• Please review our updated Terms and Rules here

Intel Plato P54C-90 (430nx) won't boot

P

purple toupee

Member
Joined
Sep 21, 2023
Messages
10
Location
Seattle, WA, USA
Hey friends, I'd appreciate a sanity check on a theory. I've been assembling PCs since the late 80s, but I haven't spent a ton of time with it since it got old, so I'm far from an expert here.

I recently bought a supposedly-working Intel "Premiere/PCI II" Plato 430NX motherboard with a 90MHz P54C in it. The system appears healthy, and will run the memory check, but it hard freezes after. Whether or not I try to enter setup, it just freezes (keyboard stops responding to lock keys, etc). My theory is that this *might* be due to the original Dallas RTC, but I'm a little skeptical because this seems like an extreme reaction to an old RTC (and it's a hard fix to validate, for me).

What I've tried:
  • unplugging all things (even video card) except for RAM, case LEDs/buttons/speaker, keyboard, CPU & cpu fan.
  • clear CMOS jumper (was actually set to clear when I got it)
  • many repeat attempts

Possible causes?:
  • RTC -- it's an original, soldered Dallas RTC. I have seen the "Dremel away the side & solder a CR2032" fix. Is that what I'm left with?
  • Bad RAM? Seems unlikely, but I can't get to memtest so I can't rule it out.
  • Random hardware failure somewhere in the board...? (in other words, "everything else" I guess)

What do you all think?
 
Oh, I left one thing off. I doubt it matters, but: I'm using an ATX2AT converter with current limiting that's not quite up to what the motherboard needs. It maxes out at 4.75A on +12V, which this motherboard trips on first power-on. So it usually takes 2-3 tries to turn it on. I'm waiting for a more conventional (just wires) power adapter to rule this out though.
 
There is a known issue like that in the BIOSes for those early Intel brand Pentium motherboards described elsewhere in the forum history. IIRC it sometimes involves not being able to get into Setup, but the system will still let you continue past the CMOS warning and boot from a floppy and set the date and time from DOS, at which point it will let you into Setup. It's something to do with being unable to handle bad dates.

The Dallas chips, unlike a barrel battery board, don't retain settings if the battery is completely dead--even while the system is powered on. As soon as you exit setup (or the POST attempts to initialize values) CMOS registers are lost. So you might need a working Dallas before the above approach will even help.

If all else fails, once you have a socketed, working Dallas, you might try temporarily putting it in another board that takes Dallas--even a completely different one--to at least set a time and date and some sense of settings into it, then transplant it back.
 
Makefile said:
There is a known issue like that in the BIOSes for those early Intel brand Pentium motherboards described elsewhere in the forum history. IIRC it sometimes involves not being able to get into Setup, but the system will still let you continue past the CMOS warning and boot from a floppy and set the date and time from DOS, at which point it will let you into Setup. It's something to do with being unable to handle bad dates.
Click to expand...
Possibly the thread at [here].
 
Ah, right. So, the theory is something like this: This board needs its NVRAM initialized with certain settings. (Maybe this is a valid date, maybe it's other stuff like RAM settings and bus timings and all that other gory detail the BIOS configures.) This is nothing new, right? Since at least the 386 era, it's normal for BIOSes to control this stuff, and need valid settings to get to the boot sequence.

This typically isn't a problem. because we expect EITHER the "CMOS checksum failure" logic OR the "CMOS clear" jumper to properly initialize settings to a safe state. Even when they cannot retain the settings after power-down, the vast majority of machines can at least retain their settings and function until powered down. (I don't know much about this stuff. Maybe it's typical for a BIOS to cache the NVRAM settings it requires in normal RAM past boot-up?)

But for whatever reason, this particular BIOS with this particular RTC cannot retain the NVRAM settings even while powered on, so the board cannot reach the boot sequence. Probably it powers up, writes valid settings, loses them, and crashes when the garbage settings kick in.

Do I have that about right? And if so, it's reasonable to hope that a powered RTC might fix this particular problem?

Man, it's funny. I used to think the barrel batteries were pure evil. And I guess they are, at least the Dallas-style RTC doesn't kill the whole board, but at least getting rid of the barrel batteries was easy! I would like to have words with the genius at Intel who thought soldering a 24-pin DIP with a battery in it was a good way to save money.
 
purple toupee said:
Ah, right. So, the theory is something like this: This board needs its NVRAM initialized with certain settings. (Maybe this is a valid date, maybe it's other stuff like RAM settings and bus timings and all that other gory detail the BIOS configures.) This is nothing new, right? Since at least the 386 era, it's normal for BIOSes to control this stuff, and need valid settings to get to the boot sequence.
Click to expand...

Yes, suicide batteries have been a thing for a very long time. As to what happens when a battery gets low / goes bad on a machine to machine basis is always different. Some symptoms are benign, some are annoyances, some are serious random faults that can look like other problems, or the system can appear to be completely dead and not boot at all. In the absolute worst case, some machines have settings that require a special setup disk, and if you don't have that, you're hosed if the settings are lost.

purple toupee said:
This typically isn't a problem. because we expect EITHER the "CMOS checksum failure" logic OR the "CMOS clear" jumper to properly initialize settings to a safe state. Even when they cannot retain the settings after power-down, the vast majority of machines can at least retain their settings and function until powered down. (I don't know much about this stuff. Maybe it's typical for a BIOS to cache the NVRAM settings it requires in normal RAM past boot-up?)
Click to expand...

Clearing the CMOS memory doesn't always reset the settings to a sane state. Memory cells have a natural tenancy to drift if they don't have power, so even if you set everything to 00/FF/etc. reading those values back at a later time can return garbage. This is why some older machines will zero out the entire memory space to start with a known value. Because if a program stores some value in uninitialized memory, doesn't set it to a default value and goes back and reads it, it can result in erratic behavior.

purple toupee said:
But for whatever reason, this particular BIOS with this particular RTC cannot retain the NVRAM settings even while powered on, so the board cannot reach the boot sequence. Probably it powers up, writes valid settings, loses them, and crashes when the garbage settings kick in.

Do I have that about right? And if so, it's reasonable to hope that a powered RTC might fix this particular problem?
Click to expand...

Yep.

purple toupee said:
Man, it's funny. I used to think the barrel batteries were pure evil. And I guess they are, at least the Dallas-style RTC doesn't kill the whole board, but at least getting rid of the barrel batteries was easy! I would like to have words with the genius at Intel who thought soldering a 24-pin DIP with a battery in it was a good way to save money.
Click to expand...

You have to look at the RTC module in a historical perspective. The Dallas RTC module was designed for an operational life of 10 years, and many greatly exceeded that. I've had some Dallas RTC modules that lasted nearly 25 years before I even started getting low battery alerts on them. I don't think you'd want to reprimand a person that opted to spend more money on the design and use an RTC module vs a cheap crappy NiCD/CR2032 battery that'd at most give you 5 years of service before they died.

During the time your machine was built, computer technology was moving at a lightning speed, that machine could be obsolete in a year. Having a 10 year RTC was a step up from the norm of garbage NiCD batteries and was a lot more expensive to implement.
 
purple toupee said:
Man, it's funny. I used to think the barrel batteries were pure evil. And I guess they are, at least the Dallas-style RTC doesn't kill the whole board, but at least getting rid of the barrel batteries was easy! I would like to have words with the genius at Intel who thought soldering a 24-pin DIP with a battery in it was a good way to save money.
Click to expand...
It's unfortunate that it was soldered on. Other than saving money on a socket the only reason I can think of that you would do it is if you were afraid it would fall out?

Would the soldering process even require special handling of the Dallas chip since it's heat sensitive because of the battery?

Equivalent Dallas chips such as the 12887A+ or 12C887A+ (whatever's cheaper when you order from Mouser/DigiKey) are still being made for less than $20. The A-less versions too, if those are the only ones in stock, will work, except that you can't clear their settings externally. I know many immediately jump to modifying the old chip but a new one is still an option.
 
Makefile said:
It's unfortunate that it was soldered on. Other than saving money on a socket the only reason I can think of that you would do it is if you were afraid it would fall out?
Click to expand...

The Dallas RTC was specced to far outlive the service life of the machine, there was no reason to spend even more money on a socket.

Sockets also introduce unreliability. As they get older, they develop problems from oxide on the contacts that can cause connectivity issues. This can rapidly accelerate in humid climates. If I were replacing an RTC module today, I'd solder it on just like the factory did. With the machine being so old, there's a higher chance of something else on the board failing than the RTC module again.
 
During the time your machine was built, computer technology was moving at a lightning speed, that machine could be obsolete in a year. Having a 10 year RTC was a step up from the norm of garbage NiCD batteries and was a lot more expensive to implement.
Click to expand...
You're right, of course. I actually had a predecessor to this board, Batman, back in the day -- and I recall that the Intel motherboards were a clear step up. They set a new bar for quality, and it's easy to forget just how absurd it was to think about wanting to use a Pentium in ten years. I think I had my 486 for a year. RIP Moore's Law, I guess.

I know many immediately jump to modifying the old chip but a new one is still an option.
Click to expand...
From my reading, this seems to be from a fear that the old chips have been specially initialized in the factory with settings specific to the motherboard, and replacements just don't work. On the face of it, that doesn't quite smell right to me; it would EITHER require the RTC package to have some physical, internal alterations specific to the motherboard (highly, highly unlikely, given the uniform part numbers and sealed package) OR require the chip to have some special, separate NVRAM that survives the battery death, but that also just seems vanishingly unlikely, and uneconomical. Neither of those seem compatible with mass manufacture. It just doesn't make sense.

...says the guy planning on trying the hack to keep the old part...but I'm also planning on soldering on a socket, since I'll probably destroy the RTC anyway...
 
purple toupee said:
From my reading, this seems to be from a fear that the old chips have been specially initialized in the factory with settings specific to the motherboard, and replacements just don't work. On the face of it, that doesn't quite smell right to me; it would EITHER require the RTC package to have some physical, internal alterations specific to the motherboard (highly, highly unlikely, given the uniform part numbers and sealed package) OR require the chip to have some special, separate NVRAM that survives the battery death, but that also just seems vanishingly unlikely, and uneconomical. Neither of those seem compatible with mass manufacture. It just doesn't make sense.

...says the guy planning on trying the hack to keep the old part...but I'm also planning on soldering on a socket, since I'll probably destroy the RTC anyway...
Click to expand...
You're right in that once the battery is dead, the contents are lost, and if there is a bug such that the BIOS can't initialize a fresh RTC, it's going to have to be worked around, modified RTC or brand new one.

I believe there are some boards that are more picky about the specific RTC you use (where the 12[C]8[8]7[A]+ variations aren't quite as interchangeable) and the new chips aren't as easy to get in every country as in the US. And the inelegance of a sealed-in battery gets to some people perhaps. Thus the modifying of old chips, and even newly made ones with an external battery.
 
purple toupee said:
But for whatever reason, this particular BIOS with this particular RTC cannot retain the NVRAM settings even while powered on, so the board cannot reach the boot sequence. Probably it powers up, writes valid settings, loses them, and crashes when the garbage settings kick in.
Click to expand...
That has nothing to do with the BIOS but with the Dallas chip. It does not accept external power. It can only be powered by the internal coin cell. If that is flat, it does not matter if the system is powered on or off - it just can't retain any settings.

You have two options: just set the CMOS reset jumper and leave it like that. The system will work just fine then. You won't have RTC, but who cares about that anyway? And unless you want to add a second floppy disk drive, the default options are appropriate in most cases.

Second option: install a new Dallas replacement. Set the CMOS reset jumper and boot into DOS. Remove the jumper (while the system is on!), set the time and date. Do a soft-reset and enter the BIOS. It should all work now as well.

ps: don't use the on-board IDE. Even with the fixes released for the CMD640/RZ1000, it will still corrupt data on the hard drive even in DOS. Seriously, though, I would not recommend to use that mainboard for anything other than e.g. a test rig. Even when it works, it's slow and unstable.
 
Last edited:
You have two options: just set the CMOS reset jumper and leave it like that. The system will work just fine then. You won't have RTC, but who cares about that anyway? And unless you want to add a second floppy disk drive, the default options are appropriate in most cases.
Click to expand...
Unfortunately, this didn't work. The board just wouldn't boot without a working RTC.

Shockingly, I managed to desolder the Dallas package, solder on a socket, AND do the hack and it actually worked! The damn thing boots now. Thanks to everyone in this thread that offered encouragement!

The specific behavior was: Once I had the RTC powered, instead of freezing, the BIOS displayed a "CMOS date invalid" prompt after the memory test, and gave the option to boot or enter setup. Trying to enter setup would lock up the machine, but booting from floppy and setting the date worked around this as expected.

On to the next problem! Which is PCI video. I'll post a separate thread though.
 
purple toupee said:
Unfortunately, this didn't work. The board just wouldn't boot without a working RTC.
Click to expand...
You probably did set the wrong jumper. It works this way, mine is in that state for years. I did desolder the RTC and put a socket in, too, but still using the old RTC.

purple toupee said:
The specific behavior was: Once I had the RTC powered, instead of freezing, the BIOS displayed a "CMOS date invalid" prompt after the memory test, and gave the option to boot or enter setup. Trying to enter setup would lock up the machine, but booting from floppy and setting the date worked around this as expected.
Click to expand...
Exactly. That's the bevaviour I already described in the thread linked to by modem7.

Glad you got it working. You still won't have much fun with that mainboard. :D
 
Last edited:
Hahaha, let me bask in my accomplishment for a few first!

You could be right. I have an emotional attachment to this board, which may be clouding my judgment. But I don't want to go and chase down yet another motherboard.(...after throwing so much at this one...)

What are the actual issues? I'm aware of the stuff with the IDE controller, and that is challenging. I have a few thoughts I was going to try, I'm curious what you think, Timo:

  1. Try the same hack I was using on my 486 -- use an XTIDE to overlay its (4gb-friendly, robust) BIOS, but still use the on-board connector. This worked great...on a 486...with a generic ISA super-io card... I am skeptical that it will magically work with the integrated IDE, but it seems worth a try. (Maybe it will work with the integrated port that's labeled ISA IDE?)
  2. Purchase an inexpensive, period-correct(!) PCI IDE adapter, perhaps like this one, and disable the on-board IDE.
 
XTIDE won't help, as it's a hardware bug in the IDE chip. So using a separate IDE/RAID/SCSI controller is the way to go. Sadly, that takes quite some part of upper memory, so you have hard times freeing up enough base memory for some memory-hungry DOS games (e.g. Ultima 7 will be a no-go). Went thru all of this already.

I have an emotional attachment to this board as well, funny enough. It was the very first retro PC part I've ever got (or rather kept, as it was not retro at that point in time) and I used it as a secondary PC for some time. That was 25 years ago! Still the same board, still working - well, sort of. ;)
 

Attachments

  • IMG_20190907_134453445.jpg
    IMG_20190907_134453445.jpg
    151.5 KB · Views: 2
  • IMG_20190907_134516287.jpg
    IMG_20190907_134516287.jpg
    221.9 KB · Views: 2
  • IMG_20190909_155319402.jpg
    IMG_20190909_155319402.jpg
    476.1 KB · Views: 2
Last edited:
Are you a gamer or a tinkerer? If the latter, I don't see any reason not to use it; if anything it could be interesting to do your own experimentation with those IDE bugs or other problems with the board. Similar to how you are figuring out which PCI video cards work in it and which don't. Or you could look at the 430NX datasheet and look at how the BIOS options affect the chipset registers and see if there are missing settings. etc.
 
Ah, I see, XTIDE won't work because the bugs aren't in the BIOS (else they would have fixed it in the later revs, instead of disabling read-ahead and hoping for the best). Anyway, I am less concerned about the bugs than the 2GB/512MB limit, to be honest.

Are you a gamer or a tinkerer?
Click to expand...
Ehh, I'm all of the above, though I wouldn't say I'm terribly interested in a deeper understanding of a particular retro chipset. Right now my main thing is programming, which I mainly do via DOSBox-x, but I use the real thing for testing & benchmarking. But yeah, games are a big draw, at least theoretically. I hadn't realized an IDE adapter would have such an impact in the upper memory, but I guess that makes sense. I was tempted to dismiss this -- after all, in theory, late DOS games should have stopped caring so much about lower memory once protected mode became common, and earlier DOS games are arguably better played on different hardware anyway. But, your example reminded me that there was a long transitional period prior to the dominance of Watcom C & DOS4G/W when it still mattered, and those times had some of the most conventional memory-hungry games...

Honestly, I think this board (or a contemporary) is close to ideal for me, except that I want to have a non-contemporary amount of storage available for all of those games. If this just isn't the right motherboard, what would you two recommend I look for, next time I'm at my local recycling center/warehouse?
 
You must log in or register to reply here.
Back
Top