How to quickly & easily fix coil-whine(coil choke noise)

[sobakasirko]

I'd love for someone who's got a problematic, whiny system to try the following:

3. Try the various combinations of power cable(s) - use the far end connectors, avoid sharp bends. Try using both connectors on one cable instead of two seperate ones, short term
(If it's a modular PSU, disconnect the unused one at the PSU!)

2. These are messy as heck because they were tied together and left in the PSU box for a year or so, but this is how it should be done for a high wattage GPU.
Use both ends of the same cable for something under 250W, or use the far ends of both cables. Avoid folding the cable back over itself!
View attachment 294317

Hi! I find your message interesting and helpful.

I've bought a used 6800xt XFX that is good in all aspects (bargain price, undervolting, frequency, low fan noise) other than coil whine. Once I beat a few games, I'll give it to my dad so that he can play his favorite COD, and it will be in his computer for a long time. So I want to fix the whine.

I have 2 PSUs to try. One is Fractal Design Edison M 650W Gold that the card is connected to right now. The card whines when connected to PSU's undetachable 2 x 6+2pin cable (the daisy-chained one, one side connected to PSU and two 6+2pin connectors on the other side to the GPU). However, I used to have a Sapphire Fury card that was consuming even a bit more power and was not whining at all connected with that single cable, same goes for my old 290X and more recently 1080ti. But that PSU is somewhat old, has seen many hours of heavy load, and the cable was bent many times, so maybe it's a factor.
Anyway, at home I have a box with additional 2 x 6+2pin modular cables (also daisy chained), I will use far ends of those like on your picture to see if it helps.

The second PSU is Thermaltake Toughpower XT 1475W Gold. It has two 12V power rails, but 1 rail is dedicated to CPU only and the second to GPU only, so I can't test the GPU whine when connected to 2 separate rails. Still, the card will have a 12V 75A (900W) rail to itself. That PSU has daisy-chained 2 x 6+2 cables, but also a few single ones. So on one side the cable connects to the PSU and on the other it's 1 x 6+2pin. I'll try 2 of those, and some other combinations maybe. The PSU, despite being over 10 years old, has never been in use and was sitting in a closet for a decade, new old stock.

Will post the results here later.

 

[Shrek]

NO!! Don't use this kind of glue. Silicone glues contain acidic solvents which WILL DESTROY your card. Acids will severely corrode the metal pathways and components.

Do not use silicone glues with electronics, EVER!!!

There are electronics grade versions, but your point is still valid; standard versions release vinegar.

ASI 388 Electronic Grade RTV Silicone (americansealantsinc.com)

 

[Nonam3]

Update on my side.

Ferrite - tested with various sizes (of ferrite beads) and locations (AC cable, PCI-E GPU cable etc.) and amount (up to 6 on AC cable) - no difference.

PSU - tested with 2 additional PSUs Corsair AX1600i 1600 W and Thermaltake Toughpower GF3 1650 W, hoping 12VHPWR 600W connector will help (the fan on this one, OMG, louder then my entire system while taking only 430 W, and that with "Smart Zero Fan" "technology", no wonder no one complains about coil whine ;/) - no difference
PSU* - as before, used those PSUs to power only GPU the rest of PC was using beQuiet 1500W. <- Dunno maybe that's the issue???

Glue - mosfets + re-glue coils - no difference

Pics of glue:


Did also notice some discolouration on gpu mounting bracket (corner area) and on some other parts of PCB (ignore old paste this was after several dis-assembly of the card, re done it already and cleaned up gpu):



Please let me know if maybe I used to little glue (used minimal amount didn't wanted to change Z height) ?

At this point I can say... money well spent /s.

As before I'm open to any / every suggestion as I'm out of the ideas (except for putting it into cabinet and insulating / sound proofing).

 

[lexluthermiester]

Please let me know if maybe I used to little glue (used minimal amount didn't wanted to change Z height) ?

No, you used a good amount. Sometimes it just doesn't work..

 

[lexluthermiester]

My latest purchase arrived today. @Mussels earlier proposed the idea that ferrite chokes can help and testing with one I already had did yield a measurable result. So I wanted to try a different approach, using them is stages.

I'm going to use two of the large ones on the input power cable and the motherboard connection cable. Next will be the PCIe power cables, then on the display cables. The goal is to surround the GPU in question with ferrite chokes to see if an effect is rendered. Testing to come... Stay tuned!

 

[Mussels]

This threads been amazing for education on a bunch of topics

Theres a PDF here that has an insanely useful amount of electronics information
Wiring Unlimited (victronenergy.com)

They're talking about a battery setup, but this applies to DC-DC just as well - you can basically say "PSU" or "prior component" when they say battery

• Varying load, causes voltages to ripple if the supplying component can't keep up
• Ripple causes AC interference
• That AC interference can then spread as EMI, RFI, crosstalk, or common ground interference to nearby components
• Added together, this is why some noise snowballs - it's diminishing returns for efficiency, with noise as the output as it goes more out of spec

This *AMAZING* image is their example of how ripple is a measurable AC voltage difference between two measurement points in a DC setup

Hi! I find your message interesting and helpful.

I've bought a used 6800xt XFX that is good in all aspects (bargain price, undervolting, frequency, low fan noise) other than coil whine. Once I beat a few games, I'll give it to my dad so that he can play his favorite COD, and it will be in his computer for a long time. So I want to fix the whine.

I have 2 PSUs to try. One is Fractal Design Edison M 650W Gold that the card is connected to right now. The card whines when connected to PSU's undetachable 2 x 6+2pin cable (the daisy-chained one, one side connected to PSU and two 6+2pin connectors on the other side to the GPU). However, I used to have a Sapphire Fury card that was consuming even a bit more power and was not whining at all connected with that single cable, same goes for my old 290X and more recently 1080ti. But that PSU is somewhat old, has seen many hours of heavy load, and the cable was bent many times, so maybe it's a factor.
Anyway, at home I have a box with additional 2 x 6+2pin modular cables (also daisy chained), I will use far ends of those like on your picture to see if it helps.

The second PSU is Thermaltake Toughpower XT 1475W Gold. It has two 12V power rails, but 1 rail is dedicated to CPU only and the second to GPU only, so I can't test the GPU whine when connected to 2 separate rails. Still, the card will have a 12V 75A (900W) rail to itself. That PSU has daisy-chained 2 x 6+2 cables, but also a few single ones. So on one side the cable connects to the PSU and on the other it's 1 x 6+2pin. I'll try 2 of those, and some other combinations maybe. The PSU, despite being over 10 years old, has never been in use and was sitting in a closet for a decade, new old stock.

Will post the results here later.

Dont forget some simple tests like with a ferrite choke or two, and making sure of some simple things like grounding within the PC. Make sure the GPU has both screws on the PCI slot covers, for example.

My latest purchase arrived today. @Mussels earlier proposed the idea that ferrite chokes can help and testing with one I already had did yield a measurable result. So I wanted to try a different approach, using them is stages.
View attachment 295826
I'm going to use two of the large ones on the input power cable and the motherboard connection cable. Next will be the PCIe power cables, then on the display cables. The goal is to surround the GPU in question with ferrite chokes to see if an effect is rendered. Testing to come... Stay tuned!

From what i've read and tested with my VDSL setup, it's best to have them close to the ends - so one at the AC plug to the PSU, one at the PSU output, one at the GPU input

Inputs prevent EMI/RFI being absorbed, outputs prevent it being transmitted (Like with JonnyGurus example of a PSU with a weak primary output capacitor)

I wish we had a way you could measure the noise, for a scientific comparison

 

[lexluthermiester]

I wish we had a way you could measure the noise, for a scientific comparison

Oscilloscope you mean? I have one, but it needs repair... A pricey repair. I'm thinking of getting a Hantek handheld model like the one Adrian Black(Adrian's Digital Basement) has as they're not expensive. I'm also thinking strongly about an isolation transformer...

However, the ferrite choke's yielded a result, but nothing noteworthy. The one ferrite choke on the PCIe power cable has an effect, but all of the others added in, did little. Removing all of them, increases noise by a noticeable amount. Going to leave the choke on the PCIe power cable. The GTX560 card is not a main use card. The testing system in question is running a GTX960 4GB as a daily.

 

[Mussels]

Oscilloscope you mean? I have one, but it needs repair... A pricey repair. I'm thinking of getting a Hantek handheld model like the one Adrian Black(Adrian's Digital Basement) has as they're not expensive. I'm also thinking strongly about an isolation transformer...

However, the ferrite choke's yielded a result, but nothing noteworthy. The one ferrite choke on the PCIe power cable has an effect, but all of the others added in, did little. Removing all of them, increases noise by a noticeable amount. Going to leave the choke on the PCIe power cable. The GTX560 card is not a main use card. The testing system in question is running a GTX960 4GB as a daily.

You can measure ripple with the AC mode as mentioned in that PDF, but no i meant if it helps, being able to record how much it helped.
More of a curiosity thing, than anything else.


Can you try the single ferrite choke on *other* in use power cables in the system? Such as the one running your molex/sata devices, to see if they're generating the noise - i can totally imagine something like an RGB controller of PWM fan hub being a source of EMI/RFI, or even fans themselves

 

[lexluthermiester]

Can you try the single ferrite choke on *other* in use power cables in the system?

I can and did. Only the one on the PCIe cable showed an audible result. So there is a result to be had, just a small one. Of course, others results may vary from nothing to more significant.

In the end, there is some merit here, so it's worth trying. The ferrite pack was $10 shipped.

 

[Shrek]

Oscilloscope you mean? I have one, but it needs repair... A pricey repair. I'm thinking of getting a Hantek handheld model like the one Adrian Black(Adrian's Digital Basement) has as they're not expensive. I'm also thinking strongly about an isolation transformer...

I hate the old oscilloscopes where one lead is grounded and not floating; I recall tripping the whole rooms breaker by touching the ground to the wrong place.

An isolating transformer is a good investment in saftey and would have saved me from tripping the breaker.

 

[lexluthermiester]

An isolating transformer is a good investment in saftey and would have saved me from tripping the breaker.

I want it more for safety than breaker tripping.

 

[Deleted member 230939]

So I could potentially alleviate the coil whine on my 7900 XT?

I am not usually one to shy away from tinkering but this was an £837 card... so I kinda don't want to ruin it.

 

[lexluthermiester]

So I could potentially alleviate the coil whine on my 7900 XT?

I am not usually one to shy away from tinkering but this was an £837 card... so I kinda don't want to ruin it.

As I said in the OP, if it's under warranty and you can live with it....

But the reality is, it's just superglue, it's not going to hurt your card.

 

[noenthu]

Not sure if this is the right thread to ask this question. I was able to RMA my 4090 fe with nvidia. The second one has slightly less coil whine than the first. It is still quite bad though.
I have had 2 gig 4090 oc and 1pny The Gigabyte ones had little to no coil whine. First one actually had 0 perceptible. The pny had some but would dampen easily with 80% power limits
The 2 4090 FE are possibly 10 times worse.
I am considering trying my luck with another rma request with Nvidia but I’m not hopeful.
If that fails I’ll have to give this superglue mod a try. Does the FE card make a good candidate?

here is my first FE for reference.

http://imgur.com/a/ss72Kea

 

[Yougotonejob]

also check this at 32:59

 

[Mussels]

I can and did. Only the one on the PCIe cable showed an audible result. So there is a result to be had, just a small one. Of course, others results may vary from nothing to more significant.

In the end, there is some merit here, so it's worth trying. The ferrite pack was $10 shipped.

Thanks, that's useful to know it's less likely to be 'feedback' from other power cables in the system and to focus on the PCI-E cables primarily, and possibly exclusively
Feel like trying your ATX cable and EPS-12V cables next? They do feed into the PCI-E slots...





Regarding the video above: Yes, capacitors are a primary method to filtering out the noise - that's why the PSU is involved, and upgrading can help... but not if your upgrade has the same type of output capacitor, as quoted by jonnyguru a few posts above.

Buildzoid states "This thing has a bunch of coil whine ... all the coil whine is from the input filtering inductors"
After his capacitor upgrade

"The voltage difference across the inductors gets smaller ... the voltage difference between them is smaller, the magnetic fields are smaller, and they vibrate less"

Multiple names for the same thing depending how it's used in a circuit is why theres so much confusion on this.



So our goal is to remove voltage differences, shield or filter magnetic fields, and reduce vibration



Notice they're ferrite devices that are used to convert electricity into a magnetic field? That's why the ferrite chokes help, they're adding another stage of that filtering, evening out some of it earlier. Turning that AC noise into magnetism, before the inductors do.

Removing ripple, the "AC" difference between all the inputs and grounds (PCI-E cables and slot cover screws, motherboard to case screws and all their respective grounds) is one first step we can reasonably work at doing. PCI-E risers could help or hinder this, for example.
Replacing filtering capacitors on GPU's is a bit beyond the average user

Then we can work on the magnetic fields, which I guess the ferrite beads help with. Maybe this is why heatsinks can help, depending on what metal they're made from??

What else do we have, we can use as a heatsink material or noise blocker?
Unless we get something pre-cut like those copper shims I love to use on GPU's and chipset cooling, it'd be tough to test this one out, especially with conductive metals.
10 Pcs Metal Strips Metal Shim Stock Metal Sheet Metal Shims Gap Filler Trim Strip 12'' Length x 0.02'' Thickness (1'' Width, Stainless Steel): Amazon.com: Industrial & Scientific
If it's magnetic interference, would stainless steel shims (with thermal pads?) be an answer?

Vibration we've worked out with the superglue, maybe the reason some cooling methods seem to help is just that they're solid and dense - thick thermal pads, heavy copper heatsinks etc might simply reduce the final vibration by pressing the parts tighter to the PCB - reducing the symptom, but doing nothing to the cause.

 

[Veseleil]

Maybe we should write to Steve Burke to make pressure on the card makers by making a huge coil whine rant video. So we could maybe get some solution from the factory, like adding those capacitors BZ did. Otherwise we won't see any kind of improvement for like ever I'm afraid.

 

[Mussels]

Maybe we should write to Steve Burke to make pressure on the card makers by making a huge coil whine rant video. So we could maybe get some solution from the factory, like adding those capacitors BZ did. Otherwise we won't see any kind of improvement for like ever I'm afraid.

Unfortunately reading all the comments JonnyGuru made before self-deleting his account, we've got no hope.

They had a PSU that worked with older GPU's fine but whined on a Nvidia 20 series cards, so they changed a capacitor as a silent revision to work better on the 20 series to benefit users going forward - this punished anyone who bought that unit or got an RMA and used it with older GPU's which now had coil whine. (AMD vega and NV10 series users, vs 20 series in his comments)

I Appreciate the heck out of his honesty and upfront discussion on why the choices were made, but it puts users in a situation where you have to gamble on a PSU and GPU being compatible because buying a super expensive high end top of the line PSU may mean nothing if the following GPU designs don't like it.

I guess this is why theres that push for pure 12V only in the industry? One voltage, less hassles.

 

[Yougotonejob]

resonance frequency (f = 1/2π√ (LC))
adding another cable usually has a Capacitor at the PSU side of the cable or at the very end of the PCIe side.
this is the same thing BZ did in the video i post above.
adding Capacitor will modify the resonance frequency, and change it to a lower frequency. If my math is correct the larger the C the smaller the f when number of inductor (the "L" part) is fixed.
but this is only adding like 400 µF on average. What BZ did in the video he was adding about 2000µF to see it effective.
see this image

https://www.techpowerup.com/review/msi-radeon-rx-6800-xt-gaming-x-trio/images/front.jpg

the bigger Caps on the right hand side is that BZ called "input filtering"
the smaller Caps after bunch of LR15 is the "output filtering".
the input and output is relevant here BZ means talking about VRM here. to give VRM better filtering is making coil less noise.
I wonder if we replace with bigger Capacitance like 330 or 470 and Low ESR for the most left input Caps would yield the same result (???)
I am in CS major not EE but i got chat GPT here. hope anyone got EE degree can help me here.

here is the gpt answer


Increasing the capacitance from 270µF to 330µF generally provides a larger capacitance value, which can have several advantages for filtering and energy storage applications. However, there are a few potential drawbacks to consider:

1. Size and Space: Larger capacitance values often correspond to physically larger capacitors. If space is limited in your application or on the PCB, increasing the capacitance may pose challenges in terms of fitting the larger capacitor within the available space.
2. Cost: Capacitors with higher capacitance values tend to be more expensive than those with lower capacitance. If cost is a significant factor in your application, increasing the capacitance might lead to higher component expenses.
3. ESR and ESL: Capacitors with higher capacitance may have higher equivalent series resistance (ESR) and equivalent series inductance (ESL). Higher ESR can lead to increased power dissipation and reduced filtering efficiency. Higher ESL can affect the capacitor's impedance at higher frequencies and impact its performance in high-frequency applications.
4. Transient Response: Capacitors with larger capacitance values generally have slower transient response times. This means they may take longer to charge or discharge when sudden changes in current or voltage occur. In applications where rapid response to load changes is critical, a larger capacitance could result in slower circuit response.
5. Ripple Current Rating: Capacitors have specific ripple current ratings that indicate their ability to handle AC current without excessive heating. When increasing the capacitance, it's important to ensure that the new capacitor can handle the anticipated ripple current levels within your application's requirements.

It's essential to carefully evaluate these potential drawbacks in relation to your specific application and its requirements. Consider factors such as space constraints, cost limitations, desired transient response, and the impact on overall circuit performance before deciding to increase the capacitance.

1 size we can make sure same size.
2 can be ignored each was like $2
3 pick low ESR already
4 GPU work load is very much like heart beat, never flat load. not sure how this is going to effect.
5 the Ripple at 12V here is most 120mA i don't worry too much about it, all caps can filter that.

 

[InVasMani]

Oscilloscope you mean? I have one, but it needs repair... A pricey repair. I'm thinking of getting a Hantek handheld model like the one Adrian Black(Adrian's Digital Basement) has as they're not expensive. I'm also thinking strongly about an isolation transformer...

However, the ferrite choke's yielded a result, but nothing noteworthy. The one ferrite choke on the PCIe power cable has an effect, but all of the others added in, did little. Removing all of them, increases noise by a noticeable amount. Going to leave the choke on the PCIe power cable. The GTX560 card is not a main use card. The testing system in question is running a GTX960 4GB as a daily.

Did you try running more than one on the PCIE cable at different ends of the cable and/or middle!? The GPU is getting most of the voltage and source of the problem so it's natural you won't see much improvement outside of it. I suppose the motherboard power cable could contribute some from the PCIE bus power and lanes, but to a lesser degree most of it's power is via the PCIE cables.

 

[lexluthermiester]

Feel like trying your ATX cable and EPS-12V cables next?

Actually, I did that one too. I had a total of 5 ferrite chokes attached. 1 on the power cord outside the system, 1 on the motherboard power cable, 1 on the CPU power cable, and 1 each behind each PCIe power connector.

Did you try running more than one on the PCIE cable at different ends of the cable and/or middle!?

See above..

 

[Shrek]

Unfortunately reading all the comments JonnyGuru made before self-deleting his account

By this do you mean the Web site?

I attach a joke from his facebook account

 

[Nonam3]

Update:

Re-Glue (two days ago and yesterday):

Two days ago I attempted third time re-glue process.
This time I put glue in between inductors like this:

On the edge in red is profile that I wanted to achieve in between them.
Also changed glue density (same brand, SANUS, they have 3 types of glue - thin, middle and strong) used middle one this time (previously thin one).
In spirit of sharing here is thermal properties of that glue (PL company) that I got directly from company representative:


Waited 24h for it to dry and yesterday re-assembled water block on it and installed.

Result (yesterday):
Furmark - coil whine like before...
OpenPose - coil whine like before (and like my 3090 MSi Gaming X Trio)

Firmware (yesterday):

At that point I was like "whatever" lets try everything.
Flashed my Gaming X Trio with:
- Asus Strix OC
- Gigabyte Gaming OC
- MSi Gaming (non X)
- MSI SuprimX

and played around with mem / power etc. using MSi Afterburner. Because I'm on water so I was like what the hell lets do +MAX on all of those sliders and see what happens. Was able to pull 550W max and got some additional FPS in FurMark. Was trying to undervolt (obviously) etc. none of those resolved coil whine. Always after 825mV card produced audible coil whine.

Was extremely tired yesterday and didn't wanted swap to 3090 so kept it in PC and my nephews did play WatchDogs legion few hours. What I did noticed is that in WD Legion card didn't whined so much and was behaving more silent, but because I didn't tested this game before using 4090, so thought that this is placebo effect.

Today:
After around, next, 24h (from gpu install yesterday) did fired up Furmark again and... no coil whine!!!. WTH ! Tested few time with different titles and there is huge different to acoustics overall.
Using OpenPose this behave same as before but, maybe, noise is more tolerable.
Tried WDL - there is hum but no coil whine (no high pitch noise, just lower frequency "electrical" noise)
Tried Everspace2 - before anything above 30FPS causes coil whine, now 60 ok, 75++ there is noise and muffled coil whine.
Shadow of the TombRider - menu - noise but muffled coil whine (huge difference compared to before)

Weirdness:

1. during initial testing, today, when I did noticed no coil whine in Furmark, did noticed drop in FPS from around 860 to 600 in Furmark. Dunno what was problem, as yesterday I did enabled in Windows Hardware-accelerated GPU Scheduling and played with firmwares. What I did was toggled off (+OS restart) and on (+OS restart) and re flashed firmware to Gaming X Trio and then again to Gigabyte Gaming OC. This resolved issue and now I hover at around 930FPS. Default setting 720 + default rest.
NOTE: I know that this could be caused by overhearing VRMS and power delivery system (i.e. inductors) and I'm afraid that this could be the case. Or maybe I had some app in background that didn't closed properly (had bunch of app like GPU-Z, HWInfo, MSi Afterburner etc. opened and was playing with them, so maybe MSi Afterburner misdetected GPU and maybe did applied profile or I did that without noticing dunno). Anyway will report if something happens (hopefully no )

2. Thermal paste on GPU has different / weird structure / consistency. I use Thermal Grizzly Kryonaut Extreme and before glue shenanigans I could re-mount GPU few times (testign different thermal pads) and paste had same consistency and was working properly. Now when I disassembly water-block and don't re-paste thermals go north 100+. Not only that thermal paste consistency is like this (this is old pix and it seams this happened after initial glue process, yesterday during assembly paster had very similiar structure tho this from pix below and this is after re-pasting from pix below):



Stickiness of it also changed now it just don't stick like before etc.
Maybe glue that I used creates some fumes.
Thermals (after applying paste and not dissembling card) are ok (GPU 59 GPU HotSpot 68 Mem 54 - NOTE such high due to 25% pump and 20% fans, speed controlled by water temp not load, after 3 min of Furmark temps are GPU 53 GPU HotSpot 62 Mem 52, hotspot delta 8-9*C so good, and yes this is with around 40% pump and fans speeds, water temp 36*C).

3. Glue - I'm afraid that glue did drop (even after initial 24h period) on thermalpads as I did seen that thermalpads, in area bettwen inductors and mosfets, did changed it's rigidity (became more rigid, glued). That maybe explains why now coil is gone? This could be because of the glue that I used (Sanus company) or not drying it enough (always 24h wait period) but this (dry time) does vary depending on glue used.

4. Hydrogen cyanide !!! health hazard / DANGER (potential) !!! - https://pubmed.ncbi.nlm.nih.gov/21737219/ If I understand this correctly forget about solder work on glued boards. Also inform new owner (when re-selling) that board was glued as any soldering attempt near super glue may / will cause fumes that could be potentially lethal / harmful to health!.

Summary:
Spend almost 1 moth trouble shooting and resolving this issue. During that time had GPU in PC only for testing now it's installed.
Wanted to say THANK YOU VERY MUCH to @lexluthermiester for sharing and testing on his GPU and all You guys for adding additional info and other approaches that one could tried.
Will be around and report if GPU dies (hopefully not :cross-finger.


My discoveries:
During troubleshooting I run into informations that buildzoid shared regarding changing frequency on drivers for NVidia 4090 GPUs so if someone (not me) is brave enough to use solder there is a way to change switching frequency but that's beyond me. Run into another YT channel that person (anna IIRC) was doing 970 fix and she mentioned coil whine and info that she changed switching frequency of drivers and that helped coil (but she also re-soldered inductors so anyone guess what helped).
MSI uses very similar design on 4090 to the one from 3090 (same noise when doing compute), same controller for memory and core.
Gigabyte uses separate controllers for core and mem and maybe that's why theirs GPU don't coil (so much) ???
Switching firmwares between brands works just fine (maybe this is common knowledge, didn't know it myself, DON"T DO IT if not needed!)
Currently using firmware from GIGABYTE RTX 4090 GAMING OC, works fine - dunno if that maybe had something to do with coil whine fix ??? most likely NO. Voltages on 12V raised from 12.1-2 to 12.3-4 and GPU Voltage at idle is now 885mV and with Gaming X Trio was under that value like 850 - 825mV (but maybe I'm mistaken, deleted pix so don't have compare).

 

[Shrek]

4. Hydrogen cyanide !!! health hazard / DANGER (potential) !!! - https://pubmed.ncbi.nlm.nih.gov/21737219/ If I understand this correctly forget about solder work on glued boards. Also inform new owner (when re-selling) that board was glued as any soldering attempt near super glue may / will cause fumes that could be potentially lethal / harmful to health!.

I'd prefer to be able to do solder work on my stuff, so will be avoiding any boards that have been super glued.

 

[Mussels]

My brain caught up with a few ideas, which resulted in some WEIRD AS SHIT DREAMS.

The short version that matches reality:
Steel etc are magnetic, not electromagnetic. We don't want that.

Copper is a better absorber of EMI/RFI etc, so copper heatsinks on our VRM's is going to be better - but it'll work better if they're *connected* to each other or touching, rather than seperate small heatsinks

Looking around at my hardware that has coil whine vs those that dont, the worst ones have alu heatsinks on the VRMs (or none) while the best have copper and are all connected to other things - like the heatsink of the card and its backplate (GPU monoblocks that touch the VRM's too), or the built in IO shield on a motherboard that connects to the case directly.

It's called common mode interference, with common also meaning ground - be it magnetic or electrical. The ferrite beads help by converting the AC into magnetic fields, so one heatsink over a group of VRM's is going to let them share a common magnetic field vs many smaller ones overlapping, and if you electrically ground it you're dissipating that energy directly.

Thermal Grizzly Kryonaut Extreme and before glue shenanigans I could re-mount GPU few times

That's a thick paste and it simply dries out when exposed to air. It's not going to last many, if any, remounts after its been exposed to high heat.
Anything that changes the voltages/voltage curve can push above or below the frequency that's causing the whine - they're using PWM where the power is pulsed out at specific times, and you're in a sour spot - getting voltages to raise or lower can solve the issue.

That sour spot can depend on the PSU it's paired with as much as anything else (JG's comments on the primary output capacitor make sense there), so yeah it can be an annoying trial and error to figure it all out

By this do you mean the Web site?

I attach a joke from his facebook account

I linked and quoted them above - he posted on a forum with a ton of detailed answers about corsairs PSU designs and how solving whine for nvidia 20 series caused it for 10 series and vega, but they went with it to benefit future owners.

They also use cheaper primary capacitors on high wattage budget line PSU's that are more likely to cause whine because 'users that buy high end GPU's buy high end PSU's'
When designing a PSU it's easy to make a design, and slap in higher rated capacitors with the same specs - even if the design itself isn't great.
Don't buy an 850W 80+ white unit when you could get a 650W platinum for the same price, sort of logic.

I dislike that attitude but it does match up with reality, a low wattage premium line PSU is going to have less ripple and noise than a higher wattage budget unit.

I'd prefer to be able to do solder work on my stuff, so will be avoiding any boards that have been super glued.

If you're soldering something with tens of grams of superglue in an enclosed space with no ventilation and huffing the fumes, i'd say you should work on that.

100 μg of HCN when 1g of cyanoacrylate monomer was heated at 280 °C

That wont hurt your donkey or fiona, the amount of superglue in a tube is tiny, and no ones putting full tubes on a GPU anyway.