I cleaned those points and it turned on! It had stopped turning on and was in the three long beep state. Now it is in an infinite boot loop: right after the memory test the system beeps and restarts a few times and eventually ends up with "CMOS Inoperational. System halted" with serveral loud short beeps (like in a loop, 50+ times).

There are two cracked tants near the cache block, the tants on the memory banks looks fine. Do you folks think I should try to recap the whole board? Do you have a recommendation on what type of caps I should use? Should I stick with tants or maybe replace them with ceramic or electrolyte-based?

Also, without an oscilloscope, is it possible to test the xtal on pins 61-62? The voltage drop on multimeter there is of about 0.9V (across pins 61 and 62).

Is just just my old eyes or are you using a 25Mhz DX with a 66Mhz crystal? Assuming it is 1/2 like the 16Mhz xtal for 8Mhz ISA bus you should use a 33Mhz cpu.... or maybe am over thinking things 😀

BitWrangler wrote on 2021-12-31, 17:56:

First step, try a battery on it.

However there's a couple of things I'm not happy with, from the getting an unknown board to exhibit full functionality point of view... i) The 66 Mhz crystal implies that the board is hardwired for 33Mhz, and I don't see enough jumpers in the right areas to be convinced that's settable. Therefore a 25Mhz CPU may be overclocked, which can lead to spurious operation unless carefully tweaked and tuned, or maybe is one of the "losing tickets" i.e. a CPU that won't overclock. ii) The cache at 25ns is a bit on the slow side for 33Mhz operation also, I would prefer to see 20ns in it.. Seems like this board might have been bare at one point and someone just stuck available parts in it, whether they were ideal or not.

Wouldn't 25ns cache be good up to 40MHz?

alpm wrote on 2021-12-31, 18:28:

Hi BitWrangler, thanks for the analysis! […]
Show full quote

Hi BitWrangler, thanks for the analysis!

BitWrangler wrote on 2021-12-31, 17:56:

First step, try a battery on it.

I'm using a CR2032, do you think this should work?

BitWrangler wrote on 2021-12-31, 17:56:

However there's a couple of things I'm not happy with, from the getting an unknown board to exhibit full functionality point of view... i) The 66 Mhz crystal implies that the board is hardwired for 33Mhz, and I don't see enough jumpers in the right areas to be convinced that's settable. Therefore a 25Mhz CPU may be overclocked, which can lead to spurious operation unless carefully tweaked and tuned, or maybe is one of the "losing tickets" i.e. a CPU that won't overclock. ii) The cache at 25ns is a bit on the slow side for 33Mhz operation also, I would prefer to see 20ns in it.. Seems like this board might have been bare at one point and someone just stuck available parts in it, whether they were ideal or not.

That's concerning. What would be the ideal approach in your opinion? Should I try to find a 33MHz 386 and replace it? As for the cache, should I try to swap the memory chips?

Sorry for the many questions folks, I'm a complete noob on these 286/386 era boards.

I would replace the crystal with a socket and a 50MHz crystal, if your budget is tight. Either that or a 33MHz CPU, but using a socket for the crystal would allow you to use any speed crystal/CPU combination. A socket and crystal would be cheaper than a CPU, even at CPU-World prices (cheapest place I know to buy the CPU, when they come up for sale)

debs3759 wrote on 2022-01-01, 21:45:

BitWrangler wrote on 2021-12-31, 17:56:

First step, try a battery on it.

However there's a couple of things I'm not happy with, from the getting an unknown board to exhibit full functionality point of view... i) The 66 Mhz crystal implies that the board is hardwired for 33Mhz, and I don't see enough jumpers in the right areas to be convinced that's settable. Therefore a 25Mhz CPU may be overclocked, which can lead to spurious operation unless carefully tweaked and tuned, or maybe is one of the "losing tickets" i.e. a CPU that won't overclock. ii) The cache at 25ns is a bit on the slow side for 33Mhz operation also, I would prefer to see 20ns in it.. Seems like this board might have been bare at one point and someone just stuck available parts in it, whether they were ideal or not.

Wouldn't 25ns cache be good up to 40MHz?

There's usually a 74F series chip in the way adding another 7ns propagation delay. If you've got a particularly fast one of those, your 5V line is strong, the 25ns chips are the top end of the spec and the moon is waning gibbous, it might work.

alpm wrote on 2022-01-01, 20:12:

I cleaned those points and it turned on! It had stopped turning on and was in the three long beep state. Now it is in an infinite boot loop: right after the memory test the system beeps and restarts a few times and eventually ends up with "CMOS Inoperational. System halted" with serveral loud short beeps (like in a loop, 50+ times).

It's possible that the problems are, as others said, that the CPU is overclocked. As a temporary fix it might be worth trying removing the J9 jumper to put it in non-turbo mode (assuming that it is in turbo mode with J9 shorted).

There are two cracked tants near the cache block, the tants on the memory banks looks fine. Do you folks think I should try to recap the whole board? Do you have a recommendation on what type of caps I should use? Should I stick with tants or maybe replace them with ceramic or electrolyte-based?

What markings do the cracked ones have? If they're of the same value then I'd be tempted to replace any matching ones, not necessarily all of them. I'd probably also go for replacing like-with-like, rather than changing types. Also, if the board has been worked on before (you mentioned that you thought the power connector looked like it'd been resoldered) then see if it looks like any capacitors might have been removed but not replaced.

Also, without an oscilloscope, is it possible to test the xtal on pins 61-62? The voltage drop on multimeter there is of about 0.9V (across pins 61 and 62).

Ah, not sure. My limited experience is that crystals are only ever just about oscillating, so putting any kind of load on the circuit will stop it oscillating. Borrowing a crystal from another board (all the RTC crystals are the same 32.768kHz) might be worth trying. Pin 65 should be a copy of the RTC (although the datasheet says this can be turned off, presumably to save power, so maybe not). Maybe the soldering problem means the clock feedback input/output has been damaged. If that's the case then maybe an oscillator (rather than crystal) can be rigged up to pin 61.

But probably easiest to go with de-turbo / swap in lower frequency oscillator / fix capacitors, before doing anything more complicated.

snufkin wrote on 2022-01-01, 23:53:

alpm wrote on 2022-01-01, 20:12:

I cleaned those points and it turned on! It had stopped turning on and was in the three long beep state. Now it is in an infinite boot loop: right after the memory test the system beeps and restarts a few times and eventually ends up with "CMOS Inoperational. System halted" with serveral loud short beeps (like in a loop, 50+ times).

It's possible that the problems are, as others said, that the CPU is overclocked. As a temporary fix it might be worth trying removing the J9 jumper to put it in non-turbo mode (assuming that it is in turbo mode with J9 shorted).

I've just tried that. Unfortunately, it is still boot looping. I've also attached a picture I took with those USB scopes of the pins 62-63.

snufkin wrote on 2022-01-01, 23:53:

There are two cracked tants near the cache block, the tants on the memory banks looks fine. Do you folks think I should try to recap the whole board? Do you have a recommendation on what type of caps I should use? Should I stick with tants or maybe replace them with ceramic or electrolyte-based?

What markings do the cracked ones have? If they're of the same value then I'd be tempted to replace any matching ones, not necessarily all of them. I'd probably also go for replacing like-with-like, rather than changing types.

The ones that are cracked (pictures attached) are 10uF 16V.

snufkin wrote on 2022-01-01, 23:53:

Also, if the board has been worked on before (you mentioned that you thought the power connector looked like it'd been resoldered) then see if it looks like any capacitors might have been removed but not replaced.

I think they were removed (see picture attached). The other caps which are on the power connector read 10uF 25V. I've ordered replacements for all missing/cracked ones (plus all others in the board).

snufkin wrote on 2022-01-01, 23:53:

Borrowing a crystal from another board (all the RTC crystals are the same 32.768kHz) might be worth trying. Pin 65 should be a copy of the RTC (although the datasheet says this can be turned off, presumably to save power, so maybe not). Maybe the soldering problem means the clock feedback input/output has been damaged. If that's the case then maybe an oscillator (rather than crystal) can be rigged up to pin 61.

Makes sense, I will also try that!

snufkin wrote on 2022-01-01, 23:53:

But probably easiest to go with de-turbo / swap in lower frequency oscillator / fix capacitors, before doing anything more complicated.

Thank you *very much* snufkin! You're awesome 😁
I will try things on this order then:

1. De-turbo (done)
2. Swap lower frequency oscillator (ordered a 50MHz one ^K)
3. Fix capacitors (only cracked and missing ones first)
4. Borrowing a crystal from another board (all the RTC crystals are the same 32.768kHz)

I will update you folks as I go through the list. Most parts should arrive by EOW, some by the 10th 😉
Again, thank you folks a lot for the support!

I saw the picture update on https://www.ultimateretro.net/en/motherboards/2091 and felt honored! Thank you!!!

I've replaced the bad/missing caps, changed the xtal to a 25Mhz, changed the time crystal and it worked! Very very happy here.

I've already dumped the BIOS (attached) and took some pictures from NSSI.

The board has 64K of cache. Really nice to see Microcache TM in action.

Good work !

alpm wrote on 2022-01-06, 01:46:

I've replaced the bad/missing caps, changed the xtal to a 25Mhz, changed the time crystal and it worked! Very very happy here.

I've already dumped the BIOS (attached) and took some pictures from NSSI.

The board has 64K of cache. Really nice to see Microcache TM in action.

Nice. Of course, now the crystal is socketed you can try seeing if the CPU will overclock...

snufkin wrote on 2022-01-06, 17:16:

alpm wrote on 2022-01-06, 01:46:

I've replaced the bad/missing caps, changed the xtal to a 25Mhz, changed the time crystal and it worked! Very very happy here.

I've already dumped the BIOS (attached) and took some pictures from NSSI.

The board has 64K of cache. Really nice to see Microcache TM in action.

Nice. Of course, now the crystal is socketed you can try seeing if the CPU will overclock...

Good idea! Will try that by EOD! Do you think this would require a heatsink?

(I also ordered a 33Mhz AMD 386 last week, anxiously waiting for it)

alpm wrote on 2022-01-06, 17:32:

Good idea! Will try that by EOD! Do you think this would require a heatsink?

(I also ordered a 33Mhz AMD 386 last week, anxiously waiting for it)

Don't think so, at least not for a quick 'does it boot' test. Beyond that, see how hot it gets and if it's too hot to keep a finger on (~50C) then stop over clocking it. Might be interesting to roughly measure the temperature at rated speed and see how much hotter it gets when overclocked.

I my case (386SX Laptop) the RTC chip (DS12885S) on the nwX287 replacement was faulty. After replacing it with another one the error was gone.