Published: Thu 23 October 2025
By PluMGMK
In Technology .
Feast your eyes on this contraption:
This, dear readers, is the result of a half-hour solder job on a Saturday night, and was what I needed to make my 2014 MSI Z97-GAMING 3 motherboard work correctly with a 2025 Corsair RM1000e power supply unit!
Background
After the spectacular failure of my fifteen-year-old 800-W power supply (which, to be honest, had probably been foreshadowed for months, but I ignored the signs), I was left in a position where I needed to order a new, reasonably-priced ATX PSU as quickly as possible. I ended up grabbing an ATX3 unit on Paradigit, made by Corsair, with a slightly higher rating of 1000 W, for future-proofing. I didn't think too long and hard about backwards compatibility – after all, ATX is ATX, right? And my motherboard was only from the 2010s, so surely it could handle the latest and greatest power supplies, right? Wrong!
I installed the PSU, and initially everything seemed to be going grand. Grand, that is, until I tried to watch a clip on YouTube, and the audio coming out of my headphones was all horrible and clipped! It didn't occur to me that this might have anything to do with the voltage rails, so I assumed that either the motherboard or the headphones had been damaged, either by the PSU failure itself or the subsequent shifting of stuff around. I went out and bought a new headset in a hurry, which didn't fix it, so I assumed the motherboard was damaged. Honestly, this saddened me quite a bit, but I soon found a workaround: plugging my old headset into my monitor and routing the audio via DisplayPort. The new headset is currently sitting upstairs, awaiting reassignment!
The next problem was even more serious, and much less easy (read: impossible) to work around. As documented here , I found that my serial port had become unidirectional: it could only receive traffic inwards (and patchily), and couldn't send any data out. At this point, I started to suspect that something might be up with negative voltage rails, given that I know that an RS-232 signal has "high" and "low" voltage levels that are equidistant from ground. However, I pushed this to the back of my mind, because I naïvely assumed that a motherboard as "new" as mine (eleven years old!) wouldn't rely on "deprecated" ATX features to run an integrated component.
Initial (Flawed) Investigation
I let the situation languish for a while, then, to try to explore my hunch about the negative voltage levels, I broke out my PicoScope 2405A and started probing around the serial port's pins. Since the serial port is connected to a header on the motherboard by a ribbon cable, and I wasn't sure that the cable was still good, this involved poking directly at pins on the motherboard, while it was live! I thought that if the problem were a missing negative voltage, I might see a clipped signal on the data transmit pin, going high but then getting stuck at zero instead of going low.
To test it, I just used the echo command (on both DOS and Linux) to write "hello" messages and such to the serial port, while holding the probes in place and checking the scope trace on my laptop. Unfortunately, I got absolutely no signal, just some low positive constant voltage. This was enough to convince me (wrongly! ) that the SuperIO chip was damaged. My working hypothesis was that since the audio and serial headers are both near the bottom of the motherboard, right above the PSU, the audio and serial circuitry had (somehow) been splashed with corrosive electrolyte when the capacitor blew.
A Dead End: StarTech PCI Adapter
With this in mind, I began a timid search for serial port expansion cards that might have some hope of being usable under DOS. I found a forum post discussing an older PCI card, the StarTech PCI2S550. This card apparently has a DOS driver that programs it to remap its I/O ports and IRQs to the traditional addresses, which sounded like exactly what I needed to make the ancient WDEB386 kernel debugger work. I was highly dubious that anything like this would be possible with newer PCIe cards, so I went ahead and ordered the PCI card on eBay.
The card arrived at the end of August. I installed it in one of my free PCI slots, which thankfully were still a thing in 2014. I connected up my null modem to my laptop as usual, tried to get some basic test communications going under Linux and – nothing changed! The serial port could still only receive, but not transmit! To make absolutely sure I was using the right port, I went into the BIOS settings and disabled the onboard one, but still the same result.
Naturally, this led to a reappraisal of my previous conclusion!
The true cause: a missing –12 V supply
I went online and looked up the standard PCI slot pin-out, and located the singular –12 V pin (luckily it's at the very back of the slot, so it's not too hard to find). I removed the card and inspected it, and found that that pin was indeed connected to some circuitry on the PCB. I started probing the motherboard again, this time using a simple multimeter. Delicately poking the probes into the –12 V and GND pins of the vacated PCI slot, I found that same small positive voltage as I had seen on the TX pin before. In other words, the –12 V network on the motherboard was completely floating!
Honestly, the realization that I had wasted so much time, and ordered a PCI card that wasn't going to help me at all, paralysed me with annoyance for at least a week! After that, I briefly ran through some hare-brained ideas. Some of these ideas involved a 12-V relay and transformer I had recently remembered were sitting disused in some old equipment upstairs. Others involved a rechargeable 12-V battery pack I happened to come across, and I also thought of buying four CR2032 coin cells and stacking them! No matter what crazy "solution" I came up with though, it would necessitate more components, or at least more connectors – so I figured I might as well do it "right".
Trying to find information on this topic on a search engine is not entirely straightforward, since "-12V" or "–12 V" aren't exactly great keywords. However, I managed to find out a couple of key things:
On Wikipedia , it actually mentions that onboard audio amplifiers can also require a –12 V supply. This realization immediately provided an innocent explanation for why my audio was suddenly clipped with the new PSU!
While there are lots of ways to do a DIY circuit to get –12 V from +12 V, thankfully Traco Power make a simple component that does the full job for you!
The Traco TMV1212DEN
When I stumbled on the spec for the TMV1212DEN , it seemed too good to be true! Given a +12 V input, it would give dual ±12 V outputs, rated at 40 mA, which is the same current rating as you would get on a normal ATX2 power supply! I had a look through the docs to see if I could find the catch, but there really didn't seem to be any – so I went ahead and ordered one on eBay. At the same time, I ordered a 24-pin ATX extension cable to attach it to, so I wouldn't have to mess around with the leads that came with the PSU itself!
Once the packages from eBay arrived, and I had the time and headspace to work on the project, I assembled the converter roughly as follows:
Strip some of the insulation from a yellow wire and a black wire on the ATX extension cable, roughly halfway along its length, as shown in the photograph above.
Thread some 24-gauge wire around / through the stripped conductors, and solder it on.
Solder the other ends to the TMV1212DEN – the yellow ATX wire (+12 V) goes to Pin 1 (Vcc), and the black ATX wire (GND) goes to Pin 2 (GND).
Solder some more 24-gauge wire to pins 5 (–Vout) and 6 (Common) of the part, and leave the other ends dangling for now.
Attach the extension cable between the plug from the PSU and the socket on the motherboard, power on the PC, and use a multimeter to measure the voltage between the dangling wires.
This voltage came out as 15 V rather than 12 V, which spooked me, but then I looked at the voltage-load curve in the documentation. The curve indicates that under no-load conditions, the voltage is expected to be 30% higher than nominal. The level should come within the required 10% tolerance once it's at "minimal reasonable load", which reassured me enough to press on!
Turn off the PC, remove the cable, and strip some of the insulation from the blue wire as before.
Thread the wire from pin 5 (–Vout) around the stripped blue wire, and solder it on.
Join the wire from pin 6 (Common) to the one from pin 2 (GND) and solder them together, to ensure there is a common ground between the two circuits.
Reinsert the extension cable in the PC, power it back on, and test it out!
I did get a few frights initially, because the motherboard didn't seem to be powering up properly, but it turned out that I had simply not inserted the plug firmly enough (it was slightly askew, so the pins on one end weren't making contact). Once I had it powering on properly, I plugged my headset back into the motherboard, listened to some audio, and found it working perfectly for the first time in months!
I got another fright with the serial port – when I tried to test it initially, I was getting I/O errors, on both Linux and DOS, so I thought maybe I really had damaged the SuperIO when I was poking around with the PicoScope previously. Then I remembered I had disabled it in the BIOS! Once I re-enabled it, it too worked perfectly for the first time in months!
Conclusions / Lessons Learned
I think the main takeaway from this is that if you have a motherboard from before 2022 , and use its serial port and/or onboard audio (or even PCI cards with same!), then you should really double-check the specs of any replacement / upgrade PSU to make sure it still provides a –12 V rail . But the good news is that if you missed that trick, the Traco voltage converter can come to the rescue with a really quick bit of soldering as detailed above!
While this solution works, it's not necessarily the best possible way to do it. Most obviously, it would be nicer to make a small PCB with ATX connectors and the voltage converter soldered directly on. This would take up less space, and eliminate exposed solder joints on loose wires inside the PC case, which are never an ideal thing to have!
It also turns out that the TMV1212DEN's dual output is unnecessary for this application. When looking at the Traco lineup, I didn't appreciate the fact that the input and output circuits are completely isolated from each other, and that I would have to connect the Common to GND to make it work. This leaves me with an extra +12 V pin that I have no use for. If I were doing it again, I would use the single-output TMV1212EN instead, and connect pin 7 (+Vout) to GND, which should give the same effect, but without the extra unused pin. So, that's something to keep in mind if you ever need to follow this guide!
Of course, I also now have a spare headset and PCI-to-serial adapter, acquired after my two misdiagnoses, so hopefully I will find a use for them in the not-too-distant future…
So anyway, thank you for reading this, and happy computing! :D