I'm not 100% sure but yes, these mobos have 2 clock sources and seem to switch between them. It makes less sense for 12->6 MHz switch, that could be done with /2 divider (with is just one flip-flop) but having two sources allow for more complex schemes, like 10MHz CPU and 6MHz de-turbo mode. That being said the mobo manufacturers quickly realized a 10MHz rated CPU will also work pretty well at 12MHz, and all this overclock needs is a simple crystal swap.

Anyway, you should have 2 crystals near 001 chip, one 12MHz (which seems to work fine if you get reliable 6MHz operation) and 24MHz one. On my mobo this one is called "25M/20M" on the solder mask. I guess even the 12MHz variant was to be overclocked to 12.5 by default. So this one I would assume is not working for you. Check it with a scope (use 10:1 probe), inspect the nearby capacitors, and the parallel resistor - in this circuit the resistor is pretty low value and quite important. I never checked but my expectation would be both of these clocks must be running all the time, rather then being switched on-demand.

Deunan wrote on 2025-08-05, 13:48:

I'm not 100% sure but yes, these mobos have 2 clock sources and seem to switch between them. It makes less sense for 12->6 MHz switch, that could be done with /2 divider (with is just one flip-flop) but having two sources allow for more complex schemes, like 10MHz CPU and 6MHz de-turbo mode. That being said the mobo manufacturers quickly realized a 10MHz rated CPU will also work pretty well at 12MHz, and all this overclock needs is a simple crystal swap.

Anyway, you should have 2 crystals near 001 chip, one 12MHz (which seems to work fine if you get reliable 6MHz operation) and 24MHz one. On my mobo this one is called "25M/20M" on the solder mask. I guess even the 12MHz variant was to be overclocked to 12.5 by default. So this one I would assume is not working for you. Check it with a scope (use 10:1 probe), inspect the nearby capacitors, and the parallel resistor - in this circuit the resistor is pretty low value and quite important. I never checked but my expectation would be both of these clocks must be running all the time, rather then being switched on-demand.

Thanks for the information. My Suntac 286 motherboard (you can see picture of it in this post) has only one active crystal oscillator (24.000 MHz rectangular one with 4 legs near Suntac 003 chip). I don't see a second rectangular active crystal oscillator for 12.000 MHz or similar value. There are also two passive cryctals with two legs for 14.31818 MHz (near 24.000 MHz active one and Suntac 003 chip) and for 16.000 MHz (near Suntac 001 chip).

If I'm not mistaken, the KB-controller chip plays some role in "turbo" mode switching.

P.S. I think, I found the problem. The resistor R16 near the active crystal oscillator is broken/partially missing. And I don't know it's value 🙁

So the passive (resonator) 12MHz one near 001 is the "standard" Suntac design, along with 2x 10pF caps and 33k resistor. Since the 24MHz active oscillator is the only thing that could drive the CPU at 12MHz it must be it. I'm not familiar with this design but it can't be much different. That clock source should also go to the 001 chip because that is what drives both the CPU and the 287 NPU. I don't think 001 can be bypassed since it needs to decode 286 Sx signals and also handle DMA req/ack, all these must be synchronous to CPU clock.

So here's my educated guess looking at your mobo: It does have a much simpler /2 divider, possibly selected by one of the KBC pins as you suspect. The 74F112 is dual flip-flip, just what you need for the clock divisor. The F10 could be selecting the clock output, along with perhaps the 04 to invert the selection signal so that only one of the F10 gates is enabled. That fits well with the selection signal being slow so you don't need F for the 04, LS is good enough - and at the same time not quick enough to act as the fan-out driver. So I guess the third F10 gate is used for that.

So first check if that 24MHz can is working and producing viable signal. Doesn't need to be square as long as it meets TTL specs. Then, since you have some clock on the CPU, check which of the F chips is connected to the 001, I belive pins 31 or 32. Start with the F10. You might want to draw the whole 04/10/112 thing to figure it out and the connections. 112 should work if the de-turbo mode works, that is the divider part is doing its job, so I guess one of the F10 gates is busted?
You might also want to check what is connected to pin 10 of 001, that's the turbo/clock switch input. Might be permanently tied to GND/VCC on your mobo or to some jumper.

I think I found the problem. The resistor R16 near the 24.000 MHz active crystal oscillator is broken/partially missing. I don't know its original value.

This one is from the picture of the closest board to mine. Is it a 15 Ohm ±5% resistor?

analog_programmer wrote on 2025-08-05, 15:01:

This one is from the picture of the closest board to mine. Is it a 15 Ohm ±5% resistor?

That's a good value, yes. It can't be too big as it would create too much of a low-pass filter with all the stray capacitance. And much lower than that (say 1.5) just makes no sense, a thin trace would do the same thing. The only thing is does here is reduce over/under shoot from the crystal to the chip inputs. Long term operation with these spikes could affect reliability of the chips, and possibly the crystal itself too. But short term, for a few hours for testing, you can just short that and see if it works. There should not be so much noise here that a short would not work work, but it might not be very stable, so if the system hangs or something it might just be the culprit.

Deunan wrote on 2025-08-05, 17:20:

That's a good value, yes. It can't be too big as it would create too much of a low-pass filter with all the stray capacitance. And much lower than that (say 1.5) just makes no sense, a thin trace would do the same thing. The only thing is does here is reduce over/under shoot from the crystal to the chip inputs. Long term operation with these spikes could affect reliability of the chips, and possibly the crystal itself too. But short term, for a few hours for testing, you can just short that and see if it works. There should not be so much noise here that a short would not work work, but it might not be very stable, so if the system hangs or something it might just be the culprit.

Thank you for pointing me to this important resistor. In my case it's R16. One of its legs is connected to the oscillator, the other - to one leg of the 74F112PC chip. It turned out that the resistor is broken in two halves with missing part of its paint finish and one of the color code strips. I desoldered it to measure its value out of the circuit and it fell apart 🙁 Now I have no clue what its real value was, except the color code strips on R16 from the picture of the motherboard, which is closest to my model. I see the strips for color code from the picture as brown, green, black, gold(?) and some online resistor color code decoder gives me 15 Ohm value. I don't want to short the resistor through holes. I'll search for at least 10 Ohm resistor in my spare parts box, if I can't find a 15 Ohm replacement.

Now I'm wandering how this board managed to work with this R16 broken (cut-off 24.000 MHz oscillator signal) in "deturbo" mode only. From where the CPU got 6 MHz signal?

analog_programmer wrote on 2025-08-05, 18:13:

Now I'm wandering how this board managed to work with this R16 broken (cut-off 24.000 MHz oscillator signal) in "deturbo" mode only. From where the CPU got 6 MHz signal?

Now I think your mobo has 3 modes of operation, except 2 result in the same CPU frequency. So most of these mobos have a jumper that selects what controls the turbo - the case switch or the DIP. I assume yours still has this setup and pins 10 (and 13) of 001 are connected to the select jumper. In that case you can set the speed with DIP-3 usually. This will force 001 to use the 12MHz source, thus resulting in 6MHz CPU clock - the 286 and 386 both require the clock input to be double of the operating frequency. It's inherited from older Intel chips that had external multi-phase clock generator chip. If you try to switch the DIP to higher clock the mobo will not run because that signal is now missing.

I assume that in the turbo mode, when the faster clock is the source for 001, this can be externally divided by 2 - this way you can have software control over it. Why do it this way and not simply route the signal from KBC to 001 to switch clocks, I have no idea. Perhaps the 001 needs that 12MHz clock for other things as well, so it just has to be there. And in that case it was easier to add an extra /2 divider externally and leave half of the original switching system in place. At least that's my current theory, you'd need to trace connections to pin 10 and 13 for 001 chip to be sure.

As for the resistor, it's really not that critical. Sure 15 ohms probably gives the best spike suppresion without degrading the shape too much, but anything from 0 to say 33 or perhaps even 51 ohms should work.

Deunan wrote on 2025-08-05, 20:03:

Now I think your mobo has 3 modes of operation, except 2 result in the same CPU frequency. So most of these mobos have a jumper that selects what controls the turbo - the case switch or the DIP. I assume yours still has this setup and pins 10 (and 13) of 001 are connected to the select jumper. In that case you can set the speed with DIP-3 usually. This will force 001 to use the 12MHz source, thus resulting in 6MHz CPU clock - the 286 and 386 both require the clock input to be double of the operating frequency. It's inherited from older Intel chips that had external multi-phase clock generator chip. If you try to switch the DIP to higher clock the mobo will not run because that signal is now missing.

I assume that in the turbo mode, when the faster clock is the source for 001, this can be externally divided by 2 - this way you can have software control over it. Why do it this way and not simply route the signal from KBC to 001 to switch clocks, I have no idea. Perhaps the 001 needs that 12MHz clock for other things as well, so it just has to be there. And in that case it was easier to add an extra /2 divider externally and leave half of the original switching system in place. At least that's my current theory, you'd need to trace connections to pin 10 and 13 for 001 chip to be sure.

As for the resistor, it's really not that critical. Sure 15 ohms probably gives the best spike suppresion without degrading the shape too much, but anything from 0 to say 33 or perhaps even 51 ohms should work.

You can see here what I managed to trace for all the jumpers and DSP switches. And there's no duplicated "turbo" function on the DSP1 for my board. The situation with cut-off CPU oscillator signal (broken R16) is really weird and still I can't explain how this board worked in "deturbo" mode only.

I found two 7.5 Ohm resistors and patched things up 'til I get a proper 15 Ohm resistor. The "turbo" button finally woks as expected 😀 Once again, thanks for pointing me to check the damaged resistor R16!

Oh, I didn't notice that. And I can't quite make out the details from the photo, but there does seem to be a TUB SEL jumper? Perhaps it's just 2-pin one then? In that case it'll permanently disable the turbo function and switch the mobo into always 6MHz CPU speed mode, again hard to argue without actually knowing connections to pin 10 and 13 for 001 chip. But I'm reasonably sure it's how it works, the 12MHz crystal near 001 was the only working source of CPU clock before the resistor fix, so that must be it.

Deunan wrote on 2025-08-06, 08:56:

Oh, I didn't notice that. And I can't quite make out the details from the photo, but there does seem to be a TUB SEL jumper? Perhaps it's just 2-pin one then? In that case it'll permanently disable the turbo function and switch the mobo into always 6MHz CPU speed mode, again hard to argue without actually knowing connections to pin 10 and 13 for 001 chip. But I'm reasonably sure it's how it works, the 12MHz crystal near 001 was the only working source of CPU clock before the resistor fix, so that must be it.

Yep, 2-pin "TUB jumper" is the connector for "turbo" button". Now, when "TUB" is shorted by the "turbo" button, the mobo finally works in 12 MHz "turbo" mode 😀 I'm not sure if the motherboard was in 6 MHz mode with broken R16, but the system was too slow - just as for "deturbo" 6 MHz mode. So, probably you are right, that the only source for some frequency signal provided to CPU was 16.000 MHz crystal resonator (divided by 2) located near Suntac oo1 chip. 8 Mhz is not much faster than 6 Mhz 😉

I assume all hardware "turbo" mysteries with this "no name/unknown" 286 Suntac motherboard are now solved.

Still I have no clue how KBC "turbo" function woks for these 286 boards. Excuse my stupid question, but when I need to change the KBC "turbo" jumper for AMI, Award, Phoenix or MR BIOSes - when the BIOS is changed or when the KBC controller is changed? I'm not sure, but I think the KB "turbo" switching (<Ctrl> + <Alt> + <+>/<->) is more of a BIOS related thing, not so KBC related thing (except for KBC pin connections).

I doubt the KBC has this turbo switch as independent functionality. There are already commands that system can send to KBC to set some of its I/O pins to certain value - this is more or less how A20 gating works. So the easiest way is to take a standard KBC, wire one more free I/O to the clock divider/mux, and let BIOS or other software control it with already existing KBC commands. This way you can use any KBC that's AT compatible and it'll still work, rather than having to come up with a very proprietary microcode version for that specific mobo.

I really don't know how this works, but on my 286 Suntac board one of the KBC "turbo" selector (missing) jumpers is hardwired/shorted by a trace. I think it's for Award setting - according to the original Award BIOS. I've just wondered why the heck are those KBC "turbo" jumpers - in my case they're "unpopulated" (with missing pin-headers).

Some old boards had a turbo hot key.
To do that, there would be some logic somewhere that opened/closed a circuit.
Most likely it was in the kbc logic since it would be the first thing intercepting keyboard commands. So having a circuit there would make sense. Any generic gpio would do.

myne wrote on 2025-08-07, 01:43:

Some old boards had a turbo hot key.

Do you mean keyboard combinations like "<Ctrl> + <Alt> + <+>/<->"?

myne wrote on 2025-08-07, 01:43:

To do that, there would be some logic somewhere that opened/closed a circuit.
Most likely it was in the kbc logic since it would be the first thing intercepting keyboard commands. So having a circuit there would make sense. Any generic gpio would do.

I'm just wondering when those jumpers have to be changed. For example now my board is using very basic Award BIOS, but what if I change it to AMI BIOS? Do I have to change those jumpers as well, if I want to still have working "turbo" mode switching keyboard combination?