Unstable Pc performance and lack of visual smoothness.

[Manul]

My current working hypothesis is not “random jitter inside the PC,” but a shift in reference conditions: chassis, ground/neutral potential drift, common-mode noise, and slow changes in supply or environmental electrical conditions that alter how digital circuits operate without causing outright errors or crashes.

I can disprove it right away. Well, not me, you yourself disproved it. All these things (some of them does not matter anyway) are disproved by running a laptop from battery or PC on UPS.

In other words, I am trying to find parameters that explain why the same digital system behaves differently without any internal state change.

If deterministic system behaves differently, by definition there must be internal change. It might be caused by external influence.

I don't know what to say. Go to backyard and if ground is not frozen yet, dig 3 meter deep hole, go inside with a laptop, cover the hole with a sheet of metal for extra shielding and test. Not a sarcasm btw.

[temperance]

Replacing the distribution transformer temporarily improved the situation, but after the first emergency power outage the effect disappeared completely.

Get yourself a variac and try to adjust the supply voltage to the computer.

[Simmed]

a) The reason I am hesitant to go down this path is not because I reject measurements, but because I already spent a long time doing functional end-to-end testing of the whole system under different conditions, and the results point away from an internal jitter problem.

b) My current working hypothesis is not “random jitter inside the PC,” but a shift in reference conditions: chassis, ground/neutral potential drift, common-mode noise, and slow changes in supply or environmental electrical conditions that alter how digital circuits operate without causing outright errors or crashes.

c) I am trying to find parameters that explain why the same digital system behaves differently without any internal state change.

 
but w/o measurements
great

[Inputdelay]

Replacing the distribution transformer temporarily improved the situation, but after the first emergency power outage the effect disappeared completely.

Get yourself a variac and try to adjust the supply voltage to the computer.

a) The reason I am hesitant to go down this path is not because I reject measurements, but because I already spent a long time doing functional end-to-end testing of the whole system under different conditions, and the results point away from an internal jitter problem.

b) My current working hypothesis is not “random jitter inside the PC,” but a shift in reference conditions: chassis, ground/neutral potential drift, common-mode noise, and slow changes in supply or environmental electrical conditions that alter how digital circuits operate without causing outright errors or crashes.

c) I am trying to find parameters that explain why the same digital system behaves differently without any internal state change.

 
but w/o measurements
great

Let me clarify an important part, which I apparently failed to state clearly before.

I do have measurements, but they are not internal PC telemetry, because internal telemetry shows no meaningful difference between a “good” location and my location.

What does change significantly is the electrical reference environment.

Here is what I can measure reliably with the equipment I have (multimeter, isolating transformer, autotransformer):

1) Neutral–ground behavior:

- Voltage between neutral and ground fluctuates heavily: from ~0.1 V up to 5–10 V AC.

- This happens when loads are switched on/off either in my house or by neighbors.

- Neutral-to-ground resistance is unstable and varies dramatically; during 1 second it can change multiple times (measured values range roughly from ~4 ohms to >100 ohms depending on time/load).

2) Phase reference instability:

- Neutral behaves worse than ground.

- Phase-to-ground voltage differs by ~10 V from phase-to-neutral.

- In a properly functioning system, neutral should not float like this and typically should stay within ~1–2 V AC relative to ground.

3) Abnormal behavior during outages:

- When the grid is down, I still measure DC components:

  ~0.3 A DC current and ~0.4–0.5 V DC present on both phase and neutral,

  while AC voltage is ~0.01 V.

- This strongly suggests coupling through the distribution wiring itself, which is acting as an antenna/reference path.

4) Grounding effect:

- I have a TT system with grounding only on one outlet (PC + monitor).

- Adding grounding reduced symptoms by ~8–12%, but did not solve the issue.

- Day/night variation completely disappeared after grounding (April 2024), which directly implicates reference stability rather than software.

5) Power quality experiments already done:

- Isolation transformer, variac, EMI filters up to 18 GHz, double-conversion UPS, portable power stations-no stable improvement.

- Variac changes cause transient degradation only; after ~4–5 seconds the system converges back to its previous state.

- This looks like internal regulation settling, not supply RMS sensitivity.

6) Internal PC metrics:

- FPS, frametimes, temperatures, clocks, voltages, and fan speeds remain identical in a location where everything feels perfect.

- Therefore, internal measurements do not correlate with the problem.

- The only variable that changes is the external electrical environment.

Because of this, I am not dismissing measurements-I am trying to measure the correct domain.

What I am explicitly looking for is guidance on:

- Which external electrical parameters (neutral drift, ground potential shift, slow common-mode variation, DC offsets, and low-frequency noise) are physically plausible to affect digital timing without causing crashes?

- and which instruments are appropriate to capture those parameters over time.

I am not trying to prove that the problem exists.

I am trying to understand what environmental variable could explain why the same digital system behaves differently with no internal state change.

[negativ3]

If concerned about RF, you need to try it in a DIY faraday cage. Depending on your design, YMMV.

[Manul]

This is going nowhere, I'm tellin' ya.

[Inputdelay]

The question is, how willing are you to put effort and set up lab tests?

Two ideas:

1. Confirm and possibly log your mouse update rate. What mouse rate do you have according to pages like this?

https://cpstest.org/polling-rate-test/

Setup something which rotates continuously in front of optical mouse sensor to simulate a constant movement (alternative: program a microcontroller to act as HID mouse simulating continuous movement). Write some piece of software, which logs and analyzes: position update rate (is it stable) and position delta per update (are there no abnormal position jumps).

2. Make an optical sensor from a photodiode with amplifier. You can plug it into sound card, cause frequencies of interest are within audio range. Make software which renders a quickly and steadily changing scene like a sliding smooth gradient (use Unity or something). Do spectrum analysis on the captured signal. Factual display update rate should be visible as a clear spectrum line (let's say 360Hz). You can write software to compute FFT in real time and log the factual display update rate. You may be able to expand on this idea and analyze ghosting too.

Otherwise I don't know what else to suggest.

No matter how many tests via Windows I do, it will change nothing, as I said. But since you are saying it's going nowhere, here are some of them.

Here's a demonstration of the mouse polling rate test, and as you can see, it's pretty stable. I was just doing circles with my mouse to perform this test in the program Mouse Tester. Here are two examples: the so-called 'ideal' and the second one at my place. It's the same PC and mouse. Don't pay attention to the CPI; I put it in manually as a marker to make sure they're 2 different screenshots in order not to accidentally confuse them.
But if you check how the wireless looks, it's not that great actually. Even though I have turned off Wi-Fi, it didn't change at all.

[Inputdelay]

If concerned about RF, you need to try it in a DIY faraday cage. Depending on your design, YMMV.

I have already tried a DIY Faraday cage, using aluminum foil, but there was no effect. But my friend did it properly with 0.5 mm copper mesh and completely covered the entire computer case, leaving a hole only for the power wires, display port, and mouse and keyboard, and of course grounded the whole thing, and for some reason it became much worse.

In my case there is a slight change; if I move the PC case by putting it on a table, everything is much better, while on the floor everything starts to feel weird and definitely worse at some point.

[NE666]

It's not such a trivial thing, constructing an effective Faraday cage.

And that notwithstanding, IF there is a strong external EMI influence then "leaving a hole only for the power wires, display port, and mouse and keyboard" would be a good way of creating an antenna with which to transfer it to the equipment within the cage.

[Inputdelay]

This is going nowhere, I'm tellin' ya.

Also, I have an electromagnetic field detector, VT-ER1. I have done those tests many, many times. But the most noticeable difference is that without the grounding there is much more RFI. And as I said before, the ground itself gave a constant, like, 8-12% better feeling.

Meanwhile, it's fluctuating and constantly changing around the PC, from 1 uW/cm² to 10 uW/cm² in front of the case and from 15 to 60 uW/cm² on top of the case, closer to the CPU, and this is happening no matter the grounding when it comes to RFI noise. But interestingly enough, the magnetic field is pretty stable and fluctuating much less; I would say almost constant. And the electric field is changing as well, but it should, as far as I understand.

And when I was trying to find anything around the mouse, keyboard, or monitor while they were working, there was almost no RFI (1-2 uW/cm²), no electric field, and no magnetic field.

[Inputdelay]

It's not such a trivial thing, constructing an effective Faraday cage.

And that notwithstanding, IF there is a strong external EMI influence then "leaving a hole only for the power wires, display port, and mouse and keyboard" would be a good way of creating an antenna with which to transfer it to the equipment within the cage.

Then the PC won't even work, since there would be no power cable, and we wouldn't see the change to see the monitor, since there would be no display port, and we wouldn't be able to use the PC via mouse and keyboard to test it.

[NE666]

As I led my response with, it isn't a trivial test to conduct. However, it is possible, using a battery-based supply and a suitable automated test harness.

My point is, what you and your friend did (with foil and copper sheets), doesn't represent an effective EMI screen / cage and therefore doesn't eliminate the possibilities that a REAL test would, or that you think it has.

[negativ3]

Have you tried cleaning your ram edge connectors with an eraser?

[NE666]

Have you tried cleaning your ram edge connectors with an eraser?

Everything the OP has already told us in multiple posts says that this affects more than one system. If we take this as fact, then we must consider hardware and software issues to have been eliminated.

"The problem is not tied to:
a specific PC (tried a lot of different ones, including notebooks and got literally the same result)"

"If I take my PC and my entire peripheral and bring it to another place in which there is no such problem, everything would be just fine."

"the problem disappears when I move the same PC and monitor to another location"

"I asked a couple of my friends to bring their PCs and notebooks, and they literally had the same problem in my place."

[negativ3]

Ahh, okay so at root it's not a PC problem, it's an environment problem manifest in PC performance.
I'd try running the PC off a battery through an inverter to rule out local power.

[NE666]

Again, if we take the OP's statements as fact, then they have already eliminated that possibility (even if they themselves don't acknowledge it). When their city has lost power, they've been running the PC on a battery-based power bank, and the effect is still observed.

3.3 Power supply
Operating from the grid with all house circuit breakers turned off - no change.
Long-term power outages in the settlement - no change.
A complete city-wide blackout - no change.

[NE666]

Based on what we've been told, I think there are just two (generalised) possibilities remaining;

1. The effect is with the user: Subjective phenomena, cognitive bias, shared delusion, possibly real visual disturbances due to pathology and/or drug effects

2. The effect is environmental but manifest in a specific location. Not related to the utility power system, and as yet unidentified


I'd be inclined to dismiss the second option, were it not for the OP's country of residence. Personally, I wouldn't discount the possibility, no matter how unlikely, that there is some form of interference as a result of electronic warfare operations in their proximity. Surveillance / counter-surveillance, battlefield comms, phased radar, signal jamming etc. etc.

[Manul]

I don't game myself, so I'm not very interested in the details of modern games, but main principles still apply (around 20 years ago I wrote a rudimentary game engine for fun):

Game engine sends data to GPU so it can render frames (time domain snapshots of the scene). Movement and other changes in the scene are calculated by the engine using:

* A timing source (on windows probably QueryPerformanceCounter or historically reading the TSC register directly), so it knows exactly how much time passed after the last frame and can advance animations accordingly and control overall pace.

* Data received from local input devices (mouse, keyboard, etc);
* Data received over network (multiplayer);
* Internally generated movement data (for instance, NPC walking);

Display's job is to display these snapshot images as they are generated. Then your brain's job is to perceive the movement by blending these snapshots into fluid motion.

These days there is also frame interpolation done by GPU to efficiently (and somewhat artificially) increase the frame rates. Cause these interpolated frames are not real frames, they are just "guesses". The downside is that it often causes visual artifacts and also adds latency.

There is of course shaders and various post processing too.

Anyway, take a look at the picture below. It is a mouse cursor on my laptop display in front of the grid (photo with long exposition time, this is not display ghosting). Dim lines every 10px, brighter lines every 100px. Mouse is not moved by me, but programmed to move exactly 5px / 1ms. Monitor updates at only 120Hz, so it does not display all mouse updates. There are some minor non-linearities, cause of 1000Hz position update rate and 120Hz refresh rate "beating". And that's because the cursor is moved a fixed 5 pixels per time unit. To achieve better linearity in motion, application must predict the next refresh moment and adjust the position accordingly in order to maintain linearity, thus some form of syncing.

Anyway, I look at the photo of my cursor and I know immediately, it is PERFECT. There is nothing else here to ask. That is exactly what was supposed to be generated and displayed to my eyes. No magic, no woo woo, no neutral to earth balance, potential drift or whatever.. These are accurate snapshots of cursor moving at fixed speed. The rest is my brain, which perceives it. There is obviously ghosting, that is visible if short exposition time is used. Ghosting depends on object and background intensities, contrast. Depends on temperature too. Usually low temperatures significantly degrade panel response time and increase ghosting, can affect color calibration and such. In any case it is intrinsic to my display panel, so it's perfect by my definition.

In games too, if frames displayed are perfect, then frames are perfect. Display panel is also perfect, if it is in spec and not defective. Quality varies: you want less ghosting, better colors, you buy better display. Simple.

And if something is clearly wrong and not smooth, then it MUST BE MEASURABLE. And when objective measurements are done, it helps troubleshoot tremendously. Reading something like "feels 8% better" is laughable frankly. And then again and again some weird hypotesis continues. Please don't get ofended, this is just a personal opinion, it is ridiculous that you tested 100 windows builds, invited radio-monitoring inspection, built improvised faraday cage, tested industrial RFI filter up to 18Ghz, measured electric, magnetic fields and there is so much more on your list, I cant remember. This is nuts. And all based on feeling? Bro, this is not how it's done.

[Inputdelay]

Based on what we've been told, I think there are just two (generalised) possibilities remaining;

1. The effect is with the user: Subjective phenomena, cognitive bias, shared delusion, possibly real visual disturbances due to pathology and/or drug effects

2. The effect is environmental but manifest in a specific location. Not related to the utility power system, and as yet unidentified


I'd be inclined to dismiss the second option, were it not for the OP's country of residence. Personally, I wouldn't discount the possibility, no matter how unlikely, that there is some form of interference as a result of electronic warfare operations in their proximity. Surveillance / counter-surveillance, battlefield comms, phased radar, signal jamming etc. etc.

I think the framing into only these two options is incomplete.

There is a third category, which is neither psychology nor exotic interference like EW or jamming.

Degraded reference conditions in the local electrical environment that affect real, physical timing margins without causing functional failure.

To be very clear:

I am not suggesting radar beams, jamming, surveillance systems, or anything remotely military-grade.

If such fields were present at levels capable of affecting modern digital electronics, we would see widespread failures, crashes, and regulatory alarms. That is not happening.

What I am describing is far more mundane and far more common in aging or poorly maintained power distribution systems:

- unstable neutral,

- shifting ground reference,

- asymmetric loading,

- DC offsets and slow potential drift,

- wiring acting as unintended coupling paths.

These effects do not “break” digital systems.

They modulate clock quality, synchronization stability, and analog margins inside otherwise digital pipelines.

This is not theoretical.

This also explains why:

- Internal telemetry (FPS, clocks, temps) looks identical in “good” and “bad” locations.

- the system never crashes,

- the effect is subtle, intermittent, and reversible,

- Replacing or adjusting upstream power infrastructure temporarily changes behavior.

- Adding a local ground reference partially improves the situation.

Regarding option (1):

The hypothesis of shared delusion or pathology does not fit the observations.

Multiple independent observers (including engineers) notice the same visual artifacts on the same equipment in the same location, and those artifacts disappear when the equipment is operated elsewhere.

That is not how cognitive bias behaves.

So the question is not “psychology vs electronic warfare.”

The real question is

which external, low-level, non-catastrophic electrical parameters can alter timing and synchronization quality in a modern PC without triggering faults-and how to measure them properly.

That is the direction I am trying to pursue.

[Inputdelay]

I don't game myself, so I'm not very interested in the details of modern games, but main principles still apply (around 20 years ago I wrote a rudimentary game engine for fun):

Game engine sends data to GPU so it can render frames (time domain snapshots of the scene). Movement and other changes in the scene are calculated by the engine using:

* A timing source (on windows probably QueryPerformanceCounter or historically reading the TSC register directly), so it knows exactly how much time passed after the last frame and can advance animations accordingly and control overall pace.

* Data received from local input devices (mouse, keyboard, etc);
* Data received over network (multiplayer);
* Internally generated movement data (for instance, NPC walking);

Display's job is to display these snapshot images as they are generated. Then your brain's job is to perceive the movement by blending these snapshots into fluid motion.

These days there is also frame interpolation done by GPU to efficiently (and somewhat artificially) increase the frame rates. Cause these interpolated frames are not real frames, they are just "guesses". The downside is that it often causes visual artifacts and also adds latency.

There is of course shaders and various post processing too.

Anyway, take a look at the picture below. It is a mouse cursor on my laptop display in front of the grid (photo with long exposition time, this is not display ghosting). Dim lines every 10px, brighter lines every 100px. Mouse is not moved by me, but programmed to move exactly 5px / 1ms. Monitor updates at only 120Hz, so it does not display all mouse updates. There are some minor non-linearities, cause of 1000Hz position update rate and 120Hz refresh rate "beating". And that's because the cursor is moved a fixed 5 pixels per time unit. To achieve better linearity in motion, application must predict the next refresh moment and adjust the position accordingly in order to maintain linearity, thus some form of syncing.

Anyway, I look at the photo of my cursor and I know immediately, it is PERFECT. There is nothing else here to ask. That is exactly what was supposed to be generated and displayed to my eyes. No magic, no woo woo, no neutral to earth balance, potential drift or whatever.. These are accurate snapshots of cursor moving at fixed speed. The rest is my brain, which perceives it. There is obviously ghosting, that is visible if short exposition time is used. Ghosting depends on object and background intensities, contrast. Depends on temperature too. Usually low temperatures significantly degrade panel response time and increase ghosting, can affect color calibration and such. In any case it is intrinsic to my display panel, so it's perfect by my definition.

In games too, if frames displayed are perfect, then frames are perfect. Display panel is also perfect, if it is in spec and not defective. Quality varies: you want less ghosting, better colors, you buy better display. Simple.

And if something is clearly wrong and not smooth, then it MUST BE MEASURABLE. And when objective measurements are done, it helps troubleshoot tremendously. Reading something like "feels 8% better" is laughable frankly. And then again and again some weird hypotesis continues. Please don't get ofended, this is just a personal opinion, it is ridiculous that you tested 100 windows builds, invited radio-monitoring inspection, built improvised faraday cage, tested industrial RFI filter up to 18Ghz, measured electric, magnetic fields and there is so much more on your list, I cant remember. This is nuts. And all based on feeling? Bro, this is not how it's done.

I understand your point, and I agree with you on one fundamental thing:

If the system produces correct, stable frames and timing, then the result must be correct and measurable.

Where I think we fundamentally diverge is in what exactly is assumed to be “inside the system.”

You describe a closed, well-defined pipeline:

timing source -> engine -> GPU -> display -> perception.

If that pipeline were isolated, stable, and self-referenced, I would fully agree with you.

My working assumption (based on observations, not feelings) is that in my case the reference conditions of that pipeline are not stable.

Let me be precise.

I am NOT claiming:

- mystical EMI,

- “moon phases,”

- or that digital logic randomly malfunctions without errors.

What I am claiming is much narrower:

The reference environment (neutral, ground) is stable enough to modulate timing margins, thresholds, and synchronization without causing crashes or obvious faults.

This is not exotic. Digital systems are not ideal abstractions.

Small changes here do not flip bits.

They change 'jitter, phase noise, and synchronization quality.'

Why I do not focus on internal telemetry:

I have compared the same PC, same monitor, same OS, and same drivers in a location where everything feels perfect.

Internal metrics (FPS, frametimes, clocks, temperatures, and voltages) are indistinguishable.

So internal telemetry does not correlate with the phenomenon.

What does correlate:

- neutral-to-ground voltage drifting from ~0.1 V to 5–10 V, resistance drifting from 4 ohms to 400 ohms sometimes.

- neutral behaving worse than ground under load,

- measurable DC components present on phase and neutral during outages,

- behavior changing after distribution transformer replacement,

- Day/night variation disappearing after adding a local ground reference.

These are not feelings. These are external electrical parameters.

About “8% better”:

That was not meant as a scientific unit but as an informal description of a consistent directional change.

I agree it is not a measurement, which is exactly why I am asking:

'What external parameters should be logged, and with what instruments, to quantify this properly?'

I am not attached to any hypothesis.

If measurements prove that neutral/ground instability cannot plausibly affect timing or synchronization in this way, I will drop it immediately.

But dismissing the problem as perception alone ignores one key fact:

The same hardware behaves differently in different electrical environments while reporting identical internal states.

That means something outside the logical “game engine -> GPU -> display” model is influencing the system.

That is the part I am trying to isolate.

[Inputdelay]

At this point I want to set a clear boundary; otherwise this discussion keeps looping.

I am not here to debate whether the problem “exists,” whether perception can be misleading, or whether subjective experience is reliable.

That phase is over.

What is already established:

- The same hardware behaves differently in different physical locations.

- Internal system metrics remain identical.

- External electrical parameters (neutral-ground voltage, resistance, DC offsets, transformer changes) correlate with behavior.

- Multiple independent observers notice the same effects at the same location.

If the conclusion is “this cannot happen,” then the discussion is finished-but that conclusion itself must be supported technically, not philosophically.

What I am asking for-and only this:

- A concrete physical mechanism that explains why external electrical reference instability cannot affect timing/synchronization quality in modern PCs.

OR

- Specific external parameters that should be measured to falsify my hypothesis, and the instruments required.

What I am not interested in anymore:

- Repeated explanations of how game engines should work in an idealized environment.

- Arguments based on “digital systems are perfect, therefore impossible.”

- Discussions about psychology, perception, or belief.

If no concrete mechanism or measurable parameter can be proposed, that is also an answer-but then there is nothing further to discuss.

I am here to move forward, not to relitigate the existence of the problem.

[Manul]

I understand your point, and I agree with you on one fundamental thing:

If the system produces correct, stable frames and timing, then the result must be correct and measurable.

If the system works incorrectly, the result is incorrect and also measurable. Most importantly, in what exact way it is incorrect often hints to where the problem is. For instance, it might hint what is more likely, PC, display link or display itself misbehaving, thus narrowing further troubleshooting. Anyway, I'm clearly not needed here, so I will refrain from further discussion.

[free_electron]

have you tried running the computer off a battery ? like through a UPS system ?

[Inputdelay]

have you tried running the computer off a battery ? like through a UPS system ?

With all due respect, could you please read the topic, since I have mentioned it many times using batteries and off-grid devices?

[Inputdelay]

I understand your point, and I agree with you on one fundamental thing:

If the system produces correct, stable frames and timing, then the result must be correct and measurable.

If the system works incorrectly, the result is incorrect and also measurable. Most importantly, in what exact way it is incorrect often hints to where the problem is. For instance, it might hint what is more likely, PC, display link or display itself misbehaving, thus narrowing further troubleshooting. Anyway, I'm clearly not needed here, so I will refrain from further discussion.

The only noticeable difference is the fluctuation of the core speed of the CPU and the constant change of bus speed; that's the only metric that stands out.