Firstly, understand the high-pitched noise can be caused by vibrations in electromagnetic coils. When a current passes through an inductor (an electronic component), it causes its copper wire to vibrate against its core with a high, audible frequency. This annoying sound is what we refer to as coil whine.
This can be fixed by simply:
1) Securing the inductor coils.
2) See if limiting the framerate, changes the amount of whine. Example, if you are in a menu with vsync off and getting 999fps, make adjustments.
3) Use a program such as MSI afterburner to change fan speeds to see if the voltage change/cooling makes a difference. -> Power limiting, overclocking, underclocking and undervolting. Test to see what makes a difference and correct the culprit.
4) All failing this -> Insulating the PC with sound dampening foam.
Alternatively, whatever size or piece of equipment I make and improve custom components. If you have the money I can provide a solution to almost any problem.
Firstly, understand the high-pitched noise can be caused by vibrations in electromagnetic coils. When a current passes through an inductor (an electronic component), it causes its copper wire to vibrate against its core with a high, audible frequency. This annoying sound is what we refer to as coil whine.
This can be fixed by simply:
1) Securing the inductor coils. They seem to be sealed.
2) See if limiting the framerate, changes the amount of whine. Example, if you are in a menu with vsync off and getting 999fps, make adjustments. Doesn't work. The frequency of the whine only spaces out. (Zzzzz to Z-z-z-z-z, matching frame rates)
3) Use a program such as MSI afterburner to change fan speeds to see if the voltage change/cooling makes a difference. -> Power limiting, overclocking, underclocking and undervolting. Test to see what makes a difference and correct the culprit. Tried this, doesn't work.
4) All failing this -> Insulating the PC with sound dampening foam. Did this with a thermal pad, worked sort of.
Alternatively, whatever size or piece of equipment I make and improve custom components. If you have the money I can provide a solution to almost any problem.
Sadly this is a Laptop, the Asus Scar 18 specifically. Seems that they simply dropped the ball with their power solution and nothing can be done. It's kind of drowned out when you have the fans running, but it makes Silent Mode when you want it for those moments of winding down not so very silent.
I guess nothing can really be done.
Unless you know a sound insulating foam that's safe to squash against a motherboard beneath the keyboard in a laptop chassis with only 5mm space? ;d
Think i got another clue into the coil whine shenanigans:
PSU's always have a happy range they're at their most efficient, and you're likely to get less noise within that range. It's possible going too far above or below that happy spot will increase noise - this explains why some people find FPS caps help and others do not, as a high wattage CPU for example could be pushing their PSU past the happy spot already (or a low wattage CPU on a 1500W PSU, it's below the happy range)
This would also explain the conflicting reports of gold/platinum PSU's helping a lot of people but not all of them, as it increases the odds of a more efficient PSU while some PSU's *just* quality for their 80+ rating, while still being overall cruddy PSUs
Seems like components are rated for noise suppression in deciblels - so they literally do mean noise noise, just outside of human hearing. When it passes outside the range those parts can suppress, the noise can become high freqency enough for humans to hear - and it's tied into ripple that's measured in PSUs, as those ripple values getting larger makes it harder to suppress said noise
a. Increase the operating frequency of the switching power supply to increase the high-frequency ripple frequency, which is conducive to suppressing the output high-frequency ripple.
b. Increase the output high-frequency filter to suppress the output high-frequency ripple.
I've seen VRM settings in some boards BIOS with control of VRM switching frequencies, i'm unaware of the drawbacks to doing so but it may be as simple as raising those values on certain boards can help. I assume various tweaks like disabling C-States is a method of artificially doing this, as higher power draw forces higher workloads to keep them in the higher working states?
This part is again beyond my current knowledge, but common mode noise is what the ferrite bead chokes remove:
5,Common Mode Ripple Noise
Due to the parasitic capacitance between the power device and the radiator bottom plate and the primary and secondary sides of the transformer, and the parasitic inductance of the wire, when the rectangular wave voltage acts on the power device, the output end of the switching power supply will generate common mode ripple noise. Reduce and control the parasitic capacitance between power devices, transformers and chassis ground, and add common mode suppression inductance and capacitance on the output side to reduce the output common mode ripple noise.
So add a ferrite choke between the devices AND on the AC cable to the PSU??
7,UHF resonant noise
The ultra-high frequency resonance noise mainly comes from the diode junction capacitance during reverse recovery of the high-frequency rectifier diode, and the resonance of the power device junction capacitance and the line parasitic inductance when the power device switches. The frequency is generally 1-10MHz. Measures such as switching tubes with small capacitance and reducing wiring length can reduce ultra-high frequency resonance noise.
Shorter cable length can remove UHF noise - i wonder how this goes with pigtail PSU cables
As an example with VDSL internet (extremely common in the UK and AU), having more than one phone socket in the house seriously harms performance as the signal goes past your modem, hits the unused far socket(s) and bounces back as out of sync interference. The unused section also picks up AC electrical interference as radio signals, which interferes with the signal. Common mode chokes are meant to filter this out, but can be over-whelmed.
Using the end cables could therefore help if it follows those rules, and ironically so could using both plugs of a single cable instead of one cable per pair (as we all advise for high wattage GPU's to prevent them melting, these days)
Assuming it follows the same rules, using the end connectors would be less noisy - but cutting them off entirely and using the closer connector, would be best overall
(Or using after market cables with just one straight connection)
Varies by country - here, they'd have to prove the user modification caused the fault.
As an example, if the card was recalled due to safety issues over the 12 pin connector, the user could send in a naked card with the stock cooler in a seperate box because it had been watercooled - and it'd be accepted because the users actions did not contribute to the reason for warranty (whatever design fault caused the product to be not fit for purpose)
Think i got another clue into the coil whine shenanigans:
PSU's always have a happy range they're at their most efficient, and you're likely to get less noise within that range. It's possible going too far above or below that happy spot will increase noise - this explains why some people find FPS caps help and others do not, as a high wattage CPU for example could be pushing their PSU past the happy spot already (or a low wattage CPU on a 1500W PSU, it's below the happy range)
This would also explain the conflicting reports of gold/platinum PSU's helping a lot of people but not all of them, as it increases the odds of a more efficient PSU while some PSU's *just* quality for their 80+ rating, while still being overall cruddy PSUs
That sounds very reasonable. I'd be very inclined to lend it towards credibility, the problem is proving it. With so very many CPU/GPU/Mobo/PSU combinations possible, it would be an extremely daunting task to prove.
Big edits to that post as i read up more on the electrical noise, what a fantastic document/resource
Key highlights:
1. Stick a common mode choke (ferrite bead) between the two noisy components at the receiving GPU end, AND one on the AC power cable right before it enters the PSU (to remove noise BEFORE it goes AC-DC). This can also help filter noise from your ground.
2. Try alternating which connector you use from pig-tail power cables, shorter cables are better in general, but un-used cabling picks up more interference as noise than active cabling (this is what the choke catches, and reduces)
3. Better PSU's create less noise and supress more noise
4. DC-DC suppresion via a choke on the GPU cable wont help if the noise is from the PC converting noisy AC-DC, or a noisy ground. Put a choke on the power cable too. They used to be included on VGA and DVI cables, this *used* to be common, even in molex and PCI-E cables in PSUs but dropped off as everything went digital
Bear in mind that the AC signal will quite comprehensively saturate the ferrite with magnetic flux, and all you are doing is introducing a voltage drop.
Bear in mind that the AC signal will quite comprehensively saturate the ferrite with magnetic flux, and all you are doing is introducing a voltage drop.
Bear in mind that the AC signal will quite comprehensively saturate the ferrite with magnetic flux, and all you are doing is introducing a voltage drop.
They're each designed to catch specific frequencies - and only from the signals on the outer surface of the wiring, not the power flow internally.
EMI + copper wiring = AC voltage, even on a power cable being used for DC power. Small amounts we usually dont care about, until it reaches a point where it makes our devices scream at us.
One key to understanding how this works is that external interference travels on the outer skin of the copper wiring, NOT the inner part of it like a power sources does. It's known as the "Skin effect" - Common Mode chokes on the outside are only filtering out the noise skimming along the outside, not the actual power/data signal internally. This is why stranded wire is more prone to interference, as it has more surface area to collect noise.
When the signal reaches the end, it's still got to deal with that extra power, even if its a small amount - but because its a VARYING small amount, devices designed for DC currents can struggle with them.
(Twisting the wiring in pairs reduces how much straight material there is to absorb these signals, people forget how BIG radio signals are, the wavelength at 2.4Ghz from bluetooth, wifi and microwaves is 12.5cm
Conductive material needs to be 12.5CM or a fraction of (1/2, 1/4 1/8 etc) to reliably receive it, so twisted wiring prevents it receiving those signals.
From NBN co's common mode choke testing:
We have carefully measured the characteristics of home-made chokes such as those above in our labs and achieved 40dB and greater common mode attenuation and virtually zero differential mode attenuation (when the pairs are fully twisted)
The radio noise interference we get is common mode noise, where it effectively makes devices act like they have a voltage difference between their ground sources - because external power is being added 'wirelessly' via EMI even in small amounts, and has to be filtered back to a common value between the devices somehow
I'll spoiler tag below as it goes off into how FM radio and AC electrics cause interference to VDSL and how the chokes work there, for those that want the technical info - some of it is beyond my understanding, for sure
Above is the measured common mode impedance R,X of the choke. It is designed to peak between 3.5 and 7MHz to afford some moderately high impedance at both frequencies.
Because the common external noise that screws with VDSL is in the 3.5MHz to 7MHz range
This is the NVBco, australian government research into this and it's extremely thorough - and then the government changed and they ignored the advice and didn't friggin implement it. nbn-Amateur-Radio-guide-publication.pdf (nbnco.com.au)
The actual frequencies used by a VDSL2 service are not set in concrete but are selected by the node and modem in collaboration from the set of all possible tones between about 20KHz and 17MHz
They explain that the devices negotiate and choose what frequencies to use, and then random interference - such as from a home users FM radio transmitter randomly appears later - and suddenly those working frequencies have issues. The default choice is that if noise is detected, shut those frequencies off and don't re-use them until the device is rebooted.
The ferrite chokes allow the noise from within the wire to continue, but filter out the ones travelling on the OUTSIDE of the wiring allowing the modems to use the frequencies above that FM radio could randomly jam up allowing you to avoid those speed drops over time that are only fixed by rebooting the DSL modem.
When I mention AC noise I dont mean from the grid - I mean *local* within the house, on the same circuit - often in the same room or power strip, but usually just sharing some wiring internally in the walls.
In the US i keep hearing stories of how you guys do ground wiring totally different to everywhere else, and that can also result in a lot more electical noise - in some situations, no ground at all.
You guys cant tell me you've never heard the noise come out from a speaker when lights flicked on or off, or seen lights flicker when a fridge compressor turns on? that's the sort of noise it helps filter out, and cheap electronics and power bricks cause a *LOT* of high frequency short range noise that good old straight unfiltered power cables can receive - be they in the walls, or the PC.
Pure sine wave UPS only helps if they're always active, otherwise that's only relevant on battery mode - not many are always active and always filtering as it trashes the batteries.
Off topic, I'd like to give huge thanks to @Mussels for a head up about ferrite cores. My ADSL signal was terrible, with a huge noise margin, ping and speed drops whenever the network updates the IP (so I need to reboot to get higher ones).
Found a bunch of ferrite cores in a electric repair shop and decided to use all of them. So I've put one group just before the phone/adsl splitter, the other right after the splitter, and a smaller group just before the router/modem header.
The results are great. Ping dropped from around 40-50ms to 3-10 range, noise margin is up to 30% lower, and it doesn't reset to the lowest speed on each IP change automatically.
On topic, super glueing my 6900XT didn't eliminate the whine, but just lowered the frequency of the pitch.
I will try to get some bigger ferrite cores, preferably the split ones, so I can experiment by putting them on the GPU power cables.
No. Ferrite chokes can not do harm. That's not how they work. Physics is a thing. You're understanding of how they work seems to need some research. Sadly, I'm not going to write a dissertation on the subject. The simple point being that Ferrite Chokes filter out "noise" in whatever signal passes through them. They can not and do not cause harm.
That would indeed be a great solution in a perfect operating environment. Sadly, most home are far from perfect. Often they are an electromagnetic soup of noise. Ferrite chokes at worst will do nothing, and at best help greatly. A UPS that filters output power is a wonderful, but expensive, solution.
They're each designed to catch specific frequencies - and only from the signals on the outer surface of the wiring, not the power flow internally.
EMI + copper wiring = AC voltage, even on a power cable being used for DC power. Small amounts we usually dont care about, until it reaches a point where it makes our devices scream at us.
One key to understanding how this works is that external interference travels on the outer skin of the copper wiring, NOT the inner part of it like a power sources does. It's known as the "Skin effect" - Common Mode chokes on the outside are only filtering out the noise skimming along the outside, not the actual power/data signal internally. This is why stranded wire is more prone to interference, as it has more surface area to collect noise.
When the signal reaches the end, it's still got to deal with that extra power, even if its a small amount - but because its a VARYING small amount, devices designed for DC currents can struggle with them.
(Twisting the wiring in pairs reduces how much straight material there is to absorb these signals, people forget how BIG radio signals are, the wavelength at 2.4Ghz from bluetooth, wifi and microwaves is 12.5cm
Conductive material needs to be 12.5CM or a fraction of (1/2, 1/4 1/8 etc) to reliably receive it, so twisted wiring prevents it receiving those signals.
From NBN co's common mode choke testing:
Notice how it affects common mode signals by 40db, but differential mode is 0db. You can guess which of those is the one we don't want filtered.
View attachment 293651
The radio noise interference we get is common mode noise, where it effectively makes devices act like they have a voltage difference between their ground sources - because external power is being added 'wirelessly' via EMI even in small amounts, and has to be filtered back to a common value between the devices somehow
I'll spoiler tag below as it goes off into how FM radio and AC electrics cause interference to VDSL and how the chokes work there, for those that want the technical info - some of it is beyond my understanding, for sure
Because the common external noise that screws with VDSL is in the 3.5MHz to 7MHz range
This is the NVBco, australian government research into this and it's extremely thorough - and then the government changed and they ignored the advice and didn't friggin implement it. nbn-Amateur-Radio-guide-publication.pdf (nbnco.com.au)
They explain that the devices negotiate and choose what frequencies to use, and then random interference - such as from a home users FM radio transmitter randomly appears later - and suddenly those working frequencies have issues. The default choice is that if noise is detected, shut those frequencies off and don't re-use them until the device is rebooted.
The ferrite chokes allow the noise from within the wire to continue, but filter out the ones travelling on the OUTSIDE of the wiring allowing the modems to use the frequencies above that FM radio could randomly jam up allowing you to avoid those speed drops over time that are only fixed by rebooting the DSL modem.
When I mention AC noise I dont mean from the grid - I mean *local* within the house, on the same circuit - often in the same room or power strip, but usually just sharing some wiring internally in the walls.
In the US i keep hearing stories of how you guys do ground wiring totally different to everywhere else, and that can also result in a lot more electical noise - in some situations, no ground at all.
You guys cant tell me you've never heard the noise come out from a speaker when lights flicked on or off, or seen lights flicker when a fridge compressor turns on? that's the sort of noise it helps filter out, and cheap electronics and power bricks cause a *LOT* of high frequency short range noise that good old straight unfiltered power cables can receive - be they in the walls, or the PC.
Pure sine wave UPS only helps if they're always active, otherwise that's only relevant on battery mode - not many are always active and always filtering as it trashes the batteries.
Bigger isn't always better. Try adding one after another and adding one to the input power cable going into the PC. After that any effect is going to be very limited.
I can't say really. Can't replicate the same environment, as I changed all the fans, and did other stuff to the system, and I'm lazy to do the paper cone thing. But I'll update when I do.
Bigger isn't always better. Try adding one after another and adding one to the input power cable going into the PC. After that any effect is going to be very limited.
