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

[Yougotonejob]

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.

I bet NVDIA and AMD has 0 employee working in NVH department.
like in car there is NVH expert to make sure your transmission and engine to avoid some frequency so that interior or exhaust etc won't make buzz sound.

Now with fancy RGB and class window and ITX more and more case is on the table. Noise is going to be a bigger issue.
I don't know who that Steve is .... sorry I from non English speaking community.

 

[Veseleil]

I bet NVDIA and AMD has 0 employee working in NVH department.
like in car there is NVH expert to make sure your transmission and engine to avoid some frequency so that interior or exhaust etc won't make buzz sound.

Now with fancy RGB and class window and ITX more and more case is on the table. Noise is going to be a bigger issue.
I don't know who that Steve is .... sorry I from non English speaking community.

Steve Burke aka Gamer's Nexus Steve.

 

[Mussels]

Steve Burke aka Gamer's Nexus Steve.

Not to be confused with Steeve the glyphid in my profile pic

I do suspect Nvidias 12 pin connector was an attempt to help with this sort of thing, by having all the power wires together from one common sources and not mixed cables of potentially different lengths - but it's really impossible for nvidia or AMD to help with anything other than reference card design

All my reference cards (not FE cards, but reference) have had zero whine, it's always the custom "overclocked" ones with issues - because they cheapen out on the components, raise it 25MHz and charge you more

 

[Yougotonejob]

Source https://www.expreview.com/79729.html
Use google translate if needed.
This article mention some way was not in this thread. the cable mod. Writer suggest put 100μF up to 3000μF in cables see if it work.
Do not exceed 3k.

My thought it kind of make sense the same mod as BZ did, just change a location it's not on the GPU PCB but it's part of the system that effect GPU. The best of this way is keep the GPU low Mod, make it as OEM as possible and also increase the Capacitance in the system. This still consider as input filter for VRM.
f = 1 / (2π × √(L × C)).
My plan is to try parallel 2x 330μF in 1,4 and 3,6.
(Edit 14 36 in 6 pin connector, maybe in 6 pin cable only do 1 and 6 is working too...)
If it's a 8 pin 15 pair and 37 pair.
ESR in parallel is 1/ESR_parallel = 1/ESR_A + 1/ESR_B.
Parallel is making the ESR as low as 4-6 mOm, as BZ mentions low ESR helps filter high frequency and big capacitance help with low frequency. Why pick 330μF ? Cuz morden Corsair has exactly same size in their cables. Of course they not have in both 1,4 and 3,6 pair they are in one of these pair. Also they not parallel them.

I am browsing Digikey now see if I can get this work and report you guys later. it will be long time until next DIY.

I tried ferrite bead too, that bead vibrate at CS GO 240 FPS even louder than the whole system. The bead sing and dancing and roaring around the PCIe cables. This happens only at high FPS situation, so it must have something to do with 100Hz to 300Hz range.
Now I try use glue to make bead stable first then re-apply to cable see if helps.

Future me here Edit, same or non-human sensible difference.

I also tried glue, it work in some degree, GPU is make less noise now and noise is focused on PSU now.

Coil Whine and extreme buzzing on HX1000 under GPU load

First of all sry for my bad english, but its not my first language, and secondly I am new to the Corsair Forum, because I would rly need some help and your opinions. Over the last three months I built a new personal system, with the help of a friend who has done dozens of systems and knows what h...

forum.corsair.com

Like this situation. JG said to that thread maybe your PC is possessed or something

Added some Edit from future me.

 

[Mussels]

Facebook- Mmmm magnets.
Specifically, magnetic fields made visible
No idea what the full video is describing as its a small snippet of someone elses longer video, but the images are fantastic

Donut shaped magnet vs cube magnets - shows a good example of how the magnetic fields spread out from a cube shape, while those ferrite beads being a toroid contain the field

This is why the internals of those ferrite chokes are in a toroid shape in the first place

But to save space they can be a less effective design like this, that cause a greater disturbance

God i love practical science

This article mention some way was not in this thread. the cable mod. Writer suggest put 100μF up to 3000μF in cables see if it work.

I did mention those earlier, they used to be included with various power supplies such as those from OCZ when GPU's used molex connectors
I tried finding pre-made filtered extensions, but with no luck.

The bead sing and dancing and roaring around the PCIe cables.

PLEASE take a video of that dancing ferrite bead as that sounds amazing!

 

[Yougotonejob]

Mussels​

before ferrite bead

http://imgur.com/oPTJ2HN

After ferrite bead

http://imgur.com/564SzKa

I took it off today earlier. As the before video state. But earlier when I put it on it shake like dancing and I put on both cable this time. if on one cable it does not have enough cable to expand to fill the space to limit the movement. it would go dancing.
I guess it has some effect with this frequency. You can see FPS and different map different render time. the FPS range is 150 to 240

 

[lexluthermiester]

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

You're just getting no love with that card.. I feel for you.

Last week, I worked on a Gigabyte 3070 that had whine. Did my usual thing and it made no difference at all, it almost sounded louder... This gal had excellent hearing and it was driving her nuts, so we swapped her out for a different card(EVGA model) for the difference in price.

It happens. This procedure isn't a 100% fix all. It works much of the time, but not always..

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.

You're welcome man! I wish it worked 100% for everyone.

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.

For the record, this is very uncommon and if you use good ventilation(as you should be anyway when you're soldering), you will never encounter a danger.

Mussels​

before ferrite bead

http://imgur.com/oPTJ2HN

After ferrite bead

http://imgur.com/564SzKa

I took it off today earlier. As the before video state. But earlier when I put it on it shake like dancing and I put on both cable this time. if on one cable it does not have enough cable to expand to fill the space to limit the movement. it would go dancing.
I guess it has some effect with this frequency. You can see FPS and different map different render time. the FPS range is 150 to 240

Ok, that was freaky! That is likely because of the location of the ferrite choke on the cable. What happens when you mount it closer to the PSU or GPU connector?

 

[Yougotonejob]

Ok, that was freaky! That is likely because of the location of the ferrite choke on the cable. What happens when you mount it closer to the PSU or GPU connector?

same

 

[Mussels]

Ok, that was freaky! That is likely because of the location of the ferrite choke on the cable. What happens when you mount it closer to the PSU or GPU connector?

This, it should be right before the GPU
It's almost like its loud and cuts out when you press down on it - like it needs a tigther fit. At no point could or should this alter FPS in any way, GPU's dont power limit like that

 

[lexluthermiester]

same

Damn odd. A ferrite choke should never do that... Looks like you're better off without it on.

This, it should be right before the GPU

Yeah, that's where mine are located.

 

[Mussels]

I wish i had a test system like that, i'd love to try it with a cable tie to ensure it's tightly connected - and then i'd run a ground wire from the case to the ferrite part of it, to see if that helps
The beads also have different properties - it's not something i know what a good or bad value is, but it's possible you'd need a larger one, or one with different specs for such a high powered GPU

This quote for example says the beads have a current limit (Wait, is this how Ammeters work?

Good old wikihow has some great example images
How to Use Ferrite Beads: 9 Steps (with Pictures) - wikiHow

A tight fit matters - your noise issue may be that it's loose inside the bead, letting it vibrate exactly like the coils do

This is not what i expected to find, googling PCI-E power filter (Expecting a cable with built in resistors/capacitors etc)

They used to be there from the factory - the link there is dead tho, silverstone changed their website.

This exists, and is what i was seeking. I remember these being in factory PSU cables on molex connectors

https://www.aliexpress.com/item/1005005521585378.html

You can just feel the howling being reduced already (it might even work)

And the same thing on ebay
GPU 8PIN(6+2) Card Extension Cable with Capacitor Eliminate Video Card Whis X9R9 | eBay

The EVGA G2 1300W has these, shrink wrapped in the stock photos

 

[Chocostick]

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!!!

Ok so I decided not to use the silicone.
Problem is I am in Europe and we do not have the same brands of superglue.

I found this one: Secondglue cyanacrylate super glue 20g UV Protect bottle thin liquid waterproof quick-drying https://amzn.eu/d/c7sai1f

It says solvent free and liquid. However they say liquid like water, should I go for maybe a bit more viscosity or do you think it’ll be fine if I try to be a bit careful?

 

[lexluthermiester]

I found this one: Secondglue cyanacrylate super glue 20g UV Protect bottle thin liquid waterproof quick-drying https://amzn.eu/d/c7sai1f

Had to look it up, but yes that stuff should be ideal!

It says solvent free and liquid. However they say liquid like water, should I go for maybe a bit more viscosity or do you think it’ll be fine if I try to be a bit careful?

No. Low viscosity is important. You need to let the glue wick into the empty spaces around the noisy parts. You can't get a good seal if you don't get the wicking effect.

 

[Chocostick]

Had to look it up, but yes that stuff should be ideal!

No. Low viscosity is important. You need to let the glue wick into the empty spaces around the noisy parts. You can't get a good seal if you don't get the wicking effect.

Fantastic, I’ll give it a try!
I’ll let you know how it goes.
Thanks!

 

[lexluthermiester]

Fantastic, I’ll give it a try!
I’ll let you know how it goes.
Thanks!

BTW, the post with the all red, bold large font that you quoted: I wasn't yelling AT you. It was more a need to strongly assert a critically important point. I hope you didn't take offense or take it personally! That statement was not at all intended to be personal or malicious.

 

[Chocostick]

BTW, the post with the all red, bold large font that you quoted: I wasn't yelling AT you. It was more a need to strongly assert a critically important point. I hope you didn't take offense or take it personally! That statement was not at all intended to be personal or malicious.

Haha no worries, I am glad it prevented me from making a potential mistake.

BTW, the post with the all red, bold large font that you quoted: I wasn't yelling AT you. It was more a need to strongly assert a critically important point. I hope you didn't take offense or take it personally! That statement was not at all intended to be personal or malicious.

Please look at the picture below from my GPU (RTX 2080 ti).

I am planning to glue the chokes which are outlined in red, but I was wondering if I should do the ones in yellow as well, I have no clue what they are meant for:

The superglue will arrive in a week.

Thank you in advance!

 

[Yougotonejob]

I am planning to glue the chokes which are outlined in red, but I was wondering if I should do the ones in yellow as well, I have no clue what they are meant for:

the left hand side one of these are your 1.8v for memory one of them is for PCIe interface and the top one of them on the top left i dont know. But i know non of these need tons of power and they should be quite ok to begin with.
on the right these are the 12v for the input side of VRM, also acting the same way when power ground return back to PSU. Them are here to stable the current 12V rails come in. The R value on it the higher they are the better to your power supply. They don't change phase as much as the red box ones do.
Oh right, one more reason to leave the right hand side is half of the bottom PCIe lanes are ground. So without any chip like regulator to force them go back to PSU then they go back to mother board via the bottom PCIe lanes.
So my suggestion is leave the yellow ones for 1st round.

 

[Chocostick]

the left hand side one of these are your 1.8v for memory one of them is for PCIe interface and the top one of them on the top left i dont know. But i know non of these need tons of power and they should be quite ok to begin with.
on the right these are the 12v for the input side of VRM, also acting the same way when power ground return back to PSU. Them are here to stable the current 12V rails come in. The R value on it the higher they are the better to your power supply. They don't change phase as much as the red box ones do.
Oh right, one more reason to leave the right hand side is half of the bottom PCIe lanes are ground. So without any chip like regulator to force them go back to PSU then they go back to mother board via the bottom PCIe lanes.
So my suggestion is leave the yellow ones for 1st round.

Thanks for the information.
Is there any down side to doing them all at once (red and yellow)?

I’d rather do it all in one time than maybe having to dismount my GPU several times.

 

[Yougotonejob]

Down side is glue is almost irreversible mod. If in the future you need to do some other mod they are kind of difficult to deal with.
I think this point is up to you to decide.

 

[Nonam3]

@Yougotonejob I had exactly same issue with ferrite beads when used one that is larger than cable diameter (location just before GPU, tested different locations like 1 - 5 - 10 cm). Putting more cables did made it more silent (not entirely) so I believe this is issue with either cable dancing inside ferrite or ferrite itself being not securely attached to the plastic housing. Also, as was already pointed out here, those are for different frequencies and maybe that causes resonance, dunno.

Also purchased 12VHPWR cable for my PSU (be quiet! Dark Power PRO 12 1500W) so I can remove 2 thick cables and will add it and one 4 pin mobo cable so I will be re-testing ferrite beads.

GPU: after few days (2-3) still works, there is coil whine under heavy load, but for rendering works surpisingly ok, gaming @60FPS is ok (some whine depending on game), compute noisy but not whines.

 

[Yougotonejob]

GPU: after few days (2-3) still works, there is coil whine under heavy load, but for rendering works surpisingly ok, gaming @60FPS is ok (some whine depending on game), compute noisy but not whines.

Sorry I could go back and check your post, but i just ask here. what else did you do to make it not whine?
you did glue but no difference. you did bios and it has yield some result right?
EDIT one more question you having a MSI 4090?

 

[Nonam3]

@Yougotonejob I have MSi Gaming X Trio 4090 - purchased used to "save" money. GPU is on water block (custom water loop) using Barrow block.

Recap what I did and tried:

1. thermal pads of different sizes to increase pressure - no difference, sometimes it whined even on idle (depending on pad thickness and PCB bend)!
2. re-assembled GPU and testing with stock cooler - no difference
3. ferrite beads - no difference (plus ferrite did whined when used incorrect size)
4. undervolt - anything above 825 mV causes coil whine and at that voltage performance is basically 3090 (my old GPU)
5. trying different power outlets (I suspect same line) and un-pluging all that I could - no difference
6. glue 1 - as in OP post, glue under inductors, wait 24h to dry - no difference
7. glue 2 - put glue on mosfets, wait 24h - no difference
8. glue 3 - put glue on side of the inductors in between them, thicker glue, wait 24h - testing no difference
9 flashing different bioses - initially no difference, keapt and still using Gigabyte Gaming OC
-> 10 wait next 24h GPU in case - no / less coil whine (depends on workload, Furmark no coil, Everspace 2 when > 60FPS cap coil whines the more FPS there is, OpenPose (compute) no difference same as before same as 3090)

How I do test: quick testing mostly, as I do put in GPU only to see if it made some difference. Furmark + OpenPose after that different workload (like games)
The strange part is that Blender workload don't cause coil whine while other workloads even at 150W cause coil.

 

[lexluthermiester]

Haha no worries, I am glad it prevented me from making a potential mistake.


Please look at the picture below from my GPU (RTX 2080 ti).

I am planning to glue the chokes which are outlined in red, but I was wondering if I should do the ones in yellow as well, I have no clue what they are meant for:
View attachment 296659

The superglue will arrive in a week.

Thank you in advance!

I usually do every choke present on the card. Saves time in the long run.

Down side is glue is almost irreversible mod.

But it's also harmless.

 

[Mussels]

It says solvent free and liquid. However they say liquid like water, should I go for maybe a bit more viscosity or do you think it’ll be fine if I try to be a bit careful?

It would have to list a maximum temperature for us to know how well it would work, not all list that

The strange part is that Blender workload don't cause coil whine while other workloads even at 150W cause coil.

It's usually uneven loads that cause whine
Example 3090 image - imagine a load that exclusively used INT32, while the VRM's are spread out so 1/3 of them powers INT32, the next third FP32, and the final third the tensor cores.
That means 1/3 is working at 100% while its neighbours idle, and they didnt design it for that so you get whining
(This is how you get uneven voltages and commons between the parts, when they arent sharing the workload)

You've seen the designs of how a motherboard might have 12 phases - imagine a workload where only two are used really hard, while the rest are ignored. That's the sort of situation that causes whine - and why things like a games load screen with an uncapped FPS cause whine (One part of the GPU can suddenly run extremely fast, while the rest of it is idling and using nothing)

The over-worked VRM's get noisy, and make more magnetic interference which then may spread out - the dancing ferrite bead is a perfect example, of how current in one location can make something else nearby move, if not secured fully.

 

[RDNAuser]

Just re-registering to show my results with ASUS TUF OC 7900XT.

This glue is not perfectly fluid it is slightly thick and stringy. But I applied a heck of a lot.

Reassembled and up and running, glue is not fully dried but I gave it 30 mins, put Borderlands 3 no Vsync on 370 FPS on the menu, the coil whine still exists but massively reduced, once it dries it will likely be almost non existent.

So a positive change.