Slender wrote: ↑18 Apr 2025, 08:28
disable ground on your panel and retest it.
Doesn't make a difference, which makes sense. If what I've found out here is really the root cause of my issues, which I am now more and more sure of it is, it's a problem with the hardware itself and how PCB's are designed. There's nothing you could ever do to fully isolate the GPU so it's only connected to ground at one point, as it needs the 2x PCIe power plugs and the GND pins in the PCIe slot at the very least, even if you could isolate it from the case.
Safety ground coming from the socket has almost no impact here. Yes, I've observed another path back to the PSU's ground via the DisplayPort cable and the monitor's 3-prong cable, but the current flowing there was very low in comparison. Flipping the switch on my apparatus wouldn't change this, because the ground loop doesn't come from "outside" of the system. My computer and the monitor are still connected to the same power strip, only that I can turn "off" the safety ground going into it. The loop is still there, the switch doesn't influence this.
I could theoretically cut the ground bars within the power strip to fully isolate every device from safety ground AND from each other, but I can already do that by using a monitor with an external power brick that has no connection to ground. This stops current from flowing on the safety ground wire alltogether. Same behaviour: Workaround wires connected means perfect gameplay, wires disconnected means problems come right back with any PSU than the newest unit.
Again, what I think the problem is that there's multiple paths back to the PSU's ground
within the system. The problem is contained within this closed system and external factors, like a conncted safety ground wire or not, doesn't make a difference.
Think of it as multiple ground loops within the computer. Current going into the PSU is equal to current returning from it. Everything up to the PSU's power plug is totally normal. If you were to measure the current on both live and neutral, they would be equal all the way to the PSU (it is, I've measured it). The ground loops occur after/behind the PSU.
ZINZIRIO wrote: ↑18 Apr 2025, 11:23
It didn't make any difference to me... I'm going to try a gas generator these days since what made a difference for 2 weeks (I thought it was fixed) was connecting my neutral to ground at the main panel, which gives me an indicator that my problem is probably coming from my electric provider.
Damn, I'm sorry to hear that. Did you make sure to use a thick wire that can carry a lot of current? I've tried it with a tiny wire and two aligator clips at both ends, and that didn't make a difference, as the clip connections are bad and the flimsy wire can't carry enough current to make a differnce. Maybe try two wires like I did?
What electrical system are you on? If you're on TN-C-S, this connection between neutral and safety ground should already be in place. If you're on TT, this connection should not be made ever, as your house should have a ground rod somewhere outside. There should be no connection between safety ground or neutral at all in a TT system. If you're on TT, you also require a RCD.
You're probably not on TN-C because if you were, there shouldn't be a grounding wire inside the main panel, as neutral and ground is separated only at the outlet. However, if you're somehow on TN-C, strongly consider investing into getting that fixed, as it is unsafe and most likely not up to electrical code, depending on where you're from.
dervu wrote: ↑18 Apr 2025, 11:34
It may be that this tweak helps some users who are close to the tipping point, while for others the offset is already well beyond that.
Possible, yes. The real problem with this is the seemingly endless amount of possibilities and variables in play. If designed with maximum perfection and only this specific problem in mind, a motherboard can probably fully account for it and "eliminate" the symptoms, even though the "problem" is still there. The GPU manufacturer could probably also make design improvements to stop this exact thing from happening. There are SO MANY of these variables in play, one singular resistor on a graphics card might be enough to make all the difference.
I'd argue it's likely every single computer has this problem to some extent. What I think is going on is that this isn't deemed a problem by hardware manufacturers. The computer works, turns on and reaches benchmark numbers after all. The little interference this creates is easily handled by error correction mechanisms without impacting the computer's performance at all.
Only specific combinations of hardware will show
any problems at all. And even then, most of the time these aren't even that bad, maybe a rare USB device dropout or something minor. I'd argue the circumstances need to be pretty bad for "weirdness" to occur, like mouse input problems or desync.
We've seen these problems with internal ground loops come to light only after the 12VHPWR connectors started melting and people were investigating why only the 12v pins were damaged. The 12VHPWR connector is typically only found in high power graphics cards that require a lot of power. However, as the article from igor'sLAB says, it also happens on AMD cards with regular PCIe power connectors. I'd say the GPU model and it's power consumption plays a large part: Low-end or midrange GPU's don't even draw enough power to cause enough interference, even when the PCB design isn't ideal.
So much to consider with so little information about this available. Maybe this can influence even more stuff that we don't know about? We can hope that this will now be of more relevance, as this has to be addressed somehow if NVIDIA and AMD want to stop their GPU's from melting the 12v pins in the future.
dervu wrote: ↑18 Apr 2025, 11:34
I’m not sure why, in the latter case, they wouldn’t see more obvious disruptions—such as USB dropouts, etc.
I think it's because the problem is so elusive. It's the same thing for me: When informing myself about the "problem" for the first time, I noticed almost everybody who has the weird desync and hitreg problems also has some form of mouse/input lag. However, I don't have mouse problems at all and can't recall ever having them. It's likely because the motherboards I've been using keep the USB "far enough away" from the interference, while the network card is affected at the same time. This can obviously be flipped, where people have mouse problems but no desync/hitreg problems, it can be both at the same time and also nothing at all.
Also, I think we see effects of this in the wild. The article from igor'sLAB already shows some effects of this in the form of audio disruptions, I've seen threads on other forums about USB devices malfunctioning which was seemingly fixed by using another USB port, seen people fixing internet speeds with a PCIe network card instead of using the onboard adapter...
