3070ti FE Coil Whine

[Rowan88]

Hi,

I have a 3070 ti Founders Edition and I have noticed the coil whine on it is quite loud. Anything from 60fps up you hear it like a bee. If I turn my yamaha hs5 speakers on it also amplys through those too.

The card is about a week or so old and I have only started proper gaming on it.

I read some stuff with people saying it may take a few weeks to "break in" and stop.

Is there anything else I can do or try to speed up this process or check if its a defect card worth the RMA?

Cheers

 

[Sean473]

You can run some intensive benchmarks such as heaven or 3D Mark on loop - Not worth RMAing in the current climate for a new card as you might get a refurb where you have a perfeclty working new one..

 

[uscool]

depends how much it bugs you I returned 3080FE for coil whine loudest I've ever experienced got replacement within 5 days also has it but not as bad and gets drowned out once the fans kick in, we really shouldnt be having to live with it when someone else could have one without coil whine doesnt seem right to me

 

[GuruJockStrap]

I have a 3070ti FE.

Mine has coil whine but once the fans in my case spin up and I have headphones on it doesn't bother me.

Its silent on the desktop so thats all I care about.

I have undervolted my card so its pulling around 200w at load, as opposed to stock where it was around 280w. Think that's helped reduced the whine to be honest, but not eliminate it.

My previous card had coil whine at load, so its not an issue for me.

 

[Rowan88]

You can run some intensive benchmarks such as heaven or 3D Mark on loop - Not worth RMAing in the current climate for a new card as you might get a refurb where you have a perfeclty working new one..

How long should I leave it running?

I have a 3070ti FE.

Mine has coil whine but once the fans in my case spin up and I have headphones on it doesn't bother me.

Its silent on the desktop so thats all I care about.

I have undervolted my card so its pulling around 200w at load, as opposed to stock where it was around 280w. Think that's helped reduced the whine to be honest, but not eliminate it.

My previous card had coil whine at load, so its not an issue for me.

I will maybe try undervolting but I dont want to underclock as I am paying for speed I cannot use.


What about the speakers side of things?
I have a Scarlett 2i2 USB Interface with the speakers hooked into that. When I have then on its like a microphone for the whine.

Anything I can do to stop that as that is were the majority of the noise is heard while gaming.

I read up a lot of it last night and the most common thing showing was that the interference was travelling through the USB cable connected to the rear IO of motherboard. It said if you used a mains powered usb hub this can eliminate the issue. So I have one of those getting delivered today to try.

If that doesn't fix the issue I can only put it down to being a ground issue and will return the usb hub and buy a Behringer HD400Hum Destroyer and see if that fixes the problem...

Will post an update later in case this helps others..

 

[GuruJockStrap]

I will maybe try undervolting but I dont want to underclock as I am paying for speed I cannot use.

You shouldn't have to sacrifice speed.

I don't have the the exact details from from my testing results but I shaved off a chunk of voltage but the core clock speeds are nailed on at 1950MHz when underload. It is stable and does not budge at all. Temps hover around 72 under load.

At stock, wattage was higher, which pushed temps up meaning the boost core clock would fluctuate around 1950MHz.

I tested with benchmarks and with the undervolt settings, I'm getting virtually the same scores. No loss of performance but I have a cooler, lower wattage card that hits boost clocks and is stable.

I could probably tweak it further (either reduce voltage further or overclock) but for an hours worth of effort, I just wanted stock speeds but with lower power draw and temps.

I've had a coupe of Vega cards in the past and they exhibited coil whine at load. That drew a lot of power for its time (funny how people lamented AMD at the time for it yet these 30xx cards take it to another level), so unfortunately I feel its just an annoying result of modern GPU's thirst for power.

 

[uscool]

How long should I leave it running?


I will maybe try undervolting but I dont want to underclock as I am paying for speed I cannot use.


What about the speakers side of things?
I have a Scarlett 2i2 USB Interface with the speakers hooked into that. When I have then on its like a microphone for the whine.

Anything I can do to stop that as that is were the majority of the noise is heard while gaming.

I read up a lot of it last night and the most common thing showing was that the interference was travelling through the USB cable connected to the rear IO of motherboard. It said if you used a mains powered usb hub this can eliminate the issue. So I have one of those getting delivered today to try.

If that doesn't fix the issue I can only put it down to being a ground issue and will return the usb hub and buy a Behringer HD400Hum Destroyer and see if that fixes the problem...

Will post an update later in case this helps others..

at stock the clock is all over the place even if you see high boost clock it will only do it for few seconds and drop because of heat with undervolting you can sustain clock speed so giving overall better performance

Undervolting you can increase performance over stock while benefiting with better temps/ lower power and can lower fan speed, Whenever I undervolt I run benchmarks at stock note down the scores and temps/power and do the same undervolted my goal is to match stock or beat stock performance

 

[Rowan88]

You shouldn't have to sacrifice speed.

I don't have the the exact details from from my testing results but I shaved off a chunk of voltage but the core clock speeds are nailed on at 1950MHz when underload. It is stable and does not budge at all. Temps hover around 72 under load.

At stock, wattage was higher, which pushed temps up meaning the boost core clock would fluctuate around 1950MHz.

I tested with benchmarks and with the undervolt settings, I'm getting virtually the same scores. No loss of performance but I have a cooler, lower wattage card that hits boost clocks and is stable.

I could probably tweak it further (either reduce voltage further or overclock) but for an hours worth of effort, I just wanted stock speeds but with lower power draw and temps.

I've had a coupe of Vega cards in the past and they exhibited coil whine at load. That drew a lot of power for its time (funny how people lamented AMD at the time for it yet these 30xx cards take it to another level), so unfortunately I feel its just an annoying result of modern GPU's thirst for power.

Dont supose you could share the settings I should change to test this?
Mine under load stock is around 79-80 deg.

 

[Rowan88]

Not sure if I done this right...

I opened MSI Afterburner and let it run while I ran Heaven Valley Benchmark on Ultra 1440p. During the benchmark running I noticed the core clock was 1890Mhz with temps of 79 deg.

I installed MSI Afterburner, Pressed CTRL+F to open Curve Editor. In the Core Clock MHz I done "-250" then hit enter. I then dragged the 900mV point up to the 1890Mhz point and hit APPLY.

I then ran the benchmark again and noticed that the core clock was still 1890Mhz but the temps are now 69-70 deg which is a big improvement.

I do notice that the coil whine isn't as loud but still there.


I then opened Back 4 Blood and stared into the Sky with 300+ FPS and the horrible coil whine (bee mode) is still clearly noticeable through the speakers. So hopefully using the mains powered USB hub will sort this issue as if the buzz wasn't coming through the speakers you wouldn't hear it inside the case at all.

 

[Rowan88]

Got the USB mains powered hub but the buzzing was the exact same, so wonder if its a ground loop issue. May need to try the HD400 Hum Destroyer and see if that fixes it.

 

[GuruJockStrap]

Sounds like you've applied the underclock correctly.

You'll probably find with the reduced voltage and temps, the core clock will go higher. Sounds like you did some quick testing and you can always go back to tweak around to find the sweet spot.

As you've found, coil whine is reduced but its still there.

 

[Rowan88]

Sounds like you've applied the underclock correctly.

You'll probably find with the reduced voltage and temps, the core clock will go higher. Sounds like you did some quick testing and you can always go back to tweak around to find the sweet spot.

As you've found, coil whine is reduced but its still there.

I had an issue with either Heaven closing or the game closing so assumed it was unstable.

Surely if I lower to 900mV using the standard 1890Mhz it should work yes? And if I kept increasing the mV it would end up close to where it was before. The other hand is reducing the Core Mhz but then that means I am making the card weaker?

I will retest again, I done the Extreme Settings (1600x900) and it shows 1920MHz so I set that on the 900mV and it passed Heaven fine. Will try TimeSpy then test some games.

 

[GuruJockStrap]

I'll share my settings on here a bit later.

I can't remember what mine were as I did it on day one.

So basically my settings:

1. input -200 on the Core Clock
2. In voltage/frequency window (the one that opens with ctrl+f), I clicked the little dot that corresponds to 900mV, and dragged it up to 1965 mhz and applied

The difference was:

Stock
Max Clock: 1860-1875
Avg Watt: 286
Max Temp: 81
Fan Speed: 57%
Heaven Score: 4324

Undervolt
Max Clock: 1950
Avg Watt: 215
Max Temp: 72
Fan Speed: 50%
Heaven Score:4332

Ok, so the max clock is running a bit higher for the same performance. But my goal was to match perfomance but with less voltage, heat and fan noise. So achieived my goal.

I'm sure if I spent more time I could find a better sweet spot but I'm happy with it.

 

[Rowan88]

I'll share my settings on here a bit later.

I can't remember what mine were as I did it on day one.

Appreciate it m8, Right now I have it set to 1905Mhz @ 925mV. So far it passed 3 runs of Heaven, 2 of TimeSpy and a little bit of gaming. Max temps around 74 deg.

 

[Rowan88]

I'll share my settings on here a bit later.

I can't remember what mine were as I did it on day one.

So basically my settings:

1. input -200 on the Core Clock
2. In voltage/frequency window (the one that opens with ctrl+f), I clicked the little dot that corresponds to 900mV, and dragged it up to 1965 mhz and applied

The difference was:

Stock
Max Clock: 1860-1875
Avg Watt: 286
Max Temp: 81
Fan Speed: 57%
Heaven Score: 4324

Undervolt
Max Clock: 1950
Avg Watt: 215
Max Temp: 72
Fan Speed: 50%
Heaven Score:4332

Ok, so the max clock is running a bit higher for the same performance. But my goal was to match perfomance but with less voltage, heat and fan noise. So achieived my goal.

I'm sure if I spent more time I could find a better sweet spot but I'm happy with it.

Ok done a fresh test.

Set all MSI Afterburner settings to stock and ran the Extreme Heaven (1440p windowed) and the Core Clock stayed around 1875, temps were around 80 deg, Wattage around 280. Score was 2796.

I then done the same as you, -200 then dragged 900mV up to 1965 but the second I hit Apply Heaven just crashed/closed.
I reverted and tried 1920, same issue.
I reverted and tried 1910, benchmark running at 1905, temps were around 70 deg, wattage around 200, but then it crashed around 40secs into the benchmark.
I reverted and tried 1900, benchmark running at 1905, temps were around 70 deg, wattage around 200, it passed the test with a score of 2727.
I reverted and tried 1890, benchmark running at 1890, temps were around 69 deg, wattage around 190, it passed the test with a score of 2726.

Now I found this strange as this is undervolted but a higher clock and the score was clearly a lot lower than the stock setting.

What would be the logical move from here?? Would I keep moving the clock speed down until I hit 1875 (stock avg) on 900mV (not sure if this will reduce my score even further)
OR
Jump to 925mV and start from 1965 etc working down?

 

[Cadder]

When I was undervolting my 3080, I picked the frequency I wanted (1905MHz) and just moved the voltage up in increments until Heaven was stable for a full benchmark run without any stuttering.

@Rowan88 you seem to be going the other way - picking a voltage and tweaking your frequency. I'm no expert mind, I just did what I read and it worked a treat. Coil whine reduced significantly from stock and performance came in better - all the while reducing power draw and temperature.

 

[Rowan88]

When I was undervolting my 3080, I picked the frequency I wanted (1920MHz) and just moved the voltage up in increments until Heaven was stable for a full benchmark run without any stuttering.

@Rowan88 you seem to be going the other way - picking a voltage and tweaking your frequency. I'm no expert mind, I just did what I read and it worked a treat. Coil whine reduced significantly from stock and performance came in better - all the while reducing power draw and temperature.

So you tried at 900mV and if it failed moved to 925mV etc until it was stable?

I am trying something similar just now, I done the exact tests above but moved onto 925mV, the 1965 crashed.
I reverted and tried 1920, benchmark running at 1920, temps were around 73, wattage around 215, it passed the test with a score of 2780.

Now this score is still lower than my original stock.

So does that mean the fact I am using less power (mV) that my performance is actually downgrading? I understand as I have it just now, the score is only 16 apart at the cost of 5 deg.


The sad thing is, the coil whine on the card itself is actually fine, you cannot hear it inside my case.
Its the fact that its amplifying through my active monitor speakers through the usb cable plugged into the rear IO to the Scarlett 2i2 interface.

Is there any other USB cables I can try to connect these that can stop the interference?
I seem something about Ferite Chokes but I wasn't sure if this would stop that getting to the interface/speakers.

 

[Cadder]

@Rowan88 have you seen this video? - https://forums.overclockers.co.uk/posts/35075312

First thing I'd do is make sure you're using Afterburner correctly.

I tried to mod my settings to the "sweet spot" mentioned in post 1 but it crashed my Heaven benchmark. 887mV was just too low for 1905MHz on my card, but it worked for others (like @lnoton for example). So I stepped it up in the increments allowed on Afterburner - 893mV / 900mV both caused stuttering and crashes in Heaven, but the next step (906mV) worked. So my settings are 906mV/1905MHz. Others can achieve this frequency with a lower voltage.

My Unigine Heaven 4 and Crytek Neon Noir benchmark scores are better than they were at stock.

Are you using the settings for Heaven in this thread? https://forums.overclockers.co.uk/threads/unigine-heaven-4-benchmark.18487976/

 

[Rowan88]

@Rowan88 have you seen this video? - https://forums.overclockers.co.uk/posts/35075312

First thing I'd do is make sure you're using Afterburner correctly.

I tried to mod my settings to the "sweet spot" mentioned in post 1 but it crashed my Heaven benchmark. 887mV was just too low for 1905MHz on my card, but it worked for others (like @lnoton for example). So I stepped it up in the increments allowed on Afterburner - 893mV / 900mV both caused stuttering and crashes in Heaven, but the next step (906mV) worked. So my settings are 906mV/1905MHz. Others can achieve this frequency with a lower voltage.

My Unigine Heaven 4 and Crytek Neon Noir benchmark scores are better than they were at stock.

Are you using the settings for Heaven in this thread? https://forums.overclockers.co.uk/threads/unigine-heaven-4-benchmark.18487976/

Ok I changed Heaven to the 1080p settings shown in that thread and my results at stock were:
Clock : 1875-1890
Max Temp : 79
AVG Watts : 280
Score : 3968

I then changed to 1890 @ 900mV and got this for the same Heaven settings:
Clock : 1905-1920
Max Temp : 73
AVG Watts : 205
Score: 4496

Now what is weird is that passed, as other times it has failed (given I was running at 1440p heaven but that's what I game at). So I ran Back4Blood into the Lobby section of the game and after about 30secs the game crashes out so that voltage is clearly not stable.

So what I am suggesting here is although I picked 1890 in Afterburner it jumped to 1905-1920, I assume this is the boost kicking in, so should I lower this to 1875 like the minimum speed on the stock card?

I assume the error I am making is when I run stock and get my speeds, I am picking the BOOST clock score and putting it into MSI Afterburner so essentially I am overclocking the card as well.

For example I just picked 1875 in Afterburner and re ran the test and I got speeds of 1890-1905.
So for me to replicate my stock speeds I would need to lower than until I get speeds of 1875-1890?

UPDATE
I think that may be the case, I changed the speed on Afterburner to 1860 @ 900.
It gave me these stats:
Clock : 1875-1890
Max Temp : 72
AVG Watt : 205
Score : 4454

So comparing that to the Stock results essentially I am running at the same speed etc, but I have reduced the temps/wattage and somehow gained an extra 450 pts in Heaven doing so.

Does this sound correct?

The only thing confusing me if this is true is in the game (Back4Blood) on stock it runs 1875-1890, but with these settings I only get 1860, so why is it not jumping up like the benchmark did?

 

[Cadder]

I don't know the ins and outs of it all @Rowan88 but it seems like it's normal for the frequency to lock at some sort of increment - I think I read on these forums that it moves in increments of 15MHz but someone more knowledgeable should advise. As long as Afterburner is keeping your card voltage right and the frequency in the region I'd hope it would be fine.

It looks like you and I have tweaked our settings differently. I picked a frequency and adjusted voltage until stable. You picked a voltage and adjusted frequency until stable. As long as it works, is stable and you're happy that's what matters.