uniQ wrote on Today, 10:01:

However, if I completely disable the cache or use the Turbo button to slow down the CPU, Raptor runs perfectly fine.

Clean the legs and deoxit the sockets.

Do you have spare cache chips?
Looks like one is failing. You could also try less cache (IDK if you have 128 or less) and see what works to pinpoint the culprit.
---

Faulty cache chips alone won't explain why it's working if I just reduce the CPU clock or am I missing something?

Not sure if I have spare cache chips lying around, but I'll check.

Did you disable the cache?
Did you check the 12v and 5V rails?

Anonymous Coward wrote on Today, 11:56:

Did you disable the cache?
Did you check the 12v and 5V rails?

I did and it fixes the problem but so does reducing the CPU clock.
I can't check the rails, since I'm lacking the tools to do it. But the PSU works perfectly fine with other 386/486 boards, so I guess it's fine.
Another game I've just tested is Halloween Harry:

Turbo ON & 256K L2 Cache ON = Crashes instantly

Turbo ON & 256K L2 Cache OFF = Works, but too slow to enjoy
Turbo OFF & 256K L2 Cache ON = Works, but too slow to enjoy

Doom doesn’t start at all, no matter what settings I use. It just black-screens.

Could you try a vanilla DOS install?
Maybe a mistake in drivers for memory management?

Can you check the frequencies with an oscilloscope? A wrong frequency can have adverse effects. All crystals should be spot on.
OSC1, OSC2, Y1 ---> 80MHz, 14,318... MHz, 32KHz.
The 32KHz should be connected to U1 chip (guessing). When you enter bios, does it count time faster or slower compared to a watch?

Also ISA B20 pin (https://upload.wikimedia.org/wikipedia/common … SA_Bus_pins.svg)

Are you sure the BIOS is ok? Could be some bit rot happening. Try a reflash.

-- these are just ideas.

Btw, could you post pics?
How many cache chips do you have in what config? Is tag ram the right size?

I already tried a vanilla DOS install without any altering what so ever but the problem persisted.
I can reflash the BIOS an recheck. I'll also capture some video footage or picture if it helps.

Nexxen wrote on 51 minutes ago:

Could you try a vanilla DOS install? Maybe a mistake in drivers for memory management? […]
Show full quote

Could you try a vanilla DOS install?
Maybe a mistake in drivers for memory management?

Can you check the frequencies with an oscilloscope? A wrong frequency can have adverse effects. All crystals should be spot on.
OSC1, OSC2, Y1 ---> 80MHz, 14,318... MHz, 32KHz.
The 32KHz should be connected to U1 chip (guessing). When you enter bios, does it count time faster or slower compared to a watch?

Also ISA B20 pin (https://upload.wikimedia.org/wikipedia/common … SA_Bus_pins.svg)

Are you sure the BIOS is ok? Could be some bit rot happening. Try a reflash.

-- these are just ideas.

Btw, could you post pics?
How many cache chips do you have in what config? Is tag ram the right size?

I don't have an oscilloscope or even a multimeter. 😒

But I did check the clock in BIOS and it's a tiny bit fast that a normal watch.
Really not much but you can see that it jumps too fast every 3-4 seconds.

Here are some pictures from the BIOS settings, cache chips and PSU.
As mentioned in my first post, I already tested RAM and Cache timings but it didn't make a difference.

Could the CR2032 holder with diode be the culprit here? It replaced the Varta barrel battery.

uniQ wrote on 17 minutes ago:

Could the CR2032 holder with diode be the culprit here? It replaced the Varta barrel battery.

IIRC it would deplete the battery faster, not having any influence over the BIOS.
You can replace it live, you don't need to remove it while shut down. Don't use screwdriver though.

Every 3-4 seconds it looses time? 10 seconds are indeed 10 seconds, not caring for display timing?

Could be the 32KHz crystal. If the frequency wave that is outputted by the board isn't good (too much ripple) it could have adverse effects on the system stability overall like you described.
This could happen when the waves from sinusoidal to square are too bad to work correctly with other components (timings matter).

I've topped my knowledge, to troubleshoot you need more material like a multimeter, oscilloscope...
Unless someone who actually knows what is going on appears and teaches us 😀

This looks totally "rapairable" to me. Just boring 😀

Btw, would you test with just one row of cache? TAG + Bank 0, then Bank 1 (if it works) and try moving around chips until it doesn't fail.
I told you it's boring ....