I have junk AT PSU which only outputs:

5V cable: 4.84 V
-5V cable: -4.64 V
12V cable: 11.65 V
-12V cable: -11.17 V

however, the Power Good cable stays at GND, indicating that the voltage levels are not in spec, thus not letting the motherboard turn on. What did your other voltages look like? If they were as bad as mine, your MB shouldn't be powering up.

Side note: caps aren't bulging. Anyone know which components need replacing in this AT PSU (caps, regulator, inductor)?

I just recently acquired one of these CPUs and decided to push it to 100MHz also.
I must say I have not read through this entire thread so if someone else has already documented this, I apologise.

Firstly let me clear up some misconceptions regarding the voltage regulator vs the motherboard's voltage regulator.

On Socket 3 there are two sets of VCC pins, one set is VCC5, the other is VCC3. VCC5 is always connected to the +5VDC directly, where VCC3 is provided via a regulator (adjustable or not) from the motherboard.

The POD83 internally bridges these two sets of pins, bypassing the onboard regulator. You can prove this simply by measuring for continuity between +5VDC and the motherboards regulator output. When you insert the CPU, you will find a short immediately develops between them. This explains why early in this thread, it was mentioned that shorting the in-out pins on the regulator resulted in ~4.9V.

Regarding the diode mod here, I really can't understand why this is in use. The onboard regulator is clearly a 3 pin fixed voltage regulator. The only difference between a fixed and variable regulator is the resistors that set the voltage on the ADJ (adjust pin) are internal to the package. You can always adjust the voltage simply by adding an external resistor between the ground pin and ground. Conveniently this pin is the one right on the edge of the CPU package.

So how I modded my CPU:

1. Solder a wire onto the middle pin of the regulator so I can probe the voltage it's running at with a multimeter (this is the white wire).
2. Lifted the regulator pin closest to the edge of the CPU
3. Soldered a multi-turn 1k potentiometer between the lifted pin and the pad on the CPU package

Using this I am able to dial in an exact voltage, which does not drift under load as the regulator is simply performing it's intended function.
Raising my CPU from the factory 3.5v to 3.6v has made it completely stable at 100MHz.

I wasn't aware it was so simple to mod a fix-voltage VRM. The diode mod started as a quick experiment, and since it worked well, we just stuck to it. It delivers around 4.15 V and so far, we haven't run into a POD that couldn't do 100 MHz at 4.15 V.

I guess you'd adjust your trim pot's size based on what internal resistors were used. With 1 K-ohm, what voltage range are you allowed? What software did you run to determine stability at 3.60 V?

I have not done any extensive testing yet, I did however have to bump it up to 3.65V as I was getting the odd invalid opcode issue launching Windows 3.11.
I will work on getting some more software onto the system so I can properly test it and verify.

As for voltage range, 3.50 through to 4.0v seems doable, any more and I think it's hitting the limits of the regulator, or I need a higher value pot (it's hard to tell when you hit the end of a multi turn pot).

I think you might find that you don't need such a high voltage if you use this method because the voltage is stable and doesn't vary with the drop in the diode junction under different loads. More testing is needed to verify though.

Happy to report that it's solid stable at 3.65V, verified by running Quake timedemo in a loop for several hours.
Watching the CPU voltage, it dips to 3.64V while under a heavy load.

Edit: Ok, no it wasn't stable, once I went to a higher resolution of the timedemo it couldn't complete a run.

Watching the voltage over the longer run, it drops and continues to drop until it goes too low and crashes. The problem is the regulator is getting too hot and self-limiting to protect itself, a common behaviour in linear regulators. I found that if I pushed a thermal pad into the gap to allow the regulator to conduct it's heat directly into the heat sink above it, it helped, and I am now able to complete the quake time demo at 640x480.

While I have it working now, I think I might investigate building a off-board regulator that replaces the one on the CPU, I am curious to see how low the voltage can be if the rail is stable.

I find running numerous Windows 9x apps to be more telling in terms of stability than DOS Quake.

At what decreasing voltage did the CPU crash? How hot did the regulator get?

At the moment this is a pure DOS system, no 9x yet 😀

I need to attach a scope to monitor the voltage in higher resolution, but it seems if it dips below 3.64V while loaded it crashes.

As for temperature, not sure, I will attach a thermocouple when I get some time to get back to this. Suffice to say though that after adding the thermal pad, the factory heat-sink is substantially warmer in that area.

I also tested adding a 3300uF low ESR capacitor across the caps on the CPU, which did improve the voltage stability but did not help with the low voltage transient as the regulator gets hot.

feipoa wrote on 2025-09-22, 10:35:

I wasn't aware it was so simple to mod a fix-voltage VRM.

It depends on the regulator chip whether this is working well or not. In case of 486/Pentium linear regulators, it typically is that simple. If for a given family of voltage regulators, there is an adjustable model (like LM317, LM1086, LM1117 & clones, I didn't bother looking up whether the LM (National Semiconductors) or LT (Linear Technology) variants were first for 1086 and 1117 ), the "fixed voltage" chips indeed just have the voltage divider included, and modding it this way is possible.

I'd stay clear of this type of mod on the 7805-type regulators, though. The idea of this mod is to use the resistor between the ground/adj pin and actual ground to lift up the ground of the regulator to increase the voltage by a given amount. This requires the voltage drop on the resistor to be approximately constant, which means the current through that resistor needs to be approximately constant, independent of the load current. Regulators like the LM317 are supposed to be lifted from their "base voltage" (2.5V, IIRC) to the desired voltage using a voltage divider, so their current through the ground/ADJ pin is sufficiently constant by design. It's not perfectly constant, and that's why they require the divider to have sufficiently low resistance, so the current through that divider is way bigger than the variation of the ground/ADJ current of the regulator. Adding a resistor like this increases the sensitivity to ground current, and if you push it too far, the result might be that the output voltage increases slightly if the regulator gets hot (more leakage current) or the load current increases. At the scale of this thread (upping the output by up to 20% in an application that doesn't require precision in excess of 0.05V), the added ground impedance is extremely unlikely to be a concern, so don't take this explanation as a recommendation to avoid this circuit - just don't use the same approach to mod a LM1086-3.3 to output 24V instead. This might require a sufficiently high "mod resistor" that bad things (worst case: oscillations) start to happen.

On a 7805-type regulator, the ground current is not designed to be constant, but it basically consists of a quiescent current and a contribution roughly proportional to the load current. Using a resistor to lift it up will provide very poor results. If you want to lift a 7805 to 8 volts, use a 3V Z-Diode instead, which provides a way more constant voltage at varying load currents than a resistor. It will not be a precision device, but for "anything between 7.5 and 8.5V will do, and I don't want to add an 7808 to my BOM" the Z-Diode approach is good enough, but the resistor approach is not.

The VRM on my POD is an LT912CM. I wasn't able to find a datasheet, so I have no idea how this fits in to the VRM examples referenced by mkarcher. However, mkarcher's write up fired a tiny cylinder in my memory banks, for which I must correct my previous statement. Before I went the path of the 6A diode, I tried exactly what gnif is doing. Here is a direct link to that 10 year old sub-thread: Re: Modifying the POD83's voltage regulator for overclocking And a copy/paste from that sub-thread is as follows in bold:

I have determined that the VRM on the POD83 is normally fixed at 3.50 V. The 4 SMD components on the cermaic package are all 3 uF capacitors. To increase the voltage from 3.50 V using the POD's voltage regulator, splice in a variable resistor (trimmer) in series on the VRM's GND pin. I used a 1K trim pot. This allows voltage adjustment from 3.50 - 5.0 V. The problem is that the POD's VRM can only output so much current, which seems to limit the coltage increase to 3.73 V on my system (at load). Even if you increase the voltage to 3.8+ V, under load, the voltage drops to 3.70-3.73 V. Note that the red clip in the photo is so that I can measure the voltage on the VRM's centre pin.

For this modification, I selected one of my POD chips which doesn't quite cut it at 100 MHz. Before the mod, DOS Quake would not load. After upping the voltage to 3.73 V, DOS Quake loads and begins the timedemo, however it crashes shortly thereafter. So this mod may work for you if you only need an extra 0.1 to 0.2 V. If you need more voltage, it might be best to change the POD's VRM entirely, preferably to something which can output up to 5A. Note that the value of the trimmer is approx. 30-ohms with a 3.73 V output.

In short, I couldn't get the VRM to output enough current for the desired set voltage. Later on, I elected to mod the POD with a 6-A series diode here: Re: Modifying the POD83's voltage regulator for overclocking

It may be that gnif's POD sample only needs an extra 0.15 - 0.2 V to be stable, in which case, the VRM-resistor mod may work well enough. I suggest more testing in Windows 9x, especially during the summer heat, before settling on any mod.

How high of voltage you can run your POD at under load when the VRM's casing is being cooled by the heatsink? For example, if you set the VRM to 4 V, or 3.9 V, then run it under load (Quake) for a few minutes, does your VRM's output also drop to about 3.7 V?

feipoa wrote on 2025-09-23, 07:47:

The VRM on my POD is an LT912CM. I wasn't able to find a datasheet, so I have no idea how this fits in to the VRM examples referenced by mkarcher.

The assumption that this is very similar to a fixed-voltage LT1086 makes sense. You need a low drop-out (LDO) regulator at that point. I suppose the LT1084..LT1086 series is what evolved as generic LDO from customer-specific designs like this one.

feipoa wrote on 2025-09-23, 07:47:

The problem is that the POD's VRM can only output so much current, which seems to limit the coltage increase to 3.73 V on my system (at load). Even if you increase the voltage to 3.8+ V, under load, the voltage drops to 3.70-3.73 V. Note that the red clip in the photo is so that I can measure the voltage on the VRM's centre pin.

As we know now, cooling helps, so we hit the thermal limit. Actually, increasing the voltage from 3.5V to 3.8V is suppose to put less stress on the regulator, as it no longer needs to dissipate 1.5 Volts as heat, but just 1.2 Volts. This is a reduction to 80% of the original voltage drop.The fact that we still see thermal issues indicates that the reduction in voltage is more than compensated by an increase in current. A first order assumption is that the current consumed by a processor scales linearly with the voltage (twice the voltage moves twice as much charge in and out the capacitance of all the FETs inside a CMOS part on each switching operation) and also linearly with the frequency. (running at twice the frequency needs to move the charge twice as often). So going from 3.5 to 3.8 volts should just increase the current by 8%, but increasing the FSB from 33MHz to 40MHz at the same time will increase the current by another 20%. Together, the increase of the current to about 130% in fact more than compensates the voltage drop to 80% and explains the increased power dissipation required by the regulator - and this does not yet factor in that the POD gets hotter at 100MHz than it does at 83MHz, which will increase leakage currents and cause even higher power usage, while it will worsen the heat dissipation from the LT912.

gnif wrote on 2025-09-22, 17:00:

I found that if I pushed a thermal pad into the gap to allow the regulator to conduct it's heat directly into the heat sink above it, it helped, and I am now able to complete the quake time demo at 640x480.

Using the themal pad on that regulator in addition to the resistor lifting the ground of the LT912 using a resistor in fact sounds like the most elegant solution for overclocking the POD.

mkarcher wrote on 2025-09-23, 10:25:

Using the themal pad on that regulator in addition to the resistor lifting the ground of the LT912 using a resistor in fact sounds like the most elegant solution for overclocking the POD.

I agree, however only if you have a POD (at 100 MHz) which do not require more voltage/current than can be provided by the VRM. Do they all have LT912's? I am hoping that cooling the VRM with a thermal pad may unlock up to 4.0 V. Ideally, you'd first run your POD with the trimmer to determine minimum voltage needed for stability, then you'd swap it out for a fixed resistor to be even more tidy.

On the other hand, if the VRM is adding too much heat to the POD, maybe we can squeeze in under the heatsink one of those tiny 2A or 3A buck regulators. EDIT: How much current does the POD need at 100 MHz and 4 V? 3.5 A? Is the LT912 a 3A LDO?