Just as an sanity check I've replaced the MOSFETs temporarily with beefier 60v ones. But to no avail....
There's nothing else that could be wrong besides the resistors and caps and whatnot. Maybe the schottky diode but it seems to only filter out the Vcore...

Any ideas?

I think most of the "super" socket 7 boards have a dual voltage VRM that shuts down the second voltage based on a detection pin being grounded. This lets the board also support single voltage CPUs. So to test voltages without a CPU installed, you may need to ground that pin. Also, some switching power supplies don't regulate well without a load. So maybe ground a pin and add a resistor load to the output. Check the datasheet for the VRM controller chip to see what it's doing. Good luck and let us know what you find...

nickles rust wrote on 2025-05-24, 18:36:

I think most of the "super" socket 7 boards have a dual voltage VRM that shuts down the second voltage based on a detection pin being grounded. This lets the board also support single voltage CPUs. So to test voltages without a CPU installed, you may need to ground that pin. Also, some switching power supplies don't regulate well without a load. So maybe ground a pin and add a resistor load to the output. Check the datasheet for the VRM controller chip to see what it's doing. Good luck and let us know what you find...

Pulled Vcc3 to GND by an 1k resistor and got the same results of 3.54V Vcore regardless of jumpers....
Also, what is the designation of "detection pin"? Datasheets for both Pentium MMX and K6-2E don't mention any Detection pin or similar.
VRM controller seems to be an

AR7010
M6822

in SSOP8 case. No data on it at all......

Hopefully someone who fixed this board could chip in and help...

Correction. Vcc3 also has 3.54v, the same as Vcc2. (already corrected this in the post).
I also tried to desolder one leg of the Vcore inductor in hopes of cutting the power from the bad VRM to the socket in preparation of injecting Vcore, but I saw that in the socket itself there was around 0.2v leakage and the output of the VRM skyrocketed to 4.84v....

Any ideas?

Are you testing with, or without a CPU installed? Most single voltage CPUs will short both voltages together inside the CPU, independent of whatever the board is doing. Without a CPU installed there should be no direct connection between the two voltage supplies, as actual jumpers to short them together were not normally used in the later "super" socket 7 boards. With no CPU installed, you could check this condition with a multi-meter.

The pin I mentioned for detecting Vcc is usually called something like "Vcc detect" if you check the datasheets. If you power up the board with no CPU installed, 3.54v is not surprising for the I/O voltage. A poorly regulated core voltage with no load is also not surprising.

So as I mentioned, to test the VRM without a CPU, you probably need to ground that pin and add a load (something like 10-100ohm?) to the core output.

Okay, so for compatibility sake I've returned to the original MOSFETs which tested good.
I am doing testing without CPU in place (which I might've killed already...)
Now, I have tried to pull down the Vcore with 33ohm and 10ohm resistors, but all resulted in 3.54v.

Most single voltage CPUs will short both voltages together inside the CPU...

What do you mean by this? Do you mean short both voltages as in the detect pin and Vcore?

There is a VCCDET pin that needs to be shorted to ground to make the VRM output both voltages. It's floating (not connected to anything) on a single-voltage CPU and grounded on dual-voltage CPUs. 3.54V is a little high for a single-voltage Pentium (it's normally either 3.3V or 3.45V), but that might have been an intentional design decision from Epox or just the lack of a load.

The lack of an obvious 3.3V regulator is interesting. Maybe that mystery VRM controller is a combination switching/linear controller (switching for core voltage and linear for I/O voltage). Revision 1.2 of that board has a US3033, which is a combination controller like that. However, that later revision also has a linear regulator next to the RAM slots. It looks really strange, and I'd like to see a schematic or have a board in front of me to see what they're doing.

kotel wrote on 2025-05-25, 08:33:

I also tried to desolder one leg of the Vcore inductor in hopes of cutting the power from the bad VRM to the socket in preparation of injecting Vcore, but I saw that in the socket itself there was around 0.2v leakage and the output of the VRM skyrocketed to 4.84v....

That's normal. The VRM controller senses the voltage at the CPU socket, so removing the inductor makes it sense 0V (or 0.2V in your case). Then it increases the duty cycle to the maximum.

kotel wrote on 2025-05-25, 18:33:

Most single voltage CPUs will short both voltages together inside the CPU...

What do you mean by this? Do you mean short both voltages as in the detect pin and Vcore?

Single-voltage CPUs will short VCC2 and VCC3 together. I'm used to boards connecting VCC2 and VCC3 together with a FET, but this board doesn't have enough FETs to do that. However, since you're getting the same voltage on both VCC2 and VCC3, that means that something must be connecting the two together without a CPU installed. This is weird.

This is the pinout I could figure out for that weird chip.

So from my understanding, now I need to check if Vcc2 and Vcc3 are shorted, and then short Vccdet pin to GND directly (no resistors), correct?

Okay, I have tried to short Vccdet to Gnd but same behavior. Vcc3 and Vcc2 aren't shorted together.
Attaching an picture of the VRM side. No other chips are related to VRM around the ram slots. This is also an rev 0.5 board

I'm thinking there is a problem with the circuit connected to VCC2DET. Since there is no schematic, you'll just have to follow the traces to see where the feedback jumpers connect. I'm guessing that there is a small transistor (most likely a 2N7002 FET) to bypass the jumpers if VCC2DET isn't grounded, followed by a circuit to disable the 3.3V regulator that normally generates VCC3. Can you get some pictures of the VRM area on both sides of the board?

I see that you have a wire in the D15 pin on the socket. Were you checking other signals?

lti wrote on 2025-05-27, 01:08:

I'm thinking there is a problem with the circuit connected to VCC2DET. Since there is no schematic, you'll just have to follow the traces to see where the feedback jumpers connect. I'm guessing that there is a small transistor (most likely a 2N7002 FET) to bypass the jumpers if VCC2DET isn't grounded, followed by a circuit to disable the 3.3V regulator that normally generates VCC3.

Might be. But judging by the 3.3v LED on my POST card, the 3.3v is not generated onboard. With AT the 3.3v LED is off, while with ATX it is on.

lti wrote on 2025-05-27, 01:08:

Can you get some pictures of the VRM area on both sides of the board?

Yup, attaching whole board pictures and VRM (sorry for the quality, but there's no sunlight here)

lti wrote on 2025-05-27, 01:08:

I see that you have a wire in the D15 pin on the socket. Were you checking other signals?

The pin in the socket is actually in A29, where the Vccdet pin is (at least according to this(NOTE: the pinout has flipped it's numbers and letters! On my board the side where the lever is it's letter while where the "socket 7" text is it's numbers)).

All of the SOT-3 case parts were tested good (most, if not all, were 2N7002's).
Turns out, I was groudning the wrong pin (A31 instead of AL1). Now I am grounding the right pin, but the voltage doesn't change.
On Vccdet pin I get 0.6v when it's not grounded and 0.012v when it is.

Any ideas?

kotel wrote on 2025-05-27, 04:36:

NOTE: the pinout has flipped it's numbers and letters!

Pinout diagrams will usually indicate if they are showing a top or bottom view, or pin vs socket view, etc. These will be mirrors of each other. If you're looking down at an empty socket from the top, the pin you want will be in the corner near where the release lever attaches. If you have the correct pin, 0.6v does not sound right.

Do you have a datasheet for the US3033 controller? Maybe this is similar to, or even a direct replacement for the version you have?

nickles rust wrote on 2025-05-28, 12:21:

Do you have a datasheet for the US3033 controller? Maybe this is similar to, or even a direct replacement for the version you have?

US3033 is not gonna work here without major changes in the VRM. The US3033 runs on 12v and the pinouts don't match while whatever I have on mine runs at 5v.
And even if I did adapt the VRM for 12v, that'd be way harder than to (in theory):
1. desolder the 5v coil in the VRM to remove any power to it
2. desolder the Vcore coil going into that weird AR7010 controller
3. adapt the VRM by necroware to do everything on it's board and just inject Vccdet, Vcore and Vi/o on the board

but before we go the "bruteforce" way I'd like to ask, what are the normal voltage redings for Vccdet when it's not pulled low? We'll try to get to the bottom of this before we go that way.

The 12v supply to the controller chip is to drive the gates on the power FETs. Are your FETs the kind with low voltage gates? The main power conversion is usually from 5v to Vcore on these boards. Maybe the problem you have is a bad 12v input to the controller and the FETs are not switching? If you have, and can post the US3033 datasheet here, that might at least show a schematic that is close to what you have. Maybe this could explain the source of the voltages you are measuring.

I must admit that it's difficult to follow what you're doing. In the first post there was no Vcore, then there was 5v, then 3.5v. If there is no connection to Vcore without a CPU installed (and VRM disabled), then where are the voltages coming from? Maybe tracking down where these voltages are coming from would help. It sounds like you're on the right track though.

I'm actually thinking that pin 8 (the pin you marked blue) is the PWM output, and the group of parts between the AR7010 and the diode at Q12 (they used the wrong designator prefix) is the gate drive circuit.

I found the datasheet for the US3033, and it's completely different. Here's a link to one of those datasheet websites because it's really late at night here:
https://www.alldatasheet.com/datasheet-pdf/pd … SEM/US3033.html

This layout is really confusing me. I thought that the inductor connected to pin 2 of the AR7010 and pin 3 went to the caps next to the socket, meaning that it would have some kind of internal current sensing. Now I see that the inductor and the caps are connected directly on the bottom of the board. Are pins 2 and 3 shorted together, or are they just at the same potential because of the fault this board has?

Can you see where the drain and source of the FET above Q11 connect? Is the source connected to VCC3, and does the drain connect to VCC2 or 5V? It looks like the drain connects to VCC2, which is weird because that would mean that there's no regulator to generate VCC3 from an AT power supply (on an ATX-only board, VCC3 could simply be the power supply's 3.3V rail). Maybe that's the problem all along - Epox screwed up their board design on older revisions so this baby-AT board only works with ATX power supplies. I doubt it, though.

The GND and Vcc pins don't match the oneson my board. Other could match, but Vcore goes to the left "Vcore" pin on my pinout made by measurements by an coil, and the right "Vcore" goes straight to the CPU socket.

To clear things out.
At first, there was no Vcore (most likely due to me measuring the wrong pin), but there was 3.54v Vi/o after I put the jumpers into the slowest config (60mhz FSB and 2x multiplier).
Then I put my CPU in the socket, and got no POST codes. Then I removed it and put the jumpers back as they were when I got the board (100mhz FSB and 4x multiplier). I measured the Vcore on the coil and got 3.54v. After reflowing some of the resistors and caps, I started probing the second coil and I found 5v on it. It is tied to the lower MOSFET in the VRM (right next to that schottky diode). Then, just to see if I could safely inject 2.1v Vcore (without the coil that connects Vcore generated by VRM) I desoldered that coil. I then measured 0.2v in the socket on Vcore while the VRM (without being connected to the socket) skyrocketed to 4.85v.
And now, We're stuck here trying to see what's the cause.

Hope this clears up the mess I made before.

lti wrote on 2025-05-29, 04:19:

I'm actually thinking that pin 8 (the pin you marked blue) is the PWM output, and the group of parts between the AR7010 and the […]
Show full quote

I'm actually thinking that pin 8 (the pin you marked blue) is the PWM output, and the group of parts between the AR7010 and the diode at Q12 (they used the wrong designator prefix) is the gate drive circuit.

I found the datasheet for the US3033, and it's completely different. Here's a link to one of those datasheet websites because it's really late at night here:
https://www.alldatasheet.com/datasheet-pdf/pd … SEM/US3033.html

This layout is really confusing me. I thought that the inductor connected to pin 2 of the AR7010 and pin 3 went to the caps next to the socket, meaning that it would have some kind of internal current sensing. Now I see that the inductor and the caps are connected directly on the bottom of the board. Are pins 2 and 3 shorted together, or are they just at the same potential because of the fault this board has?

Can you see where the drain and source of the FET above Q11 connect? Is the source connected to VCC3, and does the drain connect to VCC2 or 5V? It looks like the drain connects to VCC2, which is weird because that would mean that there's no regulator to generate VCC3 from an AT power supply (on an ATX-only board, VCC3 could simply be the power supply's 3.3V rail). Maybe that's the problem all along - Epox screwed up their board design on older revisions so this baby-AT board only works with ATX power supplies. I doubt it, though.

Both Vcore pins on my AR7010 are connected together (0 ohms).
As for the FET above Q11, drain goes to Vcore, gate to 12v and source to Vi/o
For the ATX part, the board does the same thing whether I use an AT or ATX PSU. But, according to my POST card, the 3.3v is missing when I use an AT PSU, meaning there's no onboard 3.3v generation.

That is entirely possible that it may need the ATX power supply for 3.3 volts.

It could be some sort of limitation that if you use the AT plug you are limited to only single voltage CPU's. Which historically would make sense, how new of a cpu would you likely be using if you still had an ancient AT PSU.

I'd ALWAYS use the ATX if there is that option, dont even bother with the AT, HOWEVER, make sure that your power supply actually has the -12v wire, later supplies often lack this and it can make weird stuff happen