Whine like this is usually from the transformer core, which is a series of plates stacked together and glued in place. Over time, due to the magnetic fields being pushed through the core, they can slowly work loose. A good test would be to take some glue, nothing too thin, and put a few lines of it across the core plates or anywhere that you can access the core. If it's a solid core (not a series of plates) then glue can still help, just apply the glue between the core and the PCB to dampen the whine.

Often the plates are coated with something that makes it difficult to see the individual plates. It might be necessary to first remove some of this coating before applying a glue. I would go with something like JB Weld or E6000 adhesive. E6000 has a little bit of give to it, so it can actually dampen vibration and harmonics quite well.

Darn, I can't edit my post. Was going to add that I'm not familiar with that particular transformer, but it looks like you'd have to remove the clips and/or the black covering to see what's actually inside that could be causing the whine.

I will try to desolder the transformer and see if it is possible to get to the laminated plates.
I was not able to find details on the transformer, in case I'll have to order a new one.
There is a sticker on its side that says:
8311 95-3555-01
EIA 343 MEXICO

I noticed that EIA comes up when searching for transformer in google. I guess it has something to do with the type of core? An EI core maybe?

That's a ferrite transformer and the whine comes from the core itself due to magnetostriction, as function of switching frequency used in the PSU. It should be way above audible range though...

Chances are that you have to re-cap the PSU so that it is back in spec, especially the capacitors near heatsinks that wear fastest because of the elevated temperatures. New transformer will not fix this.

Tiido wrote on 2024-08-04, 13:44:

That's a ferrite transformer and the whine comes from the core itself due to magnetostriction, as function of switching frequency used in the PSU. It should be way above audible range though...

Chances are that you have to re-cap the PSU so that it is back in spec, especially the capacitors near heatsinks that wear fastest because of the elevated temperatures. New transformer will not fix this.

Well, I was trigger happy and desoldered the transformer anyway.
Found a dead bug under it, so at least that it taken care of 😉
I'll take this opportunity and clean it a bit. It is quite disgusting.

So if I understand correctly, this type of transformers is not laminated, rather a solid core?

I removed all the capacitors and checked them in a cheap multi-tester.
Indeed some seem to be below 20% of the rated capacitance. I'll order new caps.
One thing I'm not sure about, is the ESR reading. Can someone take a look at this table, and tell if I need to replace any caps due to high ESR?
I believe that the lower the capacitance, the higher the ESR can be? But how much is too high?

Switch mode PSUs basically never have iron cores, they just cannot work at these freqs without causing the core to melt (i.e induction heating 🤣). You can see from your photos that there's part of the core visible and it is the usual ferrite stuff as expected, covered by shroud in most places and flux band around it all to reduce stray field.

Most of the smaller value capacitors (22µF and smaller) seem ok, even 1ohm ESR isn't a big deal in those normally but the larger capacitors defintely have aged, they all should have under 0.1ohm ESR. That 330µF and 47µF are not good though.

The ESR look awful if thinking about mid 90s stuff and newer, but not sure what design parameters this PSU met, so maybe best to see if you can find original specs for a sample of the capacitors for around the time they were made to see how far they are off original spec.

About ESR: They say "bad caps stand out clearly", by having an ESR like 2 to 10 times higher than expected. In your table, you have a lot of 10µF/16V caps. Most are around 0.6 Ohms ESR, which seems to be the "typical" value for that type of cap. But two caps stand out: C29 has twice that value, and C19 is just awful. While you might think that C19 is a "quite good cap", as the capacitance is +18%, I'd guess that's an artiface of the capacity test method of your capacitor meter. A typical way to measure the capacity of "large" capacitors (i.e. in the micro-farad to milli-farad range) is charging them from zero to a pre-set voltage and measure how long it takes. The longer it takes to charge the cap, the higher the capacity. Right? Well, that's what the tester assumes, but this assumption is invalid if the capacitor has high leakage current and discharges itself while the tester is trying to charge it. I assume that's exactly what happens here: C19 is toast. It both has a high ESR and high leakage, the high leakage is misinterpreted as higher capacity.

C27 also has a suspiciously high ESR value. You are right that a good rule of thumb is: The higher the capacitance, the lower the expected ESR. This does not generally apply to high voltage caps, although in your case the ESR of C1 and C12 is stunningly good. Comparing the 1.2 Ohms of C27 to the 0.42 ohms of C5 (which is significantly lower capacitance), the ESR of C27 does not look good at all. I can't be sure that C27 isn't some "unimportant" cap that is from a cheap series that generally has a high ESR, but seeing the -20% capacity deviation at the same time, I tend to suspect that C27 is worn out.

My recommendation would be to swap all caps with -15% or higher capacity loss, as well as all caps with an ESR exceeding 1 Ohm. There is absolutely no need to swap C1 and C12 - they are likely as good as new.

Just wanted to update that I ended up replacing all the following caps (in gray), and now the PSU is working quietly! (well, apart from the fan, which I will tackle at another time)

I also fixed some errors in the list (uF, not pF, and missing C17) so it can be used as some reference for whoever searches for 5150 PSU.
Many thanks. This community rocks!