GeForce GTX 970 Coil Whine Concerns

[BiG StroOnZ]

The reason this comes up to today is that reports are surfacing of GeForce GTX 970 cards from various graphics card vendors exhibiting excessive coil whine or coil noise. These reports are coming in from multiple forum threads around the internet, a collection of YouTube videos of users attempting to capture the issue and even official statements from some of NVIDIA's partners. Now, just because the internet is talking about it doesn't necessarily mean it's a "big deal" relative to the number of products being sold. However, after several Twitter comments and emails requesting we look into the issue, I thought it was pertinent to start asking questions.

 

As far as I can tell today, GTX 970 cards from multiple vendors including EVGA, MSI and Gigabyte all have users reporting issues and claims of excessive coil noise. For my part here, I have two EVGA GTX 970 cards and an MSI GTX 970, none of which are producing sound at what I would call "excessive" levels. Everyone's opinion of excessive noise is going to vary, but as someone who sits next to a desk-high test bed and hears hundreds of cards a year, I am confident I have a good idea of what to listen for.

 

We are still gathering data on this potential issue, but a few of the companies mentioned above have issued official or semi-official statements on the problem.

 

From MSI:  

 

The coil whine issue is not specific to 900 series, but can happen with any high end GPU and that MSI is looking in to ways to minimize the issue. If you still have concern regarding this issue, then please contact our RMA department.

 

From EVGA:

 

We have been watching the early feedback on GTX 970 and inductor noise very closely, and have actively taken steps to improve this. We urge anyone who has this type of concern to contact our support so we can address it directly.

 

From NVIDIA:

 

We’re aware of a small percentage of users reporting excessive “coil whine” noises and are actively looking into the issue.

 

Since all of the GTX 970 cards currently shipping are non-reference, custom built PCB designs, NVIDIA's input to the problem is one mostly of recommendations. NVIDIA knows that it is their name and brand being associated with any noisy GeForce cards so I would expect a lot of discussions and calls being had behind closed doors to make sure partners are addressing user concerns.

 

Interestingly, the GeForce GTX 970 was the one card of this Maxwell release where all of NVIDIA's partners chose to go the route of custom designs rather than adopting the NVIDIA reference design. On the GTX 980, however, you'll find a mix of both and I would wager that NVIDIA's reference boards do not exhibit any above average noise levels from coils. (I have actually tested four reference GTX 980s without coil whine coming into play.) Sometimes offering all of these companies the option to be creative and to differentiate can back-fire if the utmost care isn't taken in component selection

.

Ironically the fix is simple: a little glue on those vibrating inductor coils and the problem goes away. But most of the components are sealed making the simple fix a non-starter for the end user (and I wouldn't recommend doing that anyway). It does point to a lack of leadership from board manufacturers that are willing to skimp on hardware in such a way to make this a big enough issue that I am sitting here writing about this today.

 

Source: http://www.pcper.com/news/Graphics-Cards/GeForce-GTX-970-Coil-Whine-Concerns

 

I figured since the original article had a poll, maybe we could conduct one here. I imagine our results should resonate with the rest of the internet community. I've seen a poll like this in the GPU section before, but I don't believe I've seen this actual article posted where a renowned reviewing website addresses the supposed problem. 

[Lewellyn]

My 970 from Gigabyte comes in today or tomorrow. Hopefully it doesn't have this issue, I'll have to compare it to my 760.

[IdortMasterRace]

https://www.youtube.com/watch?v=NUhwFpPKnXI

[SuperPug]

Probably because the 970's are powerful enough to push really high framerates in most games hence the whining under strain. Enable V-Sync or if not available use some sort of fps limiter.

[GoncaloMoreno]

I have a g1 gtx 970 and it only coil whines at framerates over 60 fps.

Usually in menu screens the fps are over 300 sometimes 4000, and there it coil whines like crazy.

In games such as farcry 3 , at my settings, i get 80-90 fps and it coil whines but when i turn on vsync it doesnt.

Conclusion, it only coil whines at fps >60.

[Tocsin_786]

my gigabytee 970 doesnt have it

[Jaybird]

It does seem to be more apparent from what I have seen compared to other past cards.

[Chelsi_Gamer]

Even when I overclocked my monitor to 120hz the coil whine wasn't very audible. But at 60Hz it isn't there.

[Exotath]

My 7950 had it, my GTX 780 had it, my 780 Ti has it....will find out monday if my 970's have it.

[garyjpaterson]

I havent noticed any whine from my MSI 970, and it seems pretty quiet overall tbh.

[Vrbas]

Watch that video. Harsh but true.

[Bigdaddy5414]

I have three 980 G1 gaming and the only time they do is when folding a core 15 work unit. And it's very minor. Other then that they do nothing. I guess they aren't 970s but similar.

[JAKEBAB]

Glad i skipped this gen lol id rather listen to an r9 290 vacum then hear coil whine.

[Redpalmtree]

Even when I overclocked my monitor to 120hz the coil whine wasn't very audible. But at 60Hz it isn't there.

wut

[Chelsi_Gamer]

wut

I ran my monitor at 120hz to achieve 120 visible foa. Coil whine wasn't audible.

[tovleander]

whoo "PCmasterrace" for ya'll there \o/ 


oh wait...  :rolleyes:

[tobben]

a much messier and more extensive poll can be found here : http://linustechtips.com/main/topic/229278-the-gtx-970980-coil-whine-thread/

Coil whine is extremely annoying.. i'd much rather have noisy fans.

[Mooshi]

https://www.youtube.com/watch?v=NUhwFpPKnXI

 

I love you for that. I used to think my PSU was shit, but after using Nvidia inspector, that obnoxious loud whining is gone!

 

Originally had a Fractal Integra R2 650w PSU that I RMA'd for coil whine. Fractal sent, on accident, a 750w unit and that whine was still there. You just saved me money on picking up a "quiet PSU"- @LinusTech should really do a video addressing this issue since it might also help others stop hating their hardware for the wrong reason.

[mr moose]

I've never had an issue with coil noise in pc parts. 

[FatalityDiablo]

https://www.youtube.com/watch?v=NUhwFpPKnXI

haha i loled so hard on those subtitles

 

edit: i never even had coil whine, but i guess that is just because of poor or a failing capacitors on your pcb. i often have 300+ frames in game and never notice any high pitched noise 

[Ziltoid_]

I have a reference 980 and it whines pretty bad during the windows 7 experience index test. But in games it's not noticeable.

[Snadzies]

I have a reference 980 and it whines pretty bad during the windows 7 experience index test. But in games it's not noticeable.

 

 

I haven't had a single card that didn't whine to some extend during the Windows 7 test, really.

 

 

My new MSI 4G 980 was the quietest as far as that goes while my GTX 760 was the worst.

My 6870 was somewhere in between and a couple 7850s I played around with seemed about the same as the 6870.

None of them had coil whine during normal gameplay unless there was a situation where the FPS would go into the several hundred / thousand range (Crysis splash screens and such).

I've long since gotten into the habit of setting an FPS limit of 150fps on my computers.

[Faa]

Well I'm actually one of those that have the upgrade fever, probably upgraded the GPU('s) like 15 times already and all of them were whining except one so I'm not surprised if my next card will whine. 

Also when the 600 series were released you had a bunch of people reporting coil noise, same thing with the 700 series although the 780 was so expensive & noisy that the report count was just so low. It has nothing to do with the 900 series at all. It's not because you hear coil noise that you'll be reporting it somewhere, lots of people just don't hear it or they just don't care or they think it's normal or they are running a Titan-Z for a 10 years old game like WoW where the GPU load is below 5-10%. Since the power consumption on 970/980's is so low that the noise output from the fans are almost virtually inaudible you'll hear the whining better than something like a R9 290x reference. I doubt anyone reported coil whine when the 290x reference was the only 290x available. Besides it also depends on how good your hearings are, a buddy of yours might be complaining about your PSU coil whining while you're hearing literally nothing not even from an inch away. There's no way to get rid of coils in switching power supplies, you really need them. What manufacturers usually do to prevent coil noise is glueing them but that's not done by hand and who knows there's no quality control for it at all (honestly it looks to be like this). 

I find the higher the power draw is the harder it whines, even to PSU's, try sliding the TDP to 10% in Furmark the noise would go away but once you slide it back to 100% the whining (if you have it) will be back. Cards with huge power draws like the 295 x2 are guaranteed to whine, even review samples (which they mostly don't report) mentioned it. Always better off avoiding a dual GPU imo if you want to have a chance of a card that will not whine.

We have a few fixes for this noise though. What you could do is; overclock your gpu, raise the tdp, mod your bios for a higher TDP, just do anything to make your card consume more power and run Furmark orsomething for 24 hours so the glue could break in although this is not guaranteed to work for every card. A fix that would definitely do the job; if your gpu sits at 99% load all the time in BF4 for example providing 60 fps, what you want to do is getting your GPU load lower so the power draw is lower too, simply just set the video settings to low and cap your FPS with vsync so the GPU load is around 20-30% orsomething. If you don't want to play on lower settings, simply get a 2nd card put it in SLI and as you already have 60 fps with a single card at 99% load assuming you use Vsync so the load is split across two cards so would be 50%/50%. In my experience cards at 50% weren't whining, if yours does aim for a lower load. You could go nuts about it and get 4 cards then the load of each card would be incredibly low. Expensive fix but you can't do much more than that. Most manufacturers don't offer a replacement just for coil noise, I tried lately with EVGA (actually I just hoped to get my 780 replaced with a 980) trying to RMA my card and they didn't want to. Had to BS a different excuse, but I just decided to just buy a 970 Strix rather than dealing with RMA's.

Shops like Amazon actually accept RMA's for any BS reason, so aim for those kind of shops.

Coil whine usually doesn't affect either the PSU's or the components' operation, at least not to a measurable degree, since the resonant frequencies are typically such that they're easily damped in other components, such as capacitors and other filter coils. I believe all manufacturers are actually aware of this issue, and are doing what they can to deal with it, most of the time.

Namely, there are three ways coil whine will develop. There's "self-whine" to which every coil under the Sun is susceptible (including various types of transformers) by it's very nature, and there's resonant/induced whine, which is a byproduct of resonance between VRMs on the motherboard and/or the graphics card, and the PSU's coil(s) and/or transformer(s). Let me elaborate further:

As the current passes through a coil, it creates a magnetic field, which in turn induces a current in the coil such that it tends to cancel out the change in the initial current. So if there's a constant 1A through a coil, then it jumps to 1.1A, the change in magnetic field will induce a current of -0.1A (meaning 0.1A in the opposite direction), restoring the net flow to 1A. This is how coils remove unwanted ripple/noise from the DC output of a PSU, or a DC input into a VRM.

Both the length of the coiling wire and the coil loop diameter are parts of the inductance equation, and are a variable just like the inductance is, and not constants. Well, theoretically they are constants for a given coil when it's effective inductance is calculated, but in real world, where approximations amount to a wrong result, the coil will shrink and expand under the influence of magnetostriction.

Self-whine or coil noise can be twofold - physical and electrical.

Physically, high frequency switching used in PSUs (50-150 kHz) will make the coil vibrate (from all the rapidly succeeding shrinking and expanding) at a lesser frequency, typically from one quarter to one eight of the switching frequency. This is sometimes well inside the audible range (~20 Hz - ~20 kHz, typically 30 - 18k). The lower frequency vibrations are a consequence of the finite velocity of current (rather, electrons) and the finite speed of expansion/shrinkage propagation through the coil. Not only that, but both the wire and the core are shrinking/expanding, and at a different rate and amplitude, so until everything aligns properly (rate and speed of shape change with the rate of propagation of the deformations), there will be no audible vibrations. This is part of the story.

Electrically, as the coil loops are moving and the core changes shape, both travel inside a varying magnetic field, which causes additional self-induced currents to appear. These are mostly damped out by other filtering elements, due to their very low magnitude and their relatively high frequency, but sometimes they manage to get to an amplifier in a sound card, for example, and show up as audible noise in the sound (sub)system. Additionally, every coil is a (poor) antenna for high-frequency signals (voltage changes), and it radiates those signals out into wires and PCB traces. There they are induced back from electromagnetic emissions into current and possibly amplified as per above.

The kicker is that physical noise can (and does, in larger inductors) cause electrical noise, and vice-versa. Further, any wire or other form of conductor (like a PCB trace) is also an inductor, albeit a poor one.

Resonant whine can develop between any two oscillatory systems, which coils are all by themselves, as is practically any circuit that contains them. VRM circuits on motherboards, graphics cards, hard drives, etc. pretty much always contain at least one. In order to have an electrical oscillator, you need an inductor and a capacitor. All inductors are also (poor) capacitors, and this doesn't present a problem at low frequencies, because they "see" capacitors as open circuits. Self-capacitance is a problem at high frequencies, exactly the situations where you'd want to use coils in the first place... When two coupled coils (either connected via wires and traces or magnetically coupled, or when the EM radiation of one permeates the other) reach very similar electrical self-noise frequency, the parasitic signals they produce may be (and usually are) amplified exponentially. This can, in rare cases, actually pollute the DC input/output, and there are actual cases in practice. There are some Sirfa-made PSUs in which simply moving output wires away from a regulator coil makes the PSU output voltage significantly less noisy. I still consider this a rarity, though, and it can be solved by putting a simple EMI shield (a piece of isolated metal sheet) around the offending coil or between it and the "polluted" area.

Coil whine can be lessened to an acceptable degree with a relatively simple fix. Just dampen the physical side of it by gluing or caulking down the coil, so that it's vibrations are absorbed. Another way is changing the current/voltage frequency, which is never easy, as it affects the electrical design of the device in question, or use a different coil. This could be a coil made of different materials, or of a different size, or even a different shape. I've seen coils made in the shape of the number 8 (or the infinity sign, if you need to be geeky about it), that produce significantly less noise than standard toroidal coils. I can't say how much this would add to the price, however. And let's not forget that transformers are, in effect, simply big-ass coils, and their whiny nature is much harder to deal with...

As for why coil whine would develop in time, instead of right from the get-go, well... Perhaps the dampening glue/caulk "breaks in"? Or maybe the coils very slightly change their basic (at-rest) shape, such that their resonance pattern shifts into the audible range? Who knows, it's a very complex phenomenon.
 

I've never had an issue with coil noise in pc parts. 

Well even if the coil whine is there, you might not be able to hear it but someone else would hear it from 5 meters away. It just needs to operate out of your hearing range. Our TV in the living room actually buzzes hard, they're not willing to buy a new TV just because I'm the only one being able to hear it.
 

 

edit: i never even had coil whine, but i guess that is just because of poor or a failing capacitors on your pcb. i often have 300+ frames in game and never notice any high pitched noise 

Caps aren't coils though. 99% of the time it's coils that are whining.
 

 

I have a reference 980 and it whines pretty bad during the windows 7 experience index test. But in games it's not noticeable.

It just depends on what the switching frequency of the vrm is which is variable with the type of load. 

[mr moose]

 

 

Well even if the coil whine is there, you might not be able to hear it but someone else would hear it from 5 meters away. It just needs to operate out of your hearing range. Our TV in the living room actually buzzes hard, they're not willing to buy a new TV just because I'm the only one being able to hear it.

 

 

I have my hearing tested every year,  plus I frequently carry out Fourier analysis on my room.  I can graph the frequency and SPL of my pc from  10cm and 100cm if I really wanted.  So that's not really a consideration.

[Lewellyn]

I got my 970 G1 up and running with no big coil whine thus far. Also that youtube video was just hilarious.