Your board has the block of four jumpers to the right of the 3.3V/4V select jumper. Boards with a jumper block like this typically do not support auto-voltage using the VOLTDET pin.

What you call "VRM" is in fact just a single NPN transistor. It's one of the important parts of the regulator, though. The middle pin ("collector") is connected to the 5V rail of your power supply, the right pin is the output ("emitter") and connected to the Vcc pins of the processor. The left pin is the control pin and is typically 0.5 to 0.7 volts above the output voltage when the VRM is operating.

It looks indeed like the transistor is bad. The left pin (near the ISA slot) should never be more than 0.8V higher than the right pin (near the 3.3V/4V jumper), except if an exceptionally high current is going to the left pin. I am very confident that the circuit on this board doesn't let high currents get near the left pin, and even if it did, the pin would likely be at a slightly lower voltage than the central pin. If the transistor is just overloaded, I would expect 2.8 to 3V on the left pin when the right pin is at 2.0 to 2.2V.

On the other hand, if I misunderstood what you said about left and right pins, and the pin in the middle and the pin near the 3.3V/4V jumper are both at +5V, whereas the pin near the ISA slots is at 2.2V, this would indicate that 5V is getting to Vcc on a different path than through the transistor (and thus "overrides" the VRM), and the regulator circuit in response pulls the control pin down to 2V, because the output voltage is way higher than the requested voltage of 3.3V or 4V. Two possible sources of "shorting out" the regulator are other broken components on the board, or accidentally not putting the 3.3V/5V-block into the 3.3V position.

mkarcher wrote on 2021-03-25, 14:20:

Your board has the block of four jumpers to the right of the 3.3V/4V select jumper. Boards with a jumper block like this typical […]
Show full quote

Your board has the block of four jumpers to the right of the 3.3V/4V select jumper. Boards with a jumper block like this typically do not support auto-voltage using the VOLTDET pin.

What you call "VRM" is in fact just a single NPN transistor. It's one of the important parts of the regulator, though. The middle pin ("collector") is connected to the 5V rail of your power supply, the right pin is the output ("emitter") and connected to the Vcc pins of the processor. The left pin is the control pin and is typically 0.5 to 0.7 volts above the output voltage when the VRM is operating.

It looks indeed like the transistor is bad. The left pin (near the ISA slot) should never be more than 0.8V higher than the right pin (near the 3.3V/4V jumper), except if an exceptionally high current is going to the left pin. I am very confident that the circuit on this board doesn't let high currents get near the left pin, and even if it did, the pin would likely be at a slightly lower voltage than the central pin. If the transistor is just overloaded, I would expect 2.8 to 3V on the left pin when the right pin is at 2.0 to 2.2V.

On the other hand, if I misunderstood what you said about left and right pins, and the pin in the middle and the pin near the 3.3V/4V jumper are both at +5V, whereas the pin near the ISA slots is at 2.2V, this would indicate that 5V is getting to Vcc on a different path than through the transistor (and thus "overrides" the VRM), and the regulator circuit in response pulls the control pin down to 2V, because the output voltage is way higher than the requested voltage of 3.3V or 4V. Two possible sources of "shorting out" the regulator are other broken components on the board, or accidentally not putting the 3.3V/5V-block into the 3.3V position.

You were right with your original assumption. With the Cyrix chip installed, the pin nearest the ISA slots and the middle pin were both roughy 5v, whereas the pin closest to the jumper block was just above 2v.

paradigital wrote on 2021-03-25, 16:24:

You were right with your original assumption. With the Cyrix chip installed, the pin nearest the ISA slots and the middle pin were both roughy 5v, whereas the pin closest to the jumper block was just above 2v.

Please check that I didn't mix up the pinout: The right pin of that transistor is supposed to be directly connected to the processor VCC pins. If that is indeed the case, you definitely need a replacement:

You need a NPN transistor, from pictures in the Internet of that board I guess insulated case is preferred, and it should handle 3A to 4A collector current to be on the safe side. In Europe, the BD131 or the BD175 provide a good substitute. Make sure you install the replacement with the pin called "base" near the ISA slots. If I understand the datasheet of the suggested transistor correctly, that means you install it in the orientation that you can read the marking, but you better double-check. You can easily identify the base pin using your meter. Switch it to diode test mode, touch the suspected "base" pin with the red lead. If you touch either of the two other pins with the black lead, the display should display something around 650. If it doesn't, either you didn't find the base pin, or the transistor is broken. This only works reliably if the transistor is not soldered into a circuit.

If I understand the behaviour of the broken transistor correctly, it doesn't hurt to leave it in. The two transistor spots are completely connected, I suppose (so continuity between the left pins, between the center pins and between the right pins). I can easily see the continuity at the left at center pins from mainboard photos, but the trace for the right pin is likely on the bottom side of the board. If the spots are indeed completely parallel, there can't be a need for extra components to support the second transistor, and adding a replacement into that sport should work immediately.

Finally got around to fitting the two BD131 NPN transistors that I ordered. CPU will now happily boot at 3.3v, even when running 40MHz FSB for a 160MHz CPU clock.

Different issue now appears to be that setting LBA mode on either of the IDE channels results in the board hanging during IDE detection, which limits me to 504MB in DOS 6.22 🙁

You can always use XT-IDE bios on an ethernet card to bypass any BIOS limitations.

Also Disk overlay software tends to work OK, but I dislike it because it often prevents you accessing the drive safely on other systems.

canthearu wrote on 2021-04-07, 14:36:

You can always use XT-IDE bios on an ethernet card to bypass any BIOS limitations.

Also Disk overlay software tends to work OK, but I dislike it because it often prevents you accessing the drive safely on other systems.

Completely forgot about that, time to dig out a 3com and the minipro! Thanks!