Need some help repairing a Dell Precision Workstation 210.

[Mr. Perfect]

Hey all, I just snagged a Dell Precision Workstation 210 from a company that shut down and put everything out at the curb. Unfortunately it needs a little help.

First thing I did was open it up and do a visual inspection. Apart from an unholy amount of dust that I cleaned out with a canned air duster, there wasn't any visible damage apart from the rear case fan having a seized bearing. Internals include a 450MHz P2 (and a blanked out second Slot 1), 440BX chipset, 4x128MB 100MHz ECC ram sticks, an AGP 2x slot with a video card called a Fire 1K Pro in it (I'm not familiar with this, some sort of workstation card?), and a bunch of drives. The service tag indicates it was built 2/22/98 (the 2s are amusing).

Next I hooked it up to a monitor, mouse and keyboard. As soon as I plugged in the PSU the PC turned itself on (which I find suspicious). The drives and PSU fan spun up, but with no video, no POST and no beep codes. The front panel lights also lit up, but with the HDD activity light solid on and the power button lit up amber (the manual indicates amber is Sleep mode, On is green). It didn't respond to the power or reset buttons. After pulling the power I unplugged as many peripherals as I could, it was down to just the PSU, MB, CPU, video card and a single stick of RAM. As soon as the power cable was plugged in it did the same thing as the first time. For a third attempt I swapped the RAM stick to one of the others, reseated the CPU and VGA, and pulled the KB and mouse just in case it didn't like one of those. This time it made a tiny squeak with the internal speaker, but was otherwise the same. I unplugged it after that since I was out of ideas and didn't want to risk damaging it.

A day later I gingerly opened the lid on the PSU for a visual inspection. One of the big scary caps on the primary side vented sometime in the past, hard enough to knock the little plastic protection disk off the top. The other primary cap was still sleeved up, so couldn't be seen. Two smaller caps on the secondary side had also burst. That was enough to tell me this PSU was probably no good, so sealed it up and started looking for replacements. Unfortunately the board uses an ATX 24 pin rewired to some bonkers Dell proprietary pinout. Ebay has some of the same part number PSUs, but the options are either $50+ for untested 27-year-old PSUs or $200+ for "working pull" 27-year-old PSUs.

Here are pictures of the pinout. I find it a little odd that it's using an adapter to get from older 20 + 6 pin connectors to the 24pin. Especially since the label on the adapter is from September '98, a good seven months after the manufacture date on the case's tag.



So now for some questions:

1. Does this behavior sound like a bad PSU, or something more?
2. Can anyone share the Service Manual for the 210? I've found a User Manual online, but it's literally just how to use the system with no repair information.
3. Are there 24pin ATX adapters to let me hook up a standard PSU to this board?
4. If so, does anyone still make PSUs with a -5v line? The OEM PSU has one, but even my oldest spare PSU does not.

Well, thanks for reading all that. If anyone can help, this system is really quite interesting. I never had a slot CPU back in the day (skipped over them entirely), much less a dualy.

 

[GiGaBiTe]

video card called a Fire 1K Pro in it (I'm not familiar with this, some sort of workstation card?)

That's probably a Diamond Multimedia FireGL 1000 Pro. It's based on the 3D Labs Perimedia 2 video chip. The Perimedia 2 was 3D Labs attempt to try and stay relevant in the rapidly accelerating 3D graphics arms race, which failed.

It's a strange oddity being in that machine, because the Perimedia 2 was released a year prior to the Pentium II 450, which at the time was ancient history at that point. I usually see those types of machines with an ATI Rage something, a TNT/TNT2 or some 3dfx card. While it could theoretically be original to the machine, it's more likely someone added it later.

Regardless, finding drivers for that card can be a headache. Diamond used to keep a driver archive for their ancient cards, but some years ago they hid it in some archive subsection of their site that was a PITA to find. If you want to play 3D games on that machine, you'll definitely need to swap it out. If you just want to do 2D stuff and old DOS games, it should be fine for that. I have the PCI version of that card, and while I haven't used it in well over a decade, it was a good 2D card.

A day later I gingerly opened the lid on the PSU for a visual inspection. One of the big scary caps on the primary side vented sometime in the past, hard enough to knock the little plastic protection disk off the top. The other primary cap was still sleeved up, so couldn't be seen. Two smaller caps on the secondary side had also burst. That was enough to tell me this PSU was probably no good, so sealed it up and started looking for replacements. Unfortunately the board uses an ATX 24 pin rewired to some bonkers Dell proprietary pinout. Ebay has some of the same part number PSUs, but the options are either $50+ for untested 27-year-old PSUs or $200+ for "working pull" 27-year-old PSUs.

Whenever a primary side capacitor blows up like that, it's usually curtains for the supply. I'm VERY surprised it turned on enough to power the machine. If that's the case, I'd recommend trying to repair it. You may be able to get away with just replacing all of the capacitors in it. That can be a tedious job, depending on how cramped it is inside. Sometimes sub-boards and mosfets have to be removed to get access to capacitors buried deep between components.

I don't recommend buying another exact replacement for the reasons you describe. Other units are going to be the same age and have the same problems. Dell power supplies are notorious for capacitor issues, and any unit you buy will need to be completely recapped. With all of the tariff nonsense, a capacitor kit will easily cost as much or more than the price of the entire used unit. I'm glad that you didn't try to just plug a random ATX supply into the machine, as that would have probably blown the motherboard.

Here are pictures of the pinout. I find it a little odd that it's using an adapter to get from older 20 + 6 pin connectors to the 24pin. Especially since the label on the adapter is from September '98, a good seven months after the manufacture date on the case's tag.

A picture of the harness isn't terribly helpful. Dell has rarely observed industry standard color coding for power wires. There's probably a page out there somewhere with the pinout for the harness, but it'd be safer to just open the supply and look at the power board. They're generally pretty consistent at having a silk screen describing what wires are what voltages. There are also some control and status wires. PS_ON, PWR_GOOD and +5v standby.

It would be possible to build your own adapter harness to convert it to ATX, but it would be pretty expensive. I build harnesses and just the tools and wire will set you back a bit. You can get all of the Molex connectors and socket pins off of Mouser or Digikey.

So now for some questions:

1. Does this behavior sound like a bad PSU, or something more?
2. Can anyone share the Service Manual for the 210? I've found a User Manual online, but it's literally just how to use the system with no repair information.
3. Are there 24pin ATX adapters to let me hook up a standard PSU to this board?
4. If so, does anyone still make PSUs with a -5v line? The OEM PSU has one, but even my oldest spare PSU does not.

Well, thanks for reading all that. If anyone can help, this system is really quite interesting. I never had a slot CPU back in the day (skipped over them entirely), much less a dualy.

1. YES, the power supply is bad. Do not plug it in anymore with the blown capacitors. You risk damaging the PSU further, and the components its powering.
2. Dell does not provide repair manuals to end users. Even for Dell service techs (I was one back in the day) they only provide service manuals so far as to replace assemblies, they never did component level repair.
3. I'm sure that once upon a time, an ATX adapter harness was sold for this machine, but not in decades. There's no market for them anymore. I remember decades ago there was a guy on Etsy that custom made weird ATX adapter harnesses for specific machines, but he stopped a long time ago. I got one from him for my Power Macintosh G3 when the original PSU died.
4. The -5v rail was removed from the ATX spec back in 2003 and PSUs by the late 2000s stopped including it. The only place it was used was on the ISA bus, and even then, only very old ISA cards tended to use it. Sometimes you can get away with omitting it, but sometimes not. There are motherboards out there that will check for the presence of the -5v rail and refuse to power up with it missing. There are a few options to get the -5v rail. First, you could source an early 2000s ATX supply with it, but you'd have the same issue with capacitors. Second, you could hang an LM7905 negative voltage regulator off of the -12v rail. The third is an ISA dongle called a Voltage Blaster https://www.philscomputerlab.com/voltage-blaster--5v.html which takes the -12v rail from the slot and injects -5v back into the slot on the -5v pin. Not really something I'd recommend.

Slot 1/2/A systems were neat, but nothing special. They were a product of their time, when Intel and AMD couldn't work out how to have large caches directly on the CPU die while not having terrible yields. Performance did suffer because the cache ran at half the CPU speed. AMD's Slot A was worse, on some models, the cache could be only 33% of the CPU speed.

 

[Mr. Perfect]

That's probably a Diamond Multimedia FireGL 1000 Pro. It's based on the 3D Labs Perimedia 2 video chip. The Perimedia 2 was 3D Labs attempt to try and stay relevant in the rapidly accelerating 3D graphics arms race, which failed.

It's a strange oddity being in that machine, because the Perimedia 2 was released a year prior to the Pentium II 450, which at the time was ancient history at that point. I usually see those types of machines with an ATI Rage something, a TNT/TNT2 or some 3dfx card. While it could theoretically be original to the machine, it's more likely someone added it later.

Regardless, finding drivers for that card can be a headache. Diamond used to keep a driver archive for their ancient cards, but some years ago they hid it in some archive subsection of their site that was a PITA to find. If you want to play 3D games on that machine, you'll definitely need to swap it out. If you just want to do 2D stuff and old DOS games, it should be fine for that. I have the PCI version of that card, and while I haven't used it in well over a decade, it was a good 2D card.

Yep, it's a Diamond Multimedia card! It's got a sticker on it that might be a Dell part number?



I wouldn't be surprised if whoever bought this PC cut cost a little on the video. On one hand they sprung for a workstation class machine for their business, but on the other hand they only filled one CPU slot and apparently regretted buying only one stock 128MB stick (DIMM 1 is a Micron, but the other three are triple matching Kingstons I'm guessing where aftermarket). Since the Fire is nothing useful, I've got a Voodoo 5 5500 that may or may not work. Could be fun having a Glide machine to play games with again.

Whenever a primary side capacitor blows up like that, it's usually curtains for the supply. I'm VERY surprised it turned on enough to power the machine. If that's the case, I'd recommend trying to repair it. You may be able to get away with just replacing all of the capacitors in it. That can be a tedious job, depending on how cramped it is inside. Sometimes sub-boards and mosfets have to be removed to get access to capacitors buried deep between components.

I don't recommend buying another exact replacement for the reasons you describe. Other units are going to be the same age and have the same problems. Dell power supplies are notorious for capacitor issues, and any unit you buy will need to be completely recapped. With all of the tariff nonsense, a capacitor kit will easily cost as much or more than the price of the entire used unit. I'm glad that you didn't try to just plug a random ATX supply into the machine, as that would have probably blown the motherboard.

Er, I'm not actually any good with a soldering iron. I picked up an ESD safe Weller station with the intention to learn, but so far I'm only able to desolder components. Getting new components to wick solder is somehow not working the way it does for everyone on YouTube. Is there someone in the community who does recaps?

A picture of the harness isn't terribly helpful. Dell has rarely observed industry standard color coding for power wires. There's probably a page out there somewhere with the pinout for the harness, but it'd be safer to just open the supply and look at the power board. They're generally pretty consistent at having a silk screen describing what wires are what voltages. There are also some control and status wires. PS_ON, PWR_GOOD and +5v standby.

It would be possible to build your own adapter harness to convert it to ATX, but it would be pretty expensive. I build harnesses and just the tools and wire will set you back a bit. You can get all of the Molex connectors and socket pins off of Mouser or Digikey.

Hey, that'd be cool if they labeled the pins inside the PSU! I'll crack it open again tomorrow and see what's going on in there. Making an adapter might still be the only option I have the skills for.

1. YES, the power supply is bad. Do not plug it in anymore with the blown capacitors. You risk damaging the PSU further, and the components its powering.

Noted! Hopefully I didn't fry this thing poor thing already.

2. Dell does not provide repair manuals to end users. Even for Dell service techs (I was one back in the day) they only provide service manuals so far as to replace assemblies, they never did component level repair.

Oh. Well that's a bummer. Modern PCs I can at least get some documentation on by plugging the service tag into their support site. I was pleasantly surprised that Dell still has a driver library for the 210 up, but all the documentation is dead links.

3. I'm sure that once upon a time, an ATX adapter harness was sold for this machine, but not in decades. There's no market for them anymore. I remember decades ago there was a guy on Etsy that custom made weird ATX adapter harnesses for specific machines, but he stopped a long time ago. I got one from him for my Power Macintosh G3 when the original PSU died.

Makes sense, these must be few and far between nowadays. The rarity is kind of the appeal too.

4. The -5v rail was removed from the ATX spec back in 2003 and PSUs by the late 2000s stopped including it. The only place it was used was on the ISA bus, and even then, only very old ISA cards tended to use it. Sometimes you can get away with omitting it, but sometimes not. There are motherboards out there that will check for the presence of the -5v rail and refuse to power up with it missing. There are a few options to get the -5v rail. First, you could source an early 2000s ATX supply with it, but you'd have the same issue with capacitors. Second, you could hang an LM7905 negative voltage regulator off of the -12v rail. The third is an ISA dongle called a Voltage Blaster https://www.philscomputerlab.com/voltage-blaster--5v.html which takes the -12v rail from the slot and injects -5v back into the slot on the -5v pin. Not really something I'd recommend.

That makes sense, my oldest working PSU is a Seasonic S12 from 2005. Way past ISA times. That said I don't actually have any ISA cards to feed, trying without a -5V might be worth a shot if it won't damage anything. If you don't recommend the blaster, I'll avoid it.

Slot 1/2/A systems were neat, but nothing special. They were a product of their time, when Intel and AMD couldn't work out how to have large caches directly on the CPU die while not having terrible yields. Performance did suffer because the cache ran at half the CPU speed. AMD's Slot A was worse, on some models, the cache could be only 33% of the CPU speed.

This PC would really just be for fun, I suppose. Every time people reminisce about the old times Slot 1 slotkets and Celerons always come up, and it feels like I missed out a little. Neat is the appropriate word for it!

Thank you for the insights, I'll try to learn some more about the state of the PSU and post back.

 

[GiGaBiTe]

Yep, it's a Diamond Multimedia card! It's got a sticker on it that might be a Dell part number?

Yeah, that is a Dell part number. The date codes on the MoSys SGRAM show that card was made in late 1998 or 1999. 9852 = Week 52 (December) 1998. That's really late for the Perimedia 2. If you notice that some of the contacts on the AGP slot are missing, that's because the Perimedia 2 is just using the AGP slot as a 66 MHz PCI bus. It doesn't take advantage of any of the advanced acceleration that AGP has, the Voodoo5 5500 is the same, it just uses AGP as a fast PCI bus.

I noticed that it has SMD electrolytic capacitors on it, those need to be replaced ASAP. 1980s-2000s SMD electrolytic capacitors are shit and all have problems with leaking out of their base and causing severe damage to the PCB. Them being so close to the AGP connector is even more of a problem, the electrolyte loves to wick along traces and through VIAs and cause some pretty severe damage. Those don't look to have leaked outside of their base yet, which is good. The Voodoo5 is the same story. I'm actually recapping my V5 card today for the same reason. I removed the SMD caps a long time ago and bodged in some radials, but I'm tired of it looking ugly with caps sitting above the board.

If the motherboard has any SMD caps on it, those should be replaced as well. I have a Dell Dimension XPS T600 that has SMD caps on it that leaked. I replaced those and cleaned the board up before I gave it to someone else. Still have one of those machines I need to do that on.

I wouldn't be surprised if whoever bought this PC cut cost a little on the video. On one hand they sprung for a workstation class machine for their business, but on the other hand they only filled one CPU slot and apparently regretted buying only one stock 128MB stick (DIMM 1 is a Micron, but the other three are triple matching Kingstons I'm guessing where aftermarket). Since the Fire is nothing useful, I've got a Voodoo 5 5500 that may or may not work. Could be fun having a Glide machine to play games with again.

If the machine didn't have Windows NT 4.0 or Windows 2000 on it, then a second CPU was pointless. Windows 9x and DOS only support a single CPU. Usually a stubby continuity board is installed in the unused slot to keep bus reflections and noise down and preventing weird problems. Those things can be hard to find today since their function wasn't widely known and were usually thrown away. I think I still have a few of them hanging around. Having only one CPU installed in a dual CPU machine was actually a pretty common configuration for a long time. I have a Dell Precision T5610 from 2011-2012 that only came with a single CPU installed. I've since installed 2 x E5-2690v2s in it, which are 10C/20T parts each.

128 MB of RAM for Windows 9x was pretty standard at the time, and Windows 9x is pretty terrible with resource management anyway, so more memory doesn't always mean better. 512 MB is an upper limit for 98SE/ME before you start having strange issues.

Er, I'm not actually any good with a soldering iron. I picked up an ESD safe Weller station with the intention to learn, but so far I'm only able to desolder components. Getting new components to wick solder is somehow not working the way it does for everyone on YouTube. Is there someone in the community who does recaps?

I do recaps and board level repair. Not sure where you are in the world you are though, you don't have it in your profile.

That makes sense, my oldest working PSU is a Seasonic S12 from 2005. Way past ISA times. That said I don't actually have any ISA cards to feed, trying without a -5V might be worth a shot if it won't damage anything. If you don't recommend the blaster, I'll avoid it.

I don't recommend it because it gives me the heeby jeebies with injecting voltage back into the slot. It functioning correctly entirely relies on a perfect connection back into the slot, which is neigh impossible, especially on decades old slots. Oxidation is always a problem on old slots, which can cause high resistance. Since that card has no protections on it, it wouldn't be able to detect bad connections that cause heat and potentially meltdowns. It's not a fun job to replace ISA slots.

This PC would really just be for fun, I suppose. Every time people reminisce about the old times Slot 1 slotkets and Celerons always come up, and it feels like I missed out a little. Neat is the appropriate word for it!

Beware that Slotkets were generally vendor specific. There were universal Slotkets that were supposed to work with multiple different vendors, but it still wasn't a guarantee they'd work. Not all boards supported the different CPUs you could install in Slotkets. In addition to PGA370 slotkets, there were also Socket 8 Slotkets that allowed Pentium Pro CPUs to be used, but those were even more trouble to get working.

Here's an old post about Abit Slotkets bricking non-Abit boards: https://www.dell.com/community/en/c...10-bios-with-slotket/647e4b22f4ccf8a8de9365bd

Experimenting with old hardware can be fun, but experimenting with old Slotkets where the information about them has largely disappeared is not something I'd recommend.

 

[Mr. Perfect]

You where right on about the labels on the PSU PCB, they're all in there down in the weeds.



Some cable colors are actually standard: yellow is +12v, red is +5v, ground is black, white is -5v, purple is 5vSB, blue is -12v. Then they got a little creative with the other colors: brown is TFSC (though the 20+6 to 24pin adapter doesn't even pass this to the MB), blue/white is +3.3v, orange is PG, grey is PSON. And they're all in non-standard locations in the plug. They're all more or less the same number of lines, except it drops a +12v and adds a +5 line instead.

Yeah, that is a Dell part number. The date codes on the MoSys SGRAM show that card was made in late 1998 or 1999. 9852 = Week 52 (December) 1998. That's really late for the Perimedia 2. If you notice that some of the contacts on the AGP slot are missing, that's because the Perimedia 2 is just using the AGP slot as a 66 MHz PCI bus. It doesn't take advantage of any of the advanced acceleration that AGP has, the Voodoo5 5500 is the same, it just uses AGP as a fast PCI bus.

I misread the build date for the PC as a whole, it's 2/22/99 rather then 98, so that December card lines up nicely. It's interesting that the AGP cards weren't actually using AGP. It actually soothes an old grudge, because the PC we had for the Voodoo 3 generation didn't have an AGP slot and I had to settle for the PCI Voodoo. Guess it was less important then it seemed at the time.

I noticed that it has SMD electrolytic capacitors on it, those need to be replaced ASAP. 1980s-2000s SMD electrolytic capacitors are shit and all have problems with leaking out of their base and causing severe damage to the PCB. Them being so close to the AGP connector is even more of a problem, the electrolyte loves to wick along traces and through VIAs and cause some pretty severe damage. Those don't look to have leaked outside of their base yet, which is good. The Voodoo5 is the same story. I'm actually recapping my V5 card today for the same reason. I removed the SMD caps a long time ago and bodged in some radials, but I'm tired of it looking ugly with caps sitting above the board.

If the motherboard has any SMD caps on it, those should be replaced as well. I have a Dell Dimension XPS T600 that has SMD caps on it that leaked. I replaced those and cleaned the board up before I gave it to someone else. Still have one of those machines I need to do that on.

Oh, it probably does. There are only four can caps on the board as a whole, so maybe there's SMDs doing the job instead. Could be a nightmare. Here's a quick overview of the board.

If the machine didn't have Windows NT 4.0 or Windows 2000 on it, then a second CPU was pointless. Windows 9x and DOS only support a single CPU. Usually a stubby continuity board is installed in the unused slot to keep bus reflections and noise down and preventing weird problems. Those things can be hard to find today since their function wasn't widely known and were usually thrown away. I think I still have a few of them hanging around. Having only one CPU installed in a dual CPU machine was actually a pretty common configuration for a long time. I have a Dell Precision T5610 from 2011-2012 that only came with a single CPU installed. I've since installed 2 x E5-2690v2s in it, which are 10C/20T parts each.

128 MB of RAM for Windows 9x was pretty standard at the time, and Windows 9x is pretty terrible with resource management anyway, so more memory doesn't always mean better. 512 MB is an upper limit for 98SE/ME before you start having strange issues.

That's completely fair. No point in buying more hardware then will get used. The badging on the front of the case is dual Windows 98 and NT, so there's a chance for either. The one CPU probably tilts it towards 98 then. Thankfully the continuity board is still in place, though it looks like matching P2s aren't unreasonably expensive either. I was always more of a 2000 guy then a 9x one anyhow.

I do recaps and board level repair. Not sure where you are in the world you are though, you don't have it in your profile.

That's good to know. I'm in NJ, so shipping isn't entirely out of the question.

Beware that Slotkets were generally vendor specific. There were universal Slotkets that were supposed to work with multiple different vendors, but it still wasn't a guarantee they'd work. Not all boards supported the different CPUs you could install in Slotkets. In addition to PGA370 slotkets, there were also Socket 8 Slotkets that allowed Pentium Pro CPUs to be used, but those were even more trouble to get working.

Here's an old post about Abit Slotkets bricking non-Abit boards: https://www.dell.com/community/en/c...10-bios-with-slotket/647e4b22f4ccf8a8de9365bd

Experimenting with old hardware can be fun, but experimenting with old Slotkets where the information about them has largely disappeared is not something I'd recommend.

Guess I'll leave the slotkets be then. Most of my old games either fall into the "runs on current PC through GoG preservation" camp or into the "Will totally run on a P2" camp. Chasing more power for things in the second group would be a silly way to damage the PC. Dell documentation indicates this will officially run Katmai P3s too if performance is right on the edge.

 

[GiGaBiTe]

You where right on about the labels on the PSU PCB, they're all in there down in the weeds.

Some cable colors are actually standard: yellow is +12v, red is +5v, ground is black, white is -5v, purple is 5vSB, blue is -12v. Then they got a little creative with the other colors: brown is TFSC (though the 20+6 to 24pin adapter doesn't even pass this to the MB), blue/white is +3.3v, orange is PG, grey is PSON. And they're all in non-standard locations in the plug. They're all more or less the same number of lines, except it drops a +12v and adds a +5 line instead.

Not sure what TFSC is for, but if its not routed to any of the harnesses to components, then it'd be pretty straightforward to desolder the entire harness and solder it into a donor replacement supply if that one can't be repaired. I've done harness swaps and PSU restuffing before for proprietary/obsolete PSUs that are unobtanium.

I misread the build date for the PC as a whole, it's 2/22/99 rather then 98, so that December card lines up nicely. It's interesting that the AGP cards weren't actually using AGP. It actually soothes an old grudge, because the PC we had for the Voodoo 3 generation didn't have an AGP slot and I had to settle for the PCI Voodoo. Guess it was less important then it seemed at the time.

Just because the video card doesn't use advanced AGP features, doesn't mean that it equals a PCI video card. AGP is still a dedicated bus that runs at 66 MHz and has 266 MB/s of bandwidth in PCI mode. The video card on AGP has full and exclusive access to all of the bandwidth, and doesn't share it with anything else. PCI is a parallel drop bus that generally runs at 33 MHz in consumer PCs with 133 MB/s of shared bandwidth. Traffic can get very backed up on a heavily loaded PCI bus if you have heavy disk controller usage and heavy video card usage at the same time and cause significant performance loss.

There are 66 MHz PCI slots, but they're very rare in consumer PCs. You usually only find those in workstations or servers, and they're usually in the form of PCI-X slots. I've heard that the Voodoo3 can run in a 66 MHz PCI or PCI-X slot and benefit from the increased bandwidth, but I've never tried it myself.

Oh, it probably does. There are only four can caps on the board as a whole, so maybe there's SMDs doing the job instead. Could be a nightmare. Here's a quick overview of the board.

I don't see any SMD electrolytics, just a whole ton of tantalum capacitors. We tend to call them "tantrum" capacitors because their failure mode is violently exploding or going on fire when they fail. Age isn't really a factor in them going off, even new tantalum capacitors are known to be temperamental. So if your PC randomly explodes, you can go hunting for a blown tantalum after you change your pants.

That's completely fair. No point in buying more hardware then will get used. The badging on the front of the case is dual Windows 98 and NT, so there's a chance for either. The one CPU probably tilts it towards 98 then. Thankfully the continuity board is still in place, though it looks like matching P2s aren't unreasonably expensive either. I was always more of a 2000 guy then a 9x one anyhow.

If you opt to get a second PII 450, make sure that you get one with the same sSpec or stepping. Intel uses a 5 letter/number sequence that makes up the sSpec, you can just search for this on Ebay and you'll usually find what you're looking for. Just make sure you verify it, some sellers are dumb and use stock photos or have multiple CPUs and just randomly ship out parts. Mixing and matching different sSpec processors can lead to errant/undefined behavior. Some motherboards can detect that different stepping CPUs are used and may refuse to boot and/or display an error about incompatible CPUs.

Guess I'll leave the slotkets be then. Most of my old games either fall into the "runs on current PC through GoG preservation" camp or into the "Will totally run on a P2" camp. Chasing more power for things in the second group would be a silly way to damage the PC. Dell documentation indicates this will officially run Katmai P3s too if performance is right on the edge.

If the machine supports Katmai PIIIs, you can just buy those instead. The only difference between them and the PII is SSE support, which was a bastardized implementation that was really slow. But they did have higher clock speeds, up to 600 MHz. Slot 1 officially went up to 1133 MHz with Coppermine, but those high end parts today are very rare and command ridiculous prices.

I haven't looked up CPU compatibility with that machine, but the 440BX had a pretty wide range of CPU support, depending on what's in the BIOS of the motherboard.

Wikipedia has a list of Slot 1 CPU sSpecs if you want to search around for some to buy. If you scroll up from this list, Katmai parts are there as well.

List of Intel Pentium III processors - Wikipedia

en.wikipedia.org

 

[Mr. Perfect]

Not sure what TFSC is for, but if its not routed to any of the harnesses to components, then it'd be pretty straightforward to desolder the entire harness and solder it into a donor replacement supply if that one can't be repaired. I've done harness swaps and PSU restuffing before for proprietary/obsolete PSUs that are unobtanium.

I was able to chase down the function of the TFSC line, it's a Thermal Fan Speed Control. The 210 board has quite a few unpopulated solder masks, it appears the board that has those populated was the Precision Workstation 410. The 410 adds a pair of SCSI connectors, support for RDIMMs, an extra PCI slot and something they call a RAID port. That model has a Service Manual kicking around still.



The 24pin pinout is a bit different though. The 410 has a second -5VDC on pin 18 where the 210 seems to have a +5VDC.

The manual also has a cool little voltage map. The -5V is only going to the ISA like you predicted.

Just because the video card doesn't use advanced AGP features, doesn't mean that it equals a PCI video card. AGP is still a dedicated bus that runs at 66 MHz and has 266 MB/s of bandwidth in PCI mode. The video card on AGP has full and exclusive access to all of the bandwidth, and doesn't share it with anything else. PCI is a parallel drop bus that generally runs at 33 MHz in consumer PCs with 133 MB/s of shared bandwidth. Traffic can get very backed up on a heavily loaded PCI bus if you have heavy disk controller usage and heavy video card usage at the same time and cause significant performance loss.

There are 66 MHz PCI slots, but they're very rare in consumer PCs. You usually only find those in workstations or servers, and they're usually in the form of PCI-X slots. I've heard that the Voodoo3 can run in a 66 MHz PCI or PCI-X slot and benefit from the increased bandwidth, but I've never tried it myself.

Ah. Well, I'll continue to hold my grudge then. That PC had a fully utilized PCI bus: video card, audio card, NIC, modem. I bet AGP would have been useful for isolation alone.

I don't see any SMD electrolytics, just a whole ton of tantalum capacitors. We tend to call them "tantrum" capacitors because their failure mode is violently exploding or going on fire when they fail. Age isn't really a factor in them going off, even new tantalum capacitors are known to be temperamental. So if your PC randomly explodes, you can go hunting for a blown tantalum after you change your pants.

Oh. Well, that's equal parts reassuring and horrifying.

If you opt to get a second PII 450, make sure that you get one with the same sSpec or stepping. Intel uses a 5 letter/number sequence that makes up the sSpec, you can just search for this on Ebay and you'll usually find what you're looking for. Just make sure you verify it, some sellers are dumb and use stock photos or have multiple CPUs and just randomly ship out parts. Mixing and matching different sSpec processors can lead to errant/undefined behavior. Some motherboards can detect that different stepping CPUs are used and may refuse to boot and/or display an error about incompatible CPUs.

That's good advice. I've found a couple of "matching" sSpec processors on there, but didn't consider the possibility of getting a different part. I guess I should have, the picture not matching the product is a pretty standard eBay problem.

If the machine supports Katmai PIIIs, you can just buy those instead. The only difference between them and the PII is SSE support, which was a bastardized implementation that was really slow. But they did have higher clock speeds, up to 600 MHz. Slot 1 officially went up to 1133 MHz with Coppermine, but those high end parts today are very rare and command ridiculous prices.

I haven't looked up CPU compatibility with that machine, but the 440BX had a pretty wide range of CPU support, depending on what's in the BIOS of the motherboard.

Wikipedia has a list of Slot 1 CPU sSpecs if you want to search around for some to buy. If you scroll up from this list, Katmai parts are there as well.

List of Intel Pentium III processors - Wikipedia

en.wikipedia.org

I'm not entirely sure about the support either. There's a Dell document that just say it supports "Pentium IIIs", but if it was written early enough the entirety of the PIII list at time of writing would have been Katmai. The folks over at The Retro Web seem to have taken it that way at least. That Wikipedia list indicates the nicer Coppermine PIIIs used lower voltages, so I don't know if this board would fry those?

 

[GiGaBiTe]

The 24pin pinout is a bit different though. The 410 has a second -5VDC on pin 18 where the 210 seems to have a +5VDC.

Michael Dell - "We're shitty and never follow standards since forever."

Two separate -5v pins in the connector is just asinine. The current draw on that rail is 100 mA or less, when it is even used. One of the early Creative Sound Blaster ISA cards used it for opamp biasing.

Knowing there are at least two harness pinouts across similar machines, I'd want to carefully measure each pin on the motherboard to be absolutely sure where it went.

The manual also has a cool little voltage map. The -5V is only going to the ISA like you predicted.

Now the real question is if the -5v rail is monitored or not.

I'm not entirely sure about the support either. There's a Dell document that just say it supports "Pentium IIIs", but if it was written early enough the entirety of the PIII list at time of writing would have been Katmai. The folks over at The Retro Web seem to have taken it that way at least.

TRR can sometimes be a good resource, but it should be taken with a giant grain of salt. Much in the same way Wikipedia is run, TRR is run by manchildren who are belligerently intolerant of people that challenge their world views. They permabanned me from their Discord server because I made them "uncomfortable" by saying the Pentium 4 was a crap processor line, LOL.

I've found quite a bit of information on TRR to be flat out wrong and never has been corrected. But the same goes for any source on really old tech, a lot of good information that used to exist has been lost to time as companies and people come and go, or unfortunately get old and die off. Case and point is ChuckG who passed not that long ago, and took a lifetime of knowledge with him.

If you decide to download any BIOS versions from there, be sure that you have an escape plan if they fail. Nobody vets those downloads, I'd have an EEPROM programmer on hand and dump your existing BIOS before experimenting with any of them. Even if the board you have seems to match what it shown, you or the original source could have slightly different revisions of the board that have incompatible BIOSes.

That Wikipedia list indicates the nicer Coppermine PIIIs used lower voltages, so I don't know if this board would fry those?

That motherboard has automatic core voltage detection, but it does present a problem. The diagram shows 1.8-2.8v, which would cover all PII CPUs and Katmai PIIIs, but not the Coppermine parts. The question now is if that diagram is outdated and not updated. CPU clock speeds were ramping up so fast at that time that manufacturers were having a hard time keeping up. Dell may have released a later BIOS revision that added support for additional Coppermine CPUs. The only way to know may be to try one of those CPUs with the latest BIOS and see what happens.

It is entirely possible that the BIOS could support it, but the voltage regulation won't go low enough, which would slightly overvolt the CPU. But we're talking about 50-100 mV, so the worst that could happen is that the CPU would run a bit hotter than normal. You can just make sure that you have active cooling on the CPU(s). I think that case may be missing a baffle that goes over both CPUs to direct air over the heatsinks. As it is, that sole CPU is probably going to get rather warm with the fan being so far away, and more if the case is open. Those CPUs can't be passively cooled.

 

[Mr. Perfect]

Michael Dell - "We're shitty and never follow standards since forever."

Two separate -5v pins in the connector is just asinine. The current draw on that rail is 100 mA or less, when it is even used. One of the early Creative Sound Blaster ISA cards used it for opamp biasing.

Knowing there are at least two harness pinouts across similar machines, I'd want to carefully measure each pin on the motherboard to be absolutely sure where it went.

Yeah, I don't get it either. I wish there where some more of these systems around to compare to, but best I've found is one poster who was going to max out a 410 but stopped posting after a while.

Now the real question is if the -5v rail is monitored or not.

Man, I hope not. The ISA slot isn't important and being able hook up a more modern power supply without a -5v would make this all so much easier. The one PSU I've seen out there with a -5V rail is an Athena band AP-MPS3ATX30 that can be had for reasonable prices. It seems to be made for replacing old OEM supplies, but Athena isn't a name I'm familiar with. Have you heard anything about these?

TRR can sometimes be a good resource, but it should be taken with a giant grain of salt. Much in the same way Wikipedia is run, TRR is run by manchildren who are belligerently intolerant of people that challenge their world views. They permabanned me from their Discord server because I made them "uncomfortable" by saying the Pentium 4 was a crap processor line, LOL.

I mean, you're not wrong. Intel kept chasing clocks with Netburst while Athlons XPs kept being just as fast or faster at lower speeds. Clocks sold chips though.

I've found quite a bit of information on TRR to be flat out wrong and never has been corrected. But the same goes for any source on really old tech, a lot of good information that used to exist has been lost to time as companies and people come and go, or unfortunately get old and die off. Case and point is ChuckG who passed not that long ago, and took a lifetime of knowledge with him.

If you decide to download any BIOS versions from there, be sure that you have an escape plan if they fail. Nobody vets those downloads, I'd have an EEPROM programmer on hand and dump your existing BIOS before experimenting with any of them. Even if the board you have seems to match what it shown, you or the original source could have slightly different revisions of the board that have incompatible BIOSes.

Yeah, they've already got questionable information about the RAM for the 210 on there. They list it as supporting 768MB, but the manual states 512MB is the max with 4x128 UDIMMS. Even the fancier 410 needs RDIMMs to use 256MB sticks. I suppose BIOS updates could have changed that.

Thankfully Dell actually has the BIOS up on their download page still, so if this machine gets up and running and it needs a BIOS update, it can come straight from the manufacturer! I just wish it had patch notes. Modern Dells have pretty OK revision notes with BIOS updates, at least enough to mention new CPU or RAM support.

That motherboard has automatic core voltage detection, but it does present a problem. The diagram shows 1.8-2.8v, which would cover all PII CPUs and Katmai PIIIs, but not the Coppermine parts. The question now is if that diagram is outdated and not updated. CPU clock speeds were ramping up so fast at that time that manufacturers were having a hard time keeping up. Dell may have released a later BIOS revision that added support for additional Coppermine CPUs. The only way to know may be to try one of those CPUs with the latest BIOS and see what happens.

It is entirely possible that the BIOS could support it, but the voltage regulation won't go low enough, which would slightly overvolt the CPU. But we're talking about 50-100 mV, so the worst that could happen is that the CPU would run a bit hotter than normal. You can just make sure that you have active cooling on the CPU(s). I think that case may be missing a baffle that goes over both CPUs to direct air over the heatsinks. As it is, that sole CPU is probably going to get rather warm with the fan being so far away, and more if the case is open. Those CPUs can't be passively cooled.

The poster who was trying to do crazy stuff with their 410 mentions it came with 700MHz P3s, and since only Coppermine ran at those speeds I guess the board doesn't outright kill them. The last BIOS A10 came out August 2000, so potentially it could cover the whole Coppermine line if Dell was being generous.

Yeah, definitely no baffle in this case. I found the next generation Precision 220 came with P3s stock and did have a shroud forcing the air straight through the CPU heatsinks. Maybe a quick custom shroud would help.

 

[GiGaBiTe]

Man, I hope not. The ISA slot isn't important and being able hook up a more modern power supply without a -5v would make this all so much easier. The one PSU I've seen out there with a -5V rail is an Athena band AP-MPS3ATX30 that can be had for reasonable prices. It seems to be made for replacing old OEM supplies, but Athena isn't a name I'm familiar with. Have you heard anything about these?

Those Athena power supplies are junk. One of my customers bought one of those units for me to modify to use in their machine, and it was basically an IED. It had the same build quality as one of those shit "Logisys 480W" units, except the 3.3v section wasn't populated.

The -5v rail in the supply was made by an LM7905 on a small PCB screwed into the top of the secondary side heatsink. You could do the same yourself to add the -5v rail to a modern supply, just hang it off the -12v rail and send it to the ATX Minifit Jr. harness.

I mean, you're not wrong. Intel kept chasing clocks with Netburst while Athlons XPs kept being just as fast or faster at lower speeds. Clocks sold chips though.

Intel's PR pushed the MHz myth as far as they could go, until the Athlon 64 deflated their ego in 2003. But they doubled down on their stupidity again and again until they got their unreleased Tejas and finally realized that they were never getting to 10 GHz.

Yeah, they've already got questionable information about the RAM for the 210 on there. They list it as supporting 768MB, but the manual states 512MB is the max with 4x128 UDIMMS. Even the fancier 410 needs RDIMMs to use 256MB sticks. I suppose BIOS updates could have changed that.

440BX officially supported a max of 1 GB of RAM.

The poster who was trying to do crazy stuff with their 410 mentions it came with 700MHz P3s, and since only Coppermine ran at those speeds I guess the board doesn't outright kill them. The last BIOS A10 came out August 2000, so potentially it could cover the whole Coppermine line if Dell was being generous.

I've been trying to find some faster CPUs for my SuperMicro P6DBE. 700 MHz parts are all that seem to be available for not crazy prices. I currently have 2 x PIII Katmai 600 CPUs and 2 x PII 300 MHz parts available for it. the Super P6DBE supports up to 1 GHz parts.

 

[Mr. Perfect]

Those Athena power supplies are junk. One of my customers bought one of those units for me to modify to use in their machine, and it was basically an IED. It had the same build quality as one of those shit "Logisys 480W" units, except the 3.3v section wasn't populated.

The -5v rail in the supply was made by an LM7905 on a small PCB screwed into the top of the secondary side heatsink. You could do the same yourself to add the -5v rail to a modern supply, just hang it off the -12v rail and send it to the ATX Minifit Jr. harness.

Ah. Guess that's why I never heard of them before.

440BX officially supported a max of 1 GB of RAM.

Would you happen to know if that was later BIOS revisiona that added that support? Maybe 256MB UDIMMs didn't exist in August of 1998 when the 210's manual was written? At the time of writing, the manual said this board didn't support UDIMMs larger than 128MB. The 410 could run 256MB DIMMs, but only because it supported registered memory.

I've been trying to find some faster CPUs for my SuperMicro P6DBE. 700 MHz parts are all that seem to be available for not crazy prices. I currently have 2 x PIII Katmai 600 CPUs and 2 x PII 300 MHz parts available for it. the Super P6DBE supports up to 1 GHz parts.

How stable are your 600MHz Katmai? I noticed they use a higher voltage to run and when googling them an ancient Tomshardware article turned up where they had stability problems with their test CPUs. Their theory at the time was that Intel was just overvolting the CPUs to get the clocks up and stay competitive with AMD.

 

[GiGaBiTe]

Would you happen to know if that was later BIOS revisiona that added that support? Maybe 256MB UDIMMs didn't exist in August of 1998 when the 210's manual was written? At the time of writing, the manual said this board didn't support UDIMMs larger than 128MB. The 410 could run 256MB DIMMs, but only because it supported registered memory.

No clue. While the 440BX supported up to 1 GB of memory, it was up to system board vendors to how much memory was actually supported. There could be a BIOS limit, or there could be some physical hardware limit, like not all of the chip select lines were properly routed to the slots. So they could physically not support DIMMs over a specific size.

How stable are your 600MHz Katmai? I noticed they use a higher voltage to run and when googling them an ancient Tomshardware article turned up where they had stability problems with their test CPUs. Their theory at the time was that Intel was just overvolting the CPUs to get the clocks up and stay competitive with AMD.

They're stable, but they run very hot. Running UT99, they started touching the warning temperature I had set in the BIOS (65C) and the board started chirping the PC speaker. Thermal junction on these CPUs is 82C, but the cooler the better. I might buy some better fans to increase the airflow over the heatsinks at some point.

Intel was aggressively binning their CPUs in their fierce competition with AMD, but these parts weren't the ones with the trouble. The CPU that was recalled was the original Coppermine 1133 MHz part, several tech reviewers at the time noted the instability problems with the processor and Intel eventually was forced to do a recall on it and re-release it later when they could actually get it stable. This is why that specific revision is so rare today, very few of them were released.

There's nothing noteworthy about them besides them being the highest clocked Coppermine part. The Tualatin-S was a better part, higher clocks up to 1.4 GHz and had 512k of L2 cache. I have two of those in my Super P3TDDE.

 

[Mr. Perfect]

So this PC project obviously got lost in the holidays, but I got back to it today with some satisfying success. I used a molex pin extractor to turn a 24pin ATX extension into a ATX-to-Dell adapter and it worked! The trick of reading the PSU's PCB labels to figure out which line was which worked perfectly for this, so thank you for that! Apparently the -5v line is not being monitored, since this PSU doesn't have one and nothing complained. Presumably the ISA slots won't be functional.



The 210 posted, alerted about a bunch of peripherals that where unplugged, and then proceeded to boot Windows 2000 to the desktop. The system clock even had the right year, month and day. The hours and minutes where a smidge slow, but who can begrudge it that? Whoever had this last didn't bother setting a password on the user, so I've got full access to it. Event Viewer shows that it was running continuously up until the middle of the afternoon on 2/20/2023.



The hard drive might be a little sketchy, since Windows rebooted during the first boot attempt and there was a couple of clicks that reminded me of an IBM Deathstar. The first order of business will be getting some other drive to boot from. Any recommendations on this? I'm thinking one of Startek's PATA to SATA adapters hooked up to a modern drive. Reading some older threads makes it sound like there are BIOS size limits (120GB?) to what can be connected to something this old though, along with the natural limits of 32bit Windows. It would we nice to pop a SSD in there, but without TRIM or 4k sector support (isn't that a thing with old OSes and new drives?) in 2000 I'm wondering if it'll just write amplify the drive to death...

 

[GiGaBiTe]

Apparently the -5v line is not being monitored, since this PSU doesn't have one and nothing complained. Presumably the ISA slots won't be functional.

Unless you have an original Sound Blaster or other equally ancient 8 bit ISA cards, not having -5v won't really be an issue. Later 16 bit ISA cards generally didn't use the -5v rail that often. Missing -12v on the other hand would be a bigger problem, but you don't have to worry about that, since you have it. If you do end up needing the -5v rail, you can just wire up a negative charge pump circuit, or hang an LM7905 off the -12v rail.

I'm in the process of building a custom power supply for a Teradrive, and I'll be using a charge pump for the -5v rail.

The hard drive might be a little sketchy, since Windows rebooted during the first boot attempt and there was a couple of clicks that reminded me of an IBM Deathstar. The first order of business will be getting some other drive to boot from. Any recommendations on this? I'm thinking one of Startek's PATA to SATA adapters hooked up to a modern drive. Reading some older threads makes it sound like there are BIOS size limits (120GB?) to what can be connected to something this old though, along with the natural limits of 32bit Windows. It would we nice to pop a SSD in there, but without TRIM or 4k sector support (isn't that a thing with old OSes and new drives?) in 2000 I'm wondering if it'll just write amplify the drive to death...

I'm not really a fan of IDE to SATA adapter boards, many of them have issues with silent data corruption. If you don't have any period correct IDE drives, I'd recommend a DOM (Disk on Module) or a CF to IDE adapter and use a CF card.

As for limits, I'm not sure what geometry limits your machine has. Dell probably changed the limits with a later BIOS release, so you should get the latest BIOS version you can on there.

Windows 2000 has a 28 bit LBA limit of 128 GB, unless you install SP4 and the 48 bit LBA patch. I don't really recommend installing to a partition that large because formatting will take forever in setup. And while NTFS is more fault tolerant than FAT32, Windows 2000 isn't as good at dealing with huge partition sizes as later versions of Windows.

 

[Mr. Perfect]

Unless you have an original Sound Blaster or other equally ancient 8 bit ISA cards, not having -5v won't really be an issue. Later 16 bit ISA cards generally didn't use the -5v rail that often. Missing -12v on the other hand would be a bigger problem, but you don't have to worry about that, since you have it. If you do end up needing the -5v rail, you can just wire up a negative charge pump circuit, or hang an LM7905 off the -12v rail.

I'm in the process of building a custom power supply for a Teradrive, and I'll be using a charge pump for the -5v rail.

Nope, no ISA cards. I've got old Sound Blasters, one of which would be nice to install for EAX games, but they're all PCI.

Do you have a build post of your -5v project?

I'm not really a fan of IDE to SATA adapter boards, many of them have issues with silent data corruption. If you don't have any period correct IDE drives, I'd recommend a DOM (Disk on Module) or a CF to IDE adapter and use a CF card.

I have one drive, but it is a NOS 2.5" IDE drive that never made it into a laptop. Would you say the 40 pin desktop to 44 pin laptop IDE adapters are similarly sketchy? They're not actually converting a protocol at least, but they're all unbranded and cheap looking.

As for limits, I'm not sure what geometry limits your machine has. Dell probably changed the limits with a later BIOS release, so you should get the latest BIOS version you can on there.

Windows 2000 has a 28 bit LBA limit of 128 GB, unless you install SP4 and the 48 bit LBA patch. I don't really recommend installing to a partition that large because formatting will take forever in setup. And while NTFS is more fault tolerant than FAT32, Windows 2000 isn't as good at dealing with huge partition sizes as later versions of Windows.

Hmm, ok cool. Dell still has the latest A10 BIOS for the 210 on their driver site, so I can get that done. Will probably wait until I'm sure the PC is stable first, don't want to have it flake out in the middle of the flash process. Might just leave it running Memtest for a while and see how it handles that.

 

[GiGaBiTe]

Do you have a build post of your -5v project?

It's not really a project, it's just an LM2662 and a couple of capacitors, very simple to make.

LM2662 data sheet, product information and support | TI.com

TI’s LM2662 is a 200mA switched capacitor voltage converter. Find parameters, ordering and quality information

www.ti.com

I just bought premade boards on Amazon, and I'm going to mount one inside the power supply.

2PCS LM2662 Switched Capacitor Negative Voltage Converter Module +5V/-5V 200mA: Amazon.com: Industrial & Scientific

2PCS LM2662 Switched Capacitor Negative Voltage Converter Module +5V/-5V 200mA: Amazon.com: Industrial & Scientific

www.amazon.com

I have one drive, but it is a NOS 2.5" IDE drive that never made it into a laptop. Would you say the 40 pin desktop to 44 pin laptop IDE adapters are similarly sketchy? They're not actually converting a protocol at least, but they're all unbranded and cheap looking.

The only sketchy part about those converters is that sometimes they're not keyed and can allow you to plug the cable or drive in upside down, which generally results in something blowing up. So just be extra sure you have the orientation correct and you're not a row of pins off.

Hmm, ok cool. Dell still has the latest A10 BIOS for the 210 on their driver site, so I can get that done. Will probably wait until I'm sure the PC is stable first, don't want to have it flake out in the middle of the flash process. Might just leave it running Memtest for a while and see how it handles that.

My experience with Dells over the decades, it's best to get the latest BIOS version on as soon as possible. Early BIOS revisions, especially those that predate what is available from Dell tend to have significant issues.