and watch the led slowly fade away. Or just go out nearly instantly if it's bad. IIRC with many ESR meters one of the signs of bad leaks is that the measured capacitance is too HIGH because the instrument misunderstand the leak for "more capacitance" - but it's not always there so the lack of this doesn't rule out leaks.
18:20 no they aren't ;-) Pro tip: Take a single strand of bare wire from a multi-strand wire cable. That wire is 0.15mm wide which is less than the width of the trace. Bare the trace, apply flux/solder to the trace and the wire then solder the tiny wire to the trace directly but only at one end. Bend the wire using tweezers so it is the same shape as the trace and continue along the trace soldering it down. After it is done cut off the excess piece so only a tiny wire remains bridging the cut. I've done this kind of repair probably 500 times on arcade boards over about 30 years of repairing stuff
yeah, I probably used a too thick wire for this repair. It's not too bad without the 10x magnification. It's not really noticeable when the card is installed :)
It's funny when you're repairing the pin headers on that sound card.. how you're speaking of the value those cards hold to the right person today (I am NOT that person any more ha) I remember the last box (a printer paper box, the big one) of isa cards I gave away many years ago.. probably 10 real sound blasters in it, a bunch of io cards, a lot of 286 and xt stuff that to me back then was totally worthless.. all stuff that worked but was pulled out of systems I had gotten rid of prior.. but man just thinking about the 386 and 486 machines i've tossed over the years just because at the time once Pentium came out, they were junk. JUNK. Huh. I guess I moved and didn't want to pay storage for all that stuff, my mother held on to ONE old system and one CRT for me, but all the rest I just gave away to a friend. Growing up in the 80s and 90s there were 5 of us in my small town who were into computers and BBS and all that.. we had SO MUCH computer stuff back then... oh the good old days!
@@Epictronics1 Those cut traces are (from what I see in the video) for the IDE port only. But when you talk about older systems that have usually only the IDE1 port on the I/O card, you will want to use IDE2 (from the soundcard) for an optical drive, especially today when SD/CF/SATA to IDE adapters have no way to select master/slave device on an IDE channel.
@@sebastian19745 Yes, the IDE controller on these ESS cards is excellent. I would normally use this card in older systems. I don't really remember why I went with an ESS for this project. I'll probably replace it with a SB16 when I need the ESS for another project.
I suspect you have a ripple problem. Considering the 487 socket was never intended for running power hungry CPUs it probably doesn't have enough capacitance on the power rail. Try adding some additional bypass capacitor close to the socket on the 5v rail.
That is an international PSU with a switch for 120 or 230v. But now that I think about it, shouldn't those caps be rated for at least 230v? Why 200v rating I wonder
No, they don't need to be rated for 230V. They are in series. The PSU is a half bridge topology and they each have about 175V DC across them. 240V AC rectified is about 350VDC. There is nothing wrong with those high voltage capacitors. Their ESR is actually lower than the new ones.
Please get a better multimeter. those uni-t ones are indeed cheap, but also known for not even passing their own specifications. Therefore you cannot believe a uni-t datasheet
That Cirrus is clearly on local bus, those speeds are impossible on ISA bus. But, because Cirrus 542x is only a 16bit chip the speed is still half of faster VLB chips. Cirrus 543x chips are 32bit and with 40MHz VLB clock it takes almost 40Mb/s 🔥 By the way, if 5x86 won't take 160MHz then it is 150MHz so FSB to 50MHz X 3. 😁
...so we need a new PLL chip. 😅 Maybe a any_pll or what it was called... 50MHz FSB needs fast L2 cache, 12ns tag and rest of the chips 15ns and of course 60ns memory. And still wait states are very likely needed.
Why didn't you use a wirewrap tool on the pins? That would have worked great and the gauge of the wire is easier to work with . Not criticizing what you do, because you do great work. I was just curious..
Not necessarily, I had VLB motherboard that had the FSB up to 50MHz and all VLB cards (video, I/O) worked well at any frequency. May be that the memory on the video card is slow; to test what really failed, it should run a mainboard tester (like troubleshooter or check it) to see what fails and where.
20:00 "I think that looks pretty good" No it doesn't. Thats really ugly. You should've used much thinner wire. Some 0.2mm enamelled copper wire or strands from a piece of multi strand wire. Antenna coax or screened audio cable has fine copper strands in it.
Love the videos although Im not a big fan of conversion perversion. If you want to play Doom buy a newer old ass computer LOL. Im maybe just a grumpy purist who likes to keep things pretty much bone stock. Thanks for what you do!
Thanks! I'm not likely to ever play DOOM on this IBM actually. In fact, I tend to use it to take notes, because of the lovely keyboard lol. I just want to know if it can be done. Some PCs got a lot of crap over the years for being slow and un-upgradable, like this PS/1, the PCjr and some PS/2s. I find it extremely rewarding if I can make them fast :)
In my experience you couldn't overclock AMD CPUs in this generation at all. I tried and never got jack out of them. 133mhz meant 133 mhz. Even a tiny overclock would crash.
I had mine (in 1998) run at 150MHz. When I tried at 160MHz, I had to add thermal paste, a bigger heatsink and a big fan to run it stable; later I tried a water solution and itworked for months. That was before 2000 when my computer was a 486. The VLB graphics and multi I/O worked perfectly, as well the ISA sound card. I was running Windows 95 at the time.
@@sebastian19745 Do you remember if you ran any benchmarks back then? Higher FSB seems to trump higher speed. I'm guessing it was a few FPS faster at 150 than 160.
@@Epictronics1 I only had ran motherboard testing programs like CheckIt and Troubleshooter to see if the system is in a stable working state. Other benchmark programs I had at the time were Norton SysInfo and ChrisVGA but I do not remeber if I ran them or the results if so. The higher FSB is, the faster the memory access is. If the ISA clock is set as a fraction of the FSB, then the ISA and VLB bus will be faster and of course the whole system will be faster. So yes, 120MHz CPU with 40MHz FSB is faster than 133MHz CPU with 33MHz FSB. If I remember right, VLB bus works at the FSB frequency and PCI at 33 MHz fixed frequency only. TheTroubleshooter program seemsto have dissapeared, mostly CheckIt is used today, I finded it on first versions of HBCD; it is useful because it test all sub sytems of the motherboard like DMA,IRQs, cache, CPU but also sound blaster and video. Try first versions 5 to 7 of HBCD, they are bootable CDs and can even extract old utilities, benchmark, sysinfo tools from there.
I doubt it is possible to make simple interposer that would have its own oscillator. Unlike 68k, x86 CPUs have fully synchronous bus. Having motherboard and CPU run asynchronously would result in almost instant setup and hold times violations causing a crash even before the CPU had a chance to initialize. Maybe PLL frequency synthesizer would be possible to utilize but then you would have to introduce some weird wait states and the interposer would not be a simple device anymore but rather fully fledged CPU card.
Support me on patreon.com/Epictronics
Join me on Twitter: twitter.com/epictronics1
I’ve bought a bunch of 5x86 for doing overclock builds. I’ve found that 90% of chips run fine at 160mhz. 10% at 180mhz. 5% or less at 200mhz. Often the system crashes are due to the cache and/or ram not keeping up with the faster bus speeds. Also, I run the chips at 4V for 180mhz and 200mhz.
I think many of us suffer from a desire to fix what is broken, regardless of the merits or logic, the satisfaction gained is often worth more than what was actually being repaired.
True. Fixing up these machines is immensely gratifying.
Agreed.
"This video SUCKS :(" Sorry to admit i laughed a bit. That's just how it goes sometimes.
haha, we'll get there eventually!
My first comp was a IBM 33mhz
486.
Cost a lot, ran pretty good..
Wish I still had it.
They are very nice machines. Especially if they are hacked :)
Maybe another addition to the Socketblaster would be not only a clock generator, but also some jumpers to adjust the voltage regulator so you can under/overvolt the CPU? Might be able to get it up to 150-200MHz with a little kick in voltage and some more cooling.
I'm gonna try with overvoltage and better cooling next time. Jumpers for fixed voltages would require quite a lot of work I think. Maybe something like the mc34063 with different resistors could work, but it's fairly easy to adjust the voltages with the trim pot.
Measuring 40MHz with a multimeter is a little bit "how your doing" from the start I think. It is not so much only the Meter (I wouldn't have thought it goes up to even MHz range anyway), but the test leads: Those have pretty meaningfull inductances and capacitances at that frequency (dealing with RF here). Try to twist the two leads around each other maybe to get better results? But you should really be using a frequency counter with an oscilloscope style probe for these frequencies as the reflections on the unterminated multimeter leads might even break something in a bad case
Ok! I'll keep that in mind, thanks
Entirely depends on the multimeter as to how much "how you doin'" it is.
I have a GS Pro-50 DMM that can accurately go up to 30 MHz, and it's so sensitive that it can pick up mains frequency from my bench light with the probe just floating in the air. Also got one of those AliExpress specials, a "Mestek" DMM, and it goes up to 100 MHz surprisingly. I've probed 386/486/Pentium bus clocks with it, and it's fairly accurate.
On the other hand, I'm sure one of those Walmart or Harbor Freight $10 specials with a Hz setting will give you a bad time lol.
A fiberglass pencil works awesome for fixing trashed board traces......a few times with the pencil and it leaves nothing behind but the copper
I am not entirely surprised that replacing good Nichicon capacitors has not resulted in a noticeable improvement with the PSU. The Japanese companies have excellent materials science, and I learned to trust their capacitors.
Problem: Rubber don't last forever and all wet lytics use rubber seals.
So it will go bad after some time and leak its fluids all over the stuff..
@@Stefan_Payne : And if it somehow doesn't leak, it can still go bad from drying out over decades.
That is an epic repair on that sound card! Excellent work!
Thanks John!
I have one like that in a 486 build I made. It’s probably the only OPL3 card that isn’t insanely expensive now.
It may already be in the comments but.... Use enamel coated fine wire to do bodges like that. it is way thinner and looks very neat. I use it for all my track repairs. you can strip it with a blade in the areas you want to solder and the insulation will not melt easily with the iron when soldering. this will allow for many repairs to fine tracks without bulky insulation.
Yeah, I have used it in some projects. It's very nice to work with and would probably have been a better choice for this card
I have to say you have the best content on the PS/1. It's my favorite computer. Keep it going!
Thanks!
@@Epictronics1 I believe you are the only one hacking the PS/1 board like you are doing. I haven't seen anyone else on TH-cam doing it or read about it on the internet forums. Bravo! Pioneering for us to learn from and do some fun projects.
My first PC was the 33 MHz dx version
Frankly I'm glad the board didn't fly! It looks nice, it should stay nice and safe on the bench and not fly somewhere and get wrecked!
No worries, this board is gonna fly safely. Eventually...
I'm a professional and only use Fluke and Keysight meters. Even a second hand Fluke is a good buy.
I just use the oscilloscope to measure frequency for anything over a couple of MHz. And if possible I'll use the spring attachment to minimise the ground loop, although that's probably unnecessary if you're only looking at the frequency and not the shape of the waveform itself. Also be careful inserting pins into the socket, it's possible to push the contact inside at the wrong angle and bend it so it no longer makes a good connection.
It's a shame you didn't fix the power supply, but it's good you've got a spare. I'm a little surprised that it actually turns on. Usually that kind of ticking noise is the power supply failing to start because there's some issue with the bootstrap circuit. There's usually an auxiliary winding on the transformer which charges a capacitor to run the controller chip, and if that capacitor never charges enough to run the controller chip, it will just constantly pulse the transformer at that low frequency, causing that ticking sound. So I would usually suspect either the capacitor or the diode on the auxiliary winding when I hear that ticking noise. But yours actually starts, and is only making the ticking noise when it's off, which is really weird because it means the bootstrap circuit is fine.
Thanks for sharing. I haven't given up on the PSU. Since it started to work after we replaced those six caps, I think the logical thing to do is to replace some more or all the electrolytics. I'll order that diod too in case it's marginal
There is a small standby supply in the PSU so it can be turned on with the push of a button, it's probably where the noise is coming from. By the way, I always replace the small capacitors on the mains side of a SMPS and it often fixes the problem. It's quicker to eliminate them before looking for other problems.
A standby circuit on old AT power supplies? How would that even work?
@@looks-suspicious This AT PSU is a bit unusual. It has a soft start
@@looks-suspicious At the very end of the AT era, there were some weird soft start AT motherboards. I've worked on two myself, one being an FM Towns PC and the other an IBM tower like in this video.
You can actually adapt an ATX power supply to work with these soft start boards, I did such a conversion for a customer of mine a few years ago. He had a fully decked out FM Towns Pentium that was overloading the original 200W AT supply. I heavily modified an Antec SL350 to replace the ATX connector with the AT P8/P9 connectors and the 3 pin header for the soft start. One wire goes to +5vsb and the other to the PS_ON, the last being ground. Machine worked great and no longer shut off after a few minutes.
I have a few ATX supplies that make noise in standby, too. AFAICT, it's normal. Probably just a really low duty cycle, or the controller changes the topology under low-load conditions.
What a coincidence: same ESR meter and same multimeter. It sure is not that difficult, but that makes me think I might have not choosen the worst ones. (edit: oh! I might have responded too early. I just watched the rest of the video. Lucky me!)
I see by the Uni-T display you have it stock. I hacked mine to disable the Data Output, which enables the sleep timer. I left a mechanical switch hidden under one of the unused data port holes where no IR led is placed, so I can enable the output if an unlikely to happen day I want to log some data. The sleep mode is just a blast to me, as I forget it on more often than not.
After my second battery went down after very little use I did the mod it and I cannot remember how long it has been running with the same battery. Totally recommend doing it for anyone even slightly absent-minded.
I'm very happy with the UNI-T. I may have a go at the mod. Thanks
You may need to adjust the ISA bus multiplier to support higher than 40MHz. Assuming that option is visible in BIOS?
Unfortunately, there is no such option in bios
I wonder if the buck regulator on the interposer is interfering?
Definitely a great video. I was happy to see that ESS card saved.
I'll measure again without it in the socket. Some people may not agree with putting so much work on an ESS card, but I have only good things to say about those cards
I never had one of those. I always had Sound Blasters in the 1990's@@Epictronics1
@@Epictronics1 Having a way to monitor Cpu voltage might be useful as you don't know how much it drops under load, especially if you overclock the cpu, could well be the cause of the instability. If i recall Cpu Galaxy had somewhere around 4 volts when he got 180 and 200 mhz clock speeds.
@@austinmaxi We have checked the voltages in a previous video. The buck gets stable under load at the intended voltage. In the next attempt, I'm gonna try 4v. thanks
it's okay @Epictronics - it's all about the journey, right?
Yes, but only if we reach the goal :)
You still got it plenty fast, this video didn't suck 😉
Thanks! Yeah, that is one fast PS/1! But I still have an itch to get it even faster :) I'll keep trying until we find the limit
Agreed! This was a success!
As others have said a heatsink with some weight and a fan is good when using these 586’s.
In my personal experience they don’t like doing 150-180mhz without extra voltage. To stabilise 160mhz I have always needed additional voltage, they scale well with volts and this will likely fix your stability issue :)
Great, I'll give it a try. thanks
@@Epictronics1 try 3.4 all the way through to as high as 4v for very high clocks provided you use good cooling. I found I needed 3.5-3.7v to get these stable at higher clocks.
@@Mevunky That's very useful info. I'll give that a try, thanks
Maybe also a variable voltage if you need to give it more voltage.
adjust your voltage regulator on the interposer to provide 4V. this will overvolt the 5x86 but almost guarantee at least one of them will be stable at 160MHz with the overvolt
Ok! I'll give it a try. Thanks
@@Epictronics1 well really it just needs several pads added for resistors that set the voltage divider and another solder pad to link to each pad then you jump the voltage you want. Basically the same thing that is on most motherboards of that era. Then you can try 3.3 as well as 3.4, 3.5, 3.6, 3.7 etc etc. If you need help redesigning the board let me know and I can make it happen :-)
The 40mhz FSB is running everything faster hence the gains over the 33mhz, even if the CPU is slightly higher clocked.
My first PC was the SX/20 Model; damn nostalgia.
You were lucky then. Very nice machines
Same for me. The model 2123 with a 386SX/20 and...2 MB of RAM? Bought it at Costco in 1990 or 1991. I even managed to run OS/2 on it for a hot minute.
21:55, I think the 486 SX 20 would take the crown of the slowest on my first PC. Doom was a slideshow on it😅 Great to see what the platform could do in this video.
I considered mentioning the SX20 but decided not to. Because almost nobody had one back then anyways lol
Haha, I believe they call that a burn 😂
As usual, it was an absolutely awesome vid! :D Quick hint, to get off the hot glue easily, spray it with IPA and it will all disconnect from the rest super easily :)
Thanks!
Fantastic video loved it. Thank you so much. I love the desoldering gun ore the muu muu gun .
Thanks!
I noticed the socket blaster that it is using a common inexpensive buck converter. I have heard and seen that those produce quite a bit of voltage ripple, might be worth trying a linear regulator and/or adding filtering to see if you can push the overclock further.
If nothing else works, I'll try with external power
There is nothing wrong with those original 200V capacitors. The measured ESR and capacitance is actually better than the new ones. The PSU is a half bridge design and those 2 caps are in series. Rectified 240V AC is about 340V DC. They each have about 170V DC on them. 200V is fine.
Unless you measure the leakage current of the old capacitor you can't say whether it is good or not.
@@felixokeefe In this case we don't need a tester to know that the large caps are _not_ excessively leaky. If they were electrically leaky enough to matter they would be self-heating and likely bulging. Plus leaky capacitors tend to read high when tested with a multimeter which we did not see here.
Seems like those boards are very FSB dependent (or more likely the Local Bus graphics subsystem is), since with a 40MHzx3 you were getting better results than the 33MHzx4.
To do that repair, I'd have used magnet wire and run it into the throughholes, then inserted the header and soldered it and the magnet wire.
for a completely dead PSU, you should check the smallest caps in the primary area (everything before the secondary heatsink/rectification).
Especially Caps right next to heatsinks or between heatsinks and coils...
Caps with 6,3mm Diameter and lower like to dry out and go bad that way.
I'll check if there are some more small caps between the two large heat sinks, thanks
Just saw that your issue is the noise when the PSU is turned off. And since its weird that the PSU seem to have a 5V Standby CIrcuit for some reason (0,02A, +5VS), I think that might not be an issue...
@@Epictronics1
Amazing work as always!! You forgot your happy cap marker!!! 😊😂 I also agree, that ESS card is worth it - most vintage stuff is!
Oh, crap! I'll fix it in the next repair attempt :) Thanks!
@@Epictronics1 Love what you do, please keep it up! I recently got an IBM PS2 Model 55LS that I guess isn't that common - no floppy! It's dead so I am going to start with the PSU. I may need some assistance if you are willing :)
@@TheRetroRecall Thanks. Of course, I'm more than happy to help out. I suppose you'll start by checking for a RIFA or a WIMA cap :)
Good morning. How you upgrade the cache. What support chips did you use . IDE and video probably is in local bus
Unfortunately, I don't have the PS/1 on the bench to check right now. I can check when I have some more space. Thanks
Looked up the GD5424, it is a VLB chip
Excellent
You should look at enameled wire for trace repair perhaps? They are much smaller and would be much easier to route.
Yeah, that would probably have been a better choice here.
The 5x86 that almost booted dos (nr. 3 was it?), reminds me that when I overclocked those CPUs, I had same results (booted dos ok but it would no run anything, it freezed at minutes after boot), until I used a properly mounted heatsink with fan; even then, sometimes it liked to hang and freeze, until I used thermal paste (it was the first time in my life to use thermal paste on a ceramic CPU) and only then I could run Win95. It seems to me that your CPU can run at 160MHz but you really need to cool it down. (or play a bit with its voltage?).
The coil in your PSU may be singing just because the ferrite core vibrates; it may had became a bit loose, try to wiggle it to see, that may be the case. Adding a bit of silicon will fix the noise. Many coils I saw had a bit of glue or silicon to prevent them to vibrate (especially those working at audible frequecy), that noise seems to be the mains frequency. Electrically and functional it works ok.
That's very useful info. I'll make a new attempt with proper cooling. Yes, the PSU seems to work properly after I had replaced those caps. I would have replaced that coil but I can't find any part numbers or specs. The numbers written on it don't show any results on Google. Thanks!
The problem may be that there are not enough memory/L2 cache wait states for the 40 MHz bus.
Could be. The BIOS doesn't have an option for changing wait states. Unfortunately, I am not aware of a way to check wait states without having the option in the BIOS
@@Epictronics1 run cachechk utility. Also you can try to replace cache chips with 15 or 12 ns ones and make sure ram is 60ns
The display adapter is probably VL-Bus and those tend to be very sensitive to bus clock so I'd bet that it's the cause of the problems at 40Mhz.
Oh, I should have tested 160 with the onboard graphics disabled. Then we would have known
But didn't it work fine at 120MHz? That would mean the entire motherboard, including the chipset and video, work fine at 40MHz.
@@eDoc2020 Yes, It runs stable at 3x40
I do recall back in the day, my Cyrix 80Mhz could consistently outclass faster CPUs w/ 33Mhz bus... in certain workloads, it seems to matter a lot more than I would expect. I suppose Doom uses a lot of bus time, so it adds up.
Looking forward to comparing 150 to 160 when/if we ever get there. 50MHz vs 40MHz bus
@@Epictronics1 Oh, I missed the part about 50Mhz bus. I didn't even know they made those. Nice...excited to see the results (if you can get there)!
@@ruthlessadmin This MB only goes to 40, but maybe we can hack it :)
Danke!
Thank you! :)
Most multimeters don't measure frequency. Those that do are usually cheap chinese ones and usually only measure up to 10MHz. You will need to buy either a dedicated frequency meter or just use the scope. You're not measuring frequency that often and you already have the scope so just use that. No need to buy another multimeter.
lesson learned, thanks
What was that sound card circuit board designed in the shape it was? I don't think I've seen another computer expansion card who's board is cut at what I assume is a 45 degree angle, except I think in an 8 bit keys video.
Yes, It has a rather unusual design. My only guess is that it was cheaper to manufacture this way
I have an IBM PS1 tower with this identical board in it. I made a comment about it on your first IBM PS1 modding video. Mine was configured different than yours... came stock with a 486sx 33mhz.
It also came with the full 1mb video ram populated, among other things.
Oh, yeah, I was meant to ask you if your MB supports hard drives larger than 512MB. Apparently, there was an improved BIOS for this board but it was referred to as 07H1001.
@@Epictronics1 That's a good question. I hadn't officially tried it. Right now I'm running a 512mb compact flash solution. Buuuuut... it might go higher. The original owner had upgraded the computer to a Conner CFS1275A hard drive, which my research says is 1275mb. It was mostly dead when I got it, but it did briefly boot an install of windows 95 before dying for good.
So it might be worth experimenting with. Also, perhaps there some sort of reader I could get which would allow me to rip a copy of the bios and archive it? What kind of reader should I buy?
@@vhfgamer Great! I use a TL866 Plus to rip eproms but I think the new version Xgecu T48 is about the same price. If it can do 1257MB HDDs it's definitely a newer BIOS!
@@Epictronics1 Ouch, those are a lot more money than I was expecting. Are there any basic cheaper versions that would read the chip? I'm not as interested in burning anything, I just want to help out.
@@vhfgamer Ok, no worries. I don't think there is anything cheaper. The old TL866 is about the same I think
1. Buy a (used) frequency counter instead of a new DVM, as otherwise recommended. 2. look for a 133ADZ (instead of 133ADW) type CPU which is supposed to be better suited for overclocking (no guarantee however).
I'll see if I can find one, thanks
Hey Epic, I just came got my hands on a ps/1 consultant (the 1992 2133-g43 variant with 2 sockests and 2 bios chips). I just bought a DX4ODP75 hoping to bump up the specs a bit. I also have a shuttle HOT-419 I think it will work in if it doesn't work out in the consultant. Do you know if the changes you were hoping to see have been added to any socketblaster projects yet?
Hey, Unfortunately, no. Good luck with the project!
Is it possible to overvolt these AMD chips to make them stable at 160 MHz?
The thought crossed my mind, that may be the way forward
That can help, but mine worked ok at 150mhz and 3.3 volts instead of 3.45.
@@theALFEST I read somewhere that slower CPUs are just factory rejects of faster CPUs. If that is true, it totally makes sense that most 133 will run happily at 150. I really hope someone will add 50MHz to the interposer. That would be a dream solution for this IBM
@@Epictronics1 Yeah, but a way to add some wait states is definitely needed for 50 mhz. Also good specimens go to 180 or even 200 but these cpus are harder to find and you need different MB for sure.
@@Epictronics1 For later revisions of AMD 486 cpus it is more marketing than tech. Later 486dx4/100 cpus often go up to 150mhz too.
The ones you want for overclocking are the ones marked with 133ADZ. the ADWs don't overclock nearly as well.
Ok, I'll see if I can find a 133ADZ. Thanks
Not true, but the ADWs do run a little hotter so you need a heat sink and fan. I’ve found both ADZ and ADW are the best overclockers. The later 5x86s don’t overclock as well.
Honestly getting a multimeter to read 25mhz is already kind of impressive.. As elektro-peter mentioned. You'd be better off with a handheld oscilloscope for that kind of thing. (Hantek makes a decent couple, and there's a few other reputable brands making "affordable" ones too [USD-$200+ depending on quality+features])
Yeah, lesson learned. I'll just use the scope I have already. The DHO800 is small enough to move around on the bench
I've had a couple of API power supplies fail on me in the past. Astec or Magnetek would have been much better choices.
I suppose I could replace it. But I would prefer to keep the socket 4 machine as original as possible. I'll try some more
I was under the impression that the x87 socket could be used as an upgrade socket by any OverDrive chip marked with "ODPR" which has a pin that disables the main CPU negating the need for your ODP bodge hack.
Not quite. that socket is specified for a 487 Math co. The ODP has that extra pin you mentioned. the ODPR is a regular overdrive. So, without the hack, you could get to 75MHz with an ODP100. In other words, you can "only" get to 25.88 FPS without hacking the board
I used 5x86 @ 50x3=150 mhz, it was faster than 4x40=160. Also as far as I know these cpus overclock to 160 very well.
The clock chips won't allow me to run it at 50 unfortunately. Hopefully, someone will redesign the interposer to give us all that option, on any board!
In general the faster the bus speed the better you will be, quicker ram access and such, of course your other hardware needs to support such speeds. I have been looking into PCI overclocking to 66MHz for a crazy fast 486 system, apparently some of the later PCI chipsets allowed for 66MHz bus as an option. I am just at the research stage on that though.
@@nigelrhodes4330 Sounds like a fun project. good luck!
I got lucky with your hack.
My PS1 2133 571 works perfectly with both kingston turbochip and amd x5-133ADW via interposer @160Mhz. I made 1798 realtics in DOOM.
I'm desperately looking for a bios upgrade, specifically IBM P/N 07H1001 to work with Pentium overdrive and support more than 528MB hard disk.
Sweet! I hope to get there too someday. What IBM used the 07H1001 BIOS?
It's an upgrade I found in old IBM archives pages. I couldn't link it on youtube.
@@antonioesposito585 Ok, I found the article. Let's hope someone finds a chip someday
@@Epictronics1 I also tried a cyrix 5x86 but unfortunately it crashed with my PS1, while it works perfectly with Aptiva 2144 921.
Check out the comment from @vhfgamer. The BIOS he has might be what you are looking for
Wonder if the instability of the overclock could be cause by the butt converter, is it possible to move the pots outside the enclosed area to get better measurements while pc is running and beable to adjust as needed while pc is running? Another option is setup and external clean power source for testing like battery pack or power brick.
I've been told in the comments that the multimeter isn't very good at measuring frequencies. It has worked well until this project though
There's a rumor the ADZ chips are better binned and can reach 160 MHz.
Look for 100 MHz multimeter. They are pricier but not old HP level pricy.
I wonder if there is any value of wrapping those caps close to the heatsink with a little bit of Kapton tape.
Good question, hard to tell
Did the repaired sound card work? Will you test the cdrom ide connector on it?
I ran out of time. I checked for shorts and continuity but didn't do a proper test. I'll test it and report back
Back when they were still relevant I never ran across a 5x86 that wouldn't run at 160mhz. While I'm sure there are some out there that won't, it seems odd that all 3 failed at that speed. I wonder if there is something else going on with the board.
I also observed back in the day that my 3x50Mhz was faster than the same chip at 4x40Mhz. Bus speed seems to play a big part with 486 class CPUs. For what's it's worth this was on a vesa board with a Diamond Stealth 32 1mb graphics card.
It probably was mentioned already, but you're not quite 'maxing out DOOM' because it is the average FPS. When the scene is simple you probably get 50, while in busy areas it might still be 20. Of course in 1994 that would've been the equivalent of 240fps.. :-)
Maybe your onboard ram or cache limits overclocking?
Quite possible. I'll try with different chips
The caps might be electrically leaky. Neither of those meters test for that. I'm thinking of the tester that Paul uses on Mr Carlson's Lab.
Ok, I'll probably replace them all
Yeah, could be electrically leaky or it was that small cap with 20ohm ESR. Which was the cap I would expect to be the cause before even seeing the footage, because several other TH-camrs have stated it's the one that "always" goes in these SMPS.... It's close to the heat sink so it gets cooked and has to handle the high-frequency switching noise which is much harder on it than the 120/100 Hz the usually much colder bulk capacitor needs to handle.
Mr Carlson's tester would work but you can build a simple leak tester on a bread board with just a bench power supply, the tested capacitor, a resistor and a LED! It won't catch as much as his, but it'd very likely enough for this.
IE charge the capacitor with the led/resistor as load over it using the bench supply (at a voltage that's safe with the resistor/LED combo). If the capacitor is really bad the bench supply's load limit may kick and the LED might not light up at all. If it does light up, the actual testing consist of turning of the power and watch the led slowly fade away. Or just go out nearly instantly if it's bad. IIRC with many ESR meters one of the signs of bad leaks is that the measured capacitance is too HIGH because the instrument misunderstand the leak for "more capacitance" - but it's not always there so the lack of this doesn't rule out leaks.
18:20 no they aren't ;-)
Pro tip: Take a single strand of bare wire from a multi-strand wire cable. That wire is 0.15mm wide which is less than the width of the trace. Bare the trace, apply flux/solder to the trace and the wire then solder the tiny wire to the trace directly but only at one end. Bend the wire using tweezers so it is the same shape as the trace and continue along the trace soldering it down. After it is done cut off the excess piece so only a tiny wire remains bridging the cut. I've done this kind of repair probably 500 times on arcade boards over about 30 years of repairing stuff
yeah, I probably used a too thick wire for this repair. It's not too bad without the 10x magnification. It's not really noticeable when the card is installed :)
It's funny when you're repairing the pin headers on that sound card.. how you're speaking of the value those cards hold to the right person today (I am NOT that person any more ha) I remember the last box (a printer paper box, the big one) of isa cards I gave away many years ago.. probably 10 real sound blasters in it, a bunch of io cards, a lot of 286 and xt stuff that to me back then was totally worthless.. all stuff that worked but was pulled out of systems I had gotten rid of prior.. but man just thinking about the 386 and 486 machines i've tossed over the years just because at the time once Pentium came out, they were junk. JUNK. Huh. I guess I moved and didn't want to pay storage for all that stuff, my mother held on to ONE old system and one CRT for me, but all the rest I just gave away to a friend. Growing up in the 80s and 90s there were 5 of us in my small town who were into computers and BBS and all that.. we had SO MUCH computer stuff back then... oh the good old days!
Yeah, most of us threw away our best toys. Who would have thought we would miss them two decades later
Why did you do this repair to the ESS card ? Do you realy use ide header on sound cards ? Ofcorse there is more cut traces - did you see them ?
Did I miss something? The ESS is for the PS/1. It doesn't need the IDE, but eventually, this card is going to end up in a system that does
@@Epictronics1 Yes, there are more cut traces.
@@nasko8605 I can't see any, where?
@@Epictronics1 Those cut traces are (from what I see in the video) for the IDE port only. But when you talk about older systems that have usually only the IDE1 port on the I/O card, you will want to use IDE2 (from the soundcard) for an optical drive, especially today when SD/CF/SATA to IDE adapters have no way to select master/slave device on an IDE channel.
@@sebastian19745 Yes, the IDE controller on these ESS cards is excellent. I would normally use this card in older systems. I don't really remember why I went with an ESS for this project. I'll probably replace it with a SB16 when I need the ESS for another project.
I suspect you have a ripple problem. Considering the 487 socket was never intended for running power hungry CPUs it probably doesn't have enough capacitance on the power rail. Try adding some additional bypass capacitor close to the socket on the 5v rail.
There are some caps on the interposer. Five 10uf caps if I remember correctly. Not sure if it's enough though.
I understand going to all this effort for the sound portion of the card but why to fix the IDE portion..
The IDE controller on these cards is excellent. I tend to use it with MBs that give me trouble with the CD-ROM drive
I wonder if the onboard cache is fast enough to run at a 160 MHz... Have you tried to disable it just to see if DOOM doesn't crash?
Yes, unfortunately, that didn't help
@@Epictronics1 Ugh, what a bummer.
A multimeter you can't trust is garbage regardless of specs/features.
Me: scratching head and thinking why (almost margin value) 230Volts caps were chosen. Then I realize that there is 120V in America... 🤣🤣
That is an international PSU with a switch for 120 or 230v. But now that I think about it, shouldn't those caps be rated for at least 230v? Why 200v rating I wonder
No, they don't need to be rated for 230V. They are in series. The PSU is a half bridge topology and they each have about 175V DC across them. 240V AC rectified is about 350VDC. There is nothing wrong with those high voltage capacitors. Their ESR is actually lower than the new ones.
Please get a better multimeter. those uni-t ones are indeed cheap, but also known for not even passing their own specifications. Therefore you cannot believe a uni-t datasheet
Would you recommend any specific multimeter?
That Cirrus is clearly on local bus, those speeds are impossible on ISA bus. But, because Cirrus 542x is only a 16bit chip the speed is still half of faster VLB chips. Cirrus 543x chips are 32bit and with 40MHz VLB clock it takes almost 40Mb/s 🔥 By the way, if 5x86 won't take 160MHz then it is 150MHz so FSB to 50MHz X 3. 😁
Great, now we only need to figure out how to set the speed to 50. The onboard chip doesn't have suport for 50
...so we need a new PLL chip. 😅 Maybe a any_pll or what it was called... 50MHz FSB needs fast L2 cache, 12ns tag and rest of the chips 15ns and of course 60ns memory. And still wait states are very likely needed.
@@djpirtu2 I'm an optimist :)
3:16 Story of my life.
Why didn't you use a wirewrap tool on the pins? That would have worked great and the gauge of the wire is easier to work with . Not criticizing what you do, because you do great work. I was just curious..
I don't have a wirewrap tool, unfortunately. Sounds like it would be a good choice for this repair. Maybe, I should get the tool. Thanks
You cant do 133/160MHz without fan and heatsink with thermal compound... it's printed on the cpu "fan/heatsink reqd"
I'll try again with proper cooling. Thanks
i have 2 of them cards and i was never able to get them to work :(
ESS cards? Have you tried with the drivers at Phil's computer lab? www.philscomputerlab.com/ess-audiodrive-es1868.html
It's never the input caps.
I bet that the graphics chip uses VLB and 40MHz FSB is too much for it with 160MHz.
I'll try to disable it and test with an ISA card
Not necessarily, I had VLB motherboard that had the FSB up to 50MHz and all VLB cards (video, I/O) worked well at any frequency. May be that the memory on the video card is slow; to test what really failed, it should run a mainboard tester (like troubleshooter or check it) to see what fails and where.
@@sebastian19745 Yes, some MBs VLB seem to handle faster FSB better than others
20:00 "I think that looks pretty good" No it doesn't. Thats really ugly. You should've used much thinner wire. Some 0.2mm enamelled copper wire or strands from a piece of multi strand wire. Antenna coax or screened audio cable has fine copper strands in it.
Saves it from trash
A real treasure now!
@@Epictronics1 Sorry I meant the soundcard repair! 😂 It's a nice machine I like these PS/1
5x86 hangs without a fan.
Really? that quickly? I touched it to check. It was lukewarm. I'll try again with a fan, thanks
@@Epictronics1 I do not remember how long it took. But mine CPU hung at 150 mhz with faulty fan
@@theALFEST That's very useful to know. Thanks!
Love the videos although Im not a big fan of conversion perversion. If you want to play Doom buy a newer old ass computer LOL. Im maybe just a grumpy purist who likes to keep things pretty much bone stock. Thanks for what you do!
Thanks! I'm not likely to ever play DOOM on this IBM actually. In fact, I tend to use it to take notes, because of the lovely keyboard lol. I just want to know if it can be done. Some PCs got a lot of crap over the years for being slow and un-upgradable, like this PS/1, the PCjr and some PS/2s. I find it extremely rewarding if I can make them fast :)
In my experience you couldn't overclock AMD CPUs in this generation at all. I tried and never got jack out of them. 133mhz meant 133 mhz. Even a tiny overclock would crash.
Ok, I've read that some people are getting 160 out of these am5x86s. Unfortunately, not the three chips I have :/
I had mine (in 1998) run at 150MHz. When I tried at 160MHz, I had to add thermal paste, a bigger heatsink and a big fan to run it stable; later I tried a water solution and itworked for months. That was before 2000 when my computer was a 486. The VLB graphics and multi I/O worked perfectly, as well the ISA sound card. I was running Windows 95 at the time.
@@sebastian19745 Do you remember if you ran any benchmarks back then? Higher FSB seems to trump higher speed. I'm guessing it was a few FPS faster at 150 than 160.
@@Epictronics1 I only had ran motherboard testing programs like CheckIt and Troubleshooter to see if the system is in a stable working state. Other benchmark programs I had at the time were Norton SysInfo and ChrisVGA but I do not remeber if I ran them or the results if so.
The higher FSB is, the faster the memory access is. If the ISA clock is set as a fraction of the FSB, then the ISA and VLB bus will be faster and of course the whole system will be faster. So yes, 120MHz CPU with 40MHz FSB is faster than 133MHz CPU with 33MHz FSB. If I remember right, VLB bus works at the FSB frequency and PCI at 33 MHz fixed frequency only.
TheTroubleshooter program seemsto have dissapeared, mostly CheckIt is used today, I finded it on first versions of HBCD; it is useful because it test all sub sytems of the motherboard like DMA,IRQs, cache, CPU but also sound blaster and video. Try first versions 5 to 7 of HBCD, they are bootable CDs and can even extract old utilities, benchmark, sysinfo tools from there.
@@Epictronics1 maybe they have, but I never knew of anyone that did.
I doubt it is possible to make simple interposer that would have its own oscillator. Unlike 68k, x86 CPUs have fully synchronous bus. Having motherboard and CPU run asynchronously would result in almost instant setup and hold times violations causing a crash even before the CPU had a chance to initialize. Maybe PLL frequency synthesizer would be possible to utilize but then you would have to introduce some weird wait states and the interposer would not be a simple device anymore but rather fully fledged CPU card.
That sounds discouraging. Maybe, there is a better solution
LOL funny vid!