Buzz when GPU is loaded

[mhardy6647]

I would suggest that too but I have seen far too many of them to have resistances up to 20 and 30 Ohm (one had more and one had none )
(talking about new cheap ones from questionable sources)

Thank the gods for Mouser of course with their quality stuff,I never shop elsewhere for more than 10 years now.

I use only audiophile clip leads!
Yes, that's a good point. Not a problem I have had, but (seriously) most of mine are "vintage" and well constructed.

 

[JayGilb]

A ground loop can lead to a non-zero potential on the ground connection, this alters the reference point used by the GPU circuits and other devices in the PC. Noise is then introduced by signals switching and crossing the new non-zero reference and can be amplified by audio circuits.

 

[MKreroo]

What is the 5K? A DAC?

Are the speakers powered? So unbalanced interconnect from the 5K to the speakers?

Are both the 5K and the speakers grounded (3 pin power)? Are the 5K and PC connected to the same outlet?

Sorry for the confusion, 5K as in Qudelix 5K

Leading one to wonder whether @MKreroo's mains ground is in fact grounded.
I am a trust but verify kind of hominid.
Unless one is absolutely certain that an outlet is wired properly -- it is prudent to check.
It's easy enough to check with a line tester.
Here, for reference, is a US version thereof. A handy thing to have around.

View attachment 351001
This one even has a GFCI test, which I guess is good.
Mine doesn't -- but then again mine also cost a bit less than this one.

hmmm good point, the PC is running through a power bar perhaps the bar is really not grounded, I will check if they are properly grounded.

This is a long shot but since it fixed some other user's problem:

I suppose it's a custom built machine,right?
Did you make sure your PSU and the case have the lowest possible resistance?
New,nice looking parts are often not making a good physical contact,one must scratch a small ring around the screw holes sometimes.

You can test that without scratching it with a thick copper cable that will make a good contact (find bare metal) between PSU and PC case.

A long shot,but...

it is a custom built, but also 3 years old at this point, so I suppose the wear on the components should be enough? (re-assemblt it a few times through out the year).

Wow I'm really curious about this... You get this on direct connection and on bluetooth? Wild. I'm not quite following the acronyms (5k?) so please bear with me if I ask something redundant. Are you plugging the PC and audio equipment into in the same power source? Power strip, wall plugs, UPS? When you run Bluetooth is it still also plugged into a physical connection or did you disconnect that? How close is the audio equipment to the PC?

The PC and Speaker are on different power bar, but the bars are on the same wall outlet. PC is on the floor and Speakers are on the desk.

You can be right but how does this correlate with the noise being audible only when the GPU is pulling power?

Come to think of it @MKreroo to troubleshoot whether the noise is coming from your PC PSU or from the GPU you can try to stress test your CPU. Maybe it doesn't draw enough power to really reveal the issue but it can be worth trying. Use this online CPU stress test while playing music-->

CPU Stress Test - An online tool to test your CPU stability | CPU Expert

CPU Stress Test is a CPU stability test that allows you to test online your processor at heavy load

cpux.net

Yeah I stressed the CPU and no noise, if I simulate similar powerdraw on GPU (~120W) the noise is also not present. Only when the GPU draws above 200W (so I guess total system ~300w) the noise appear.

 

[markalot63]

I built a little ITX machine with a passively cooled video card (GTX 1030) for my streamer. Even better would have been an Intel or AMD with onboard video. My main machine is homebuilt as well and doesn't have much of an interference issue but it is there and I can here it when playing quiet music while stressing the card. Not loud enough to hear in games (RTX 4070ti). The only reason I noticed it was by listening to Spotify while running GPU and CPU stress testing. If I had another power supply sitting around and if it was convenient to swap it in I might test it.

I typically lock my framerate at 60FPS so during games my GPU is rarely stressed, unless playing a horribly optimized game. Not sure I helped here, maybe some new ideas?

 

[MKreroo]

Here is a recording of the noise in question (Isolator with GPU stressed.wav) compared to full ground loop noise (No Isolator.wav) that I was having some time ago.
Mic was just my phone mic but placed right before the tweeter.
Isolator used

The noise without isolator is much much louder, whereas the noise in question (Isolator with GPU stressed.wav) is barely recorded and mostly masked by PC fans, still audible in person but very quiet.

The fans are fixed at low rpm for this recording.

The Speakers is iLoud MTM, they do hiss when idle but not that static noise recorded.

 

[neRok]

You can be right but how does this correlate with the noise being audible only when the GPU is pulling power?

My DIY sub had this issue when connected to PC. At idle you couldn't hear the noise, but you could feel the driver moving if you touched it (it's 50/60Hz noise depending upon power grid). If I played a game, the noise would be very audible. I read that many GPU's have a lot of leakage out the earth circuit (by design, a lazy/poor design) when really they shouldn't.

 

[Rhodo]

Sorry for the confusion, 5K as in Qudelix 5K

So I looked that up, am I correct that it gets its power through the PC USB connection? When you try bluetooth and still get the buzz, is the 5k still getting power from the PC? I'd try connecting the 5k's USB to some other power source other than the PC and seeing if that silences the buzz while playing over bluetooth. If that solves it I'd very strongly recommend replacing the 5k with a DAC/Amp that has it's own separate power plug to the wall.

If you've already tried that (getting power to the 5k from the wall) and still got buzz then this is a challenge.... I'm a big proponent of uninterruptible power supplies, battery UPSs, for pretty much all PC equipment, howver they can be pricey particularly for the good sine wave replicating types. One of those would be a nice upgrade over a standard surge power strip though and would have the double benefit of eliminating any issues with wall power and also just extending the life of all your electronics plugged into it. However, I'd test the wall circuit as mhardy6647 recommended before buying any power equipment, it could definitely be a circuit ground issue and a plug tester is vastly cheaper than a UPS.

One other thing that comes to mind, take a look at the wiring in the PC and make sure everything is plugged in securely and no bare wire or plug ends are exposed and touching the case/motherboard/anything else they shouldn't be. Could also try keeping the power cable for the GPU physically separate from any other cables as much as is reasonably possible. It's sort of a shot in the dark but it could possibly help if there's strong current creating some kind of interference.

Edit: Oh by the way, I listened to those audio files and that's a really awful noise!

 

[antcollinet]

Sorry for the confusion, 5K as in Qudelix 5K

I'm still interested in the grounding of the 5K and the speakers.

First as pointed out by Rhodo - if you are still powering from the PC when using bluetooth, you will still have a ground loop that will couple through the analogue interconnect between 5K and speakers. In this case bluetooth or USB audio will make no difference.

Even if not - if the 5K and speakers are both grounded then it is possible for noise from the PC to still take this path, even if not connected to the PC USB, like this:

Some of the noise from the PC/graphics card be filtered** to ground. From there there are two paths to get to your building ground point.

1 - back from the PC Mains socket through the house wiring.
2 - From the PC mains socket - to the 5K mains connection, than round through the speaker ground connection - back to the mains wiring and then to the building ground point.

Noise currents will FLOW THROUGH BOTH PATHS. The amount flowing through each will depend on the relative impedances of the two paths.


**EDIT - or otherwise coupled into the ground circuit - eg through magnetic coupling.

 

[MKreroo]

I'm still interested in the grounding of the 5K and the speakers.

First as pointed out by Rhodo - if you are still powering from the PC when using bluetooth, you will still have a ground loop that will couple through the analogue interconnect between 5K and speakers. In this case bluetooth or USB audio will make no difference.

Even if not - if the 5K and speakers are both grounded then it is possible for noise from the PC to still take this path, even if not connected to the PC USB, like this:

Some of the noise from the PC/graphics card be filtered** to ground. From there there are two paths to get to your building ground point.

1 - back from the PC Mains socket through the house wiring.
2 - From the PC mains socket - to the 5K mains connection, than round through the speaker ground connection - back to the mains wiring and then to the building ground point.

Noise currents will FLOW THROUGH BOTH PATHS. The amount flowing through each will depend on the relative impedances of the two paths.


**EDIT - or otherwise coupled into the ground circuit - eg through magnetic coupling.

For the Bluetooth connection the 5K was disconnected from the PC entirely, running on it's own battery instead.
I tested the socket and bar with the plug tester @mhardy6647 mentioned earlier and all seems to be correctly wired, according to the tester.

I guess perhaps it is the noise from GPU passing into the speaker?

 

[Rhodo]

What a weird situation... Definitely a PC problem more than an audio one. Tried a little googling and ran into something interesting: If you're comfortable checking settings in the BIOS see if PCI-E Spread Spectrum is turned on or not. More info here (adblocker recommended): https://www.techarp.com/bios-guide/pcie-spread-spectrum/

Unfortunately since motherboard manufacturers have very different BIOS options I can't really tell you where to go and look for that in your PC's settings, however my feeling is that's worth checking for and enabling if it's currently turned off.

Still recommend a battery UPS for the PC, if there's something getting sent back to the wall that should prevent it. Otherwise, if it's some kind of bizarre EM radiation from the GPU then I guess you could build a Faraday cage around the PC. Added benefit to that, the space aliens won't be able to read your hard drive!

 

[MKreroo]

So out of nowhere, I decided to re-seat the right channel speaker's power cable, and took this chance to re-route it to be a little further away from PC main power. The left channel is untouched as I noticed the noise on the right channel is ever so slightly louder.

The right channel power was bundled up with other cables (USB, display, power etc) with cable ties to make them "slightly organized" (still a mess haha)
After re-doing the organizing, somehow the noise is gone from both channel.

Not sure what the cause is (def not lifting cable lol), but I don't think it's power cable seating.
Perhaps it's EM noise? from the PC main power affecting the right channel speaker, which through the power bar, seep into the left channel?
Seems unlikely because their overlapping is very short distance and both cables are fairly thick (=adequate shielding?)

 

[antcollinet]

The right channel power was bundled up with other cables (USB, display, power etc)

That could well do it.

I'm not letting the ground loop theory go. You have a DAC (even running on battery) connected with two interconnects - one to each speaker. Each speaker connected to earth via main lead.

So your loop is :

Mains earth to right speaker to dac to left speaker to mains earth.

You've then cable tied your right speaker power lead to a bunch of PC cables allowing them to magnetically couple into the loop (the loop acts like a single turn transformer secondary winding that is shorted - any magnetic field going through the area inside the loop will induce current. The coupled noise currents in the loop cause voltage differences between DAC and speakers which are added to the signal at the input to the speaker.

 

[beefkabob]

I'm having a similar issue. I have a PC with an AMD RX 6080 GPU known for coil whine. HDMI out to a receiver. Receiver goes from unbalanced to balanced to powered monitors.

Even if I move the mouse, I can hear buzzes, but the more I use the GPU, the louder it gets. When I run something on an LLM, oh boy does it get loud.

I put audio on the PC to 100% and turned down the receiver. That helped some.

I undervolted the card. That helped significantly. Between these two changes, the whine is tolerable as long as I'm not heavily using the GPU.

I switched from HDMI out to displayport out, which did nothing.

I'll look for spread spectrum settings.

All outlets are grounded, but the 5.1 speakers, etc are on at least a couple different circuits. The PC is on a UPS.

Before, everything ran through a miniDSP SHD, so it was all digital from USB. I noticed coil whine, but not at this level, and mainly from the card itself.

I will see if I can move the power cards around and seal up the case. I've had it open ever since I got the receiver.

I can test the plugs, once my AliExpress order arrives.

I had been just running the PC to the TV, and then there was audible but far less noticeable coil whine that I had assumed was straight from the card.

In 25+ years of GPU ownership, I've never had this issue, even with my good old Voodoo2s in SLI.

I wish there was a cost effective way to go 100% AES-EBU with 5.1 sound and Dirac.

 

[xNORgate]

In 25+ years of GPU ownership, I've never had this issue, even with my good old Voodoo2s in SLI.

Radeon RX6800' power needs per their website:

You've probably tried both HDMI ARC and/or outputting the audio from the PC's USB, to determine if there is a difference.
Most Radeon FX need multiple power feeds from the PS.
How old is your MoBo? Do you have 'Gold-rated' PS?

 

[beefkabob]

Radeon RX6800' power needs per their website:


You've probably tried both HDMI ARC and/or outputting the audio from the PC's USB, to determine if there is a difference.
Most Radeon FX need multiple power feeds from the PS.
How old is your MoBo? Do you have 'Gold-rated' PS?

A 600w bronze, and I don't see why that matters. The system probably maxes out at around 450w from the wall, and the coil whine is in the GPU.

 

[wwenze]

Coil whine can depend on the quality of power upstream i.e. your PSU. Back in the Athlon days I had a motherboard that was whining loud with a brandless SFX PSU but completely silent when brought to the distributor.

Opening up the case to decrease temperature can reduce coil whine in some cases.

 

[AnalogSteph]

I'm having a similar issue. I have a PC with an AMD RX 6080 GPU known for coil whine. HDMI out to a receiver. Receiver goes from unbalanced to balanced to powered monitors.

Even if I move the mouse, I can hear buzzes, but the more I use the GPU, the louder it gets. When I run something on an LLM, oh boy does it get loud.

I put audio on the PC to 100% and turned down the receiver. That helped some.

100% a ground loop. Please specify the exact topology and construction of your "unbalanced to balanced" wiring. Odds are you need the kind that only uses pins 2 and 3 and leaves 1 open. (Otherwise you'll be sending ground currents through half the receiver.) Or whack a Behringer HD400 in there.

Since GPU current contributes to the disagreement in chassis ground potential between PC and monitors that is being made audible, varying GPU power consumption is obviously going to affect how loud the whine is.

Before, everything ran through a miniDSP SHD, so it was all digital from USB. I noticed coil whine, but not at this level, and mainly from the card itself.

Balanced connections work - who knew?

 

[beefkabob]

I do not know the exact wiring of my adapters. I got the cheapest I could find. I'm running short adapters to XLR cables. The issue existed when I ran RCA to other adapters I had laying around. Is there a mod I can do to an existing adapter or XLR cable? Break a pin off the XLR?

 

[beefkabob]

My adapters are wired RCA center to, when looking at the back of the XLR, upper left, and shield to upper right. I think it needs to be shield to 3 or to upper right and 3. Gonna solder.

 

[AnalogSteph]

XLR pins should usually be labeled with numbers, at least on the female side. Either way, this is the back (solder side) view:

How to Build Your Own XLR Cables: A Step by Step Guide - Studio DIY — The Home Studio Archive

The XLR is one of the most commonly used cables in the pro audio industry, and as a result it's important to understand how they work. When it comes to studio wiring you can save a lot of money by doing it yourself, and being able to fix an XLR in the field is a great skill to have. Check out this h

www.homestudioarchive.com

The average RCA to XLR adapter wires center conductor to 2 and shield to 1+3. You want no connections to 1 in this case. It actually sounds like things may be wired up to 2 and 1 now, which would be the worst possible option.