Page 2 of 2

Re: Car V2L = Clean eletricity?

Posted: 11 May 2026, 12:21
by andrelip
kyube wrote:
30 Apr 2026, 07:33
andrelip wrote:
29 Apr 2026, 15:36
Heavy mouse feeling / negative acceleration / inconsistencies with the mouse (Superlight). Poor hit registration.
Are you using your mouse in wireless or wired mode? Depending on that, it could be either EMI causing issues to your mouse in wireless mode (can happen even in wired mode) or it can be power quality related.
Wired is ideal.
andrelip wrote:
29 Apr 2026, 15:36
A high-FPS game running on a 390Hz monitor (one of the top 5 lowest input lag displays in the world) still doesn’t feel smooth or responsive.
That's a bit of a excessive & misleading claim, as I assume you mean the XV252QF (360=>390Hz OC mode), which is still bound by slow G2G RT in specific transitions.
However, getting a more consistent display (OLED) would be beneficial in troubleshooting this.
This is to iron out any potential G2G RT instabilities affecting your perception.
andrelip wrote:
29 Apr 2026, 15:36
It usually changes from day to day. Over the past few years, I’ve been really chill about it and barely active here. I even thought it was placebo, or that the 14900K had solved it. But since I moved from a modern apartment to a house, it got much worse and consistently bad. My GC rank dropped from top 20 to 16, and my Faceit level went from 9 to 6.

Same PC, and I know it could be the game routing, the new ISP provider (despite perfect reports), a Windows update, me getting older, or other factors. But now I’m really starting to believe these “dirty electricity” claims.
Tried on the car and the noise signature in the mouse tester was really different. So 100% sure eletricit affects the components. In V2L the car is an isolated island, so ground do not work properly.

That's way too many variables to be able to claim it's electricity related with absolute certainty IMO.
As you've mentioned, networking is a very variable. You could try playing some offline game, with all networking disconnected, to see if you can experience these symptoms you're feeling.
If you could test multiple PCs in that specific apartment, that would be great too.
5 different rigs (different CPU+MB+DRAM; rest should be identical) for a longer time frame (6 months) should reveal a clearer picture whether your issue is rooted in that or not.
It would also help if you'd try playing on XMP off, with overclock dialed in to the extent where you aren't unstable (as vague as the term stability is)
The last 2 components are essential and most users in these forums misinterpret unstable hardware for EMI-related issues.
"Unstable" hardware isn't solely hardware which BSODs on launch, there's whole spectrum of micro-issues which it can produce.
Hardware can also become "unstable" from too much stress-testing / degradation of internals. I've recently come across someone with a X3D CPU with this (potential) culprit, as the 3D cache is sensitive.
It's why I keep on being vocal about ECC UDIMMs & AM5 being a great way to iron out potential issues in that regard.
Tried it on the car and the noise signature in MouseTester was really different. So I'm 100% sure that the electrical supply affects the components.

I tried to isolate as much as I could:

- Put everything on the exact same outlet to share ground potential.
- The PC is not connected by cable to anything other than the monitor.
- Tested potential between the PCIe brackets, case screws, and everything — all show continuity and the same potential.
- Moved from Ethernet to Wi-Fi 7 on the 6GHz band.
- Switched to a fiber optic HDMI cable (though there's still continuity between the connectors, so a minimal ground loop may occur — only ~0.01V difference).
- Modified MouseTester with Claude Code to avoid garbage collector pauses, list hooks on the USB devices, and identify other devices sharing the hub.
- Also modified MouseTester to send fake mouse input from an async thread, so I can isolate failed inputs from failed input *capturing*. It ran perfectly smooth with the injected inputs, which means any oscillation I see is actually coming from the real input path — not the measurement.
- Disabled RGB, Bluetooth, and other devices sharing the USB hub.

Final test: MouseTester now fluctuates between 999.98 Hz and 1000.02 Hz. The game feels butter smooth and micro-corrections feel perfect. I also disabled idle states to rule out power saving (stable 5400 MHz). Beyond frequency × time, the X, Y, and counts graphs are really revealing too — they expose batching and jitter patterns that the frequency plot alone can hide.

My diagnosis: what many people call "input lag" is actually erratic mouse polling. Sometimes the system batches inputs, causing both delays *and* batching artifacts with interpolation — and it's even worse because the rate oscillates.

This lines up with the fact that NVIDIA FrameView reports low input lag, and that AMD's Frame Latency Meter (FLM) injects fake frames during benchmarking and the game moves really fast under it. I strongly believe many "high input lag" complaints are actually *input* latency (polling-side) rather than frametime or PC-side input lag.

Next steps: Test each fix individually to see what actually moves the needle in MouseTester — including testing *without* the ground connection. Also compare against a PCIe card that supports 8000 Hz polling (ASM3142 - waiting for it to arrive).

Additional findings: The interpolated MouseTester noise have a wave oscillating at 16ms (before and after the fix). Initially I thought it was related to the 60Hz of my electrical network, but then I tested it on the car's V2L (which is 50Hz) at both PC and Monitor and it stayed at 16ms. Then I suspected the monitor refresh rate (it was 240Hz), so I dropped it to 24/30Hz — no change. And despite the lack of precision, it's so regular that it really looks like exactly 16ms, not 16.67ms.

Re: Car V2L = Clean eletricity?

Posted: 11 May 2026, 15:06
by kyube
andrelip wrote:
11 May 2026, 12:21
Tried on the car and the noise signature in the mouse tester was really different. So 100% sure eletricit affects the components. In V2L the car is an isolated island, so ground do not work properly.
Just to be certain – have you used the “Interval vs Time” setting?
It's extremely difficult to assess driver behavior consistency using the mouse... it can lead to faulty conclusions.
I'm very certain that it's likely not even possible to assess any kind of EMC-related issue with this software.
All MouseTester does is measure .etl data, namely bInterval.. or ISR Interval.
I personally haven't been able to reliably test that, for instance, grounding impacts graphs in any way possible.
andrelip wrote:
11 May 2026, 12:21
I tried to isolate as much as I could:

- Put everything on the exact same outlet to share ground potential.
- The PC is not connected by cable to anything other than the monitor.
- Tested potential between the PCIe brackets, case screws, and everything — all show continuity and the same potential.
- Moved from Ethernet to Wi-Fi 7 on the 6GHz band.
- Switched to a fiber optic HDMI cable (though there's still continuity between the connectors, so a minimal ground loop may occur — only ~0.01V difference).
- Modified MouseTester with Claude Code to avoid garbage collector pauses, list hooks on the USB devices, and identify other devices sharing the hub.
- Also modified MouseTester to send fake mouse input from an async thread, so I can isolate failed inputs from failed input *capturing*. It ran perfectly smooth with the injected inputs, which means any oscillation I see is actually coming from the real input path — not the measurement.
- Disabled RGB, Bluetooth, and other devices sharing the USB hub.

Final test: MouseTester now fluctuates between 999.98 Hz and 1000.02 Hz. The game feels butter smooth and micro-corrections feel perfect. I also disabled idle states to rule out power saving (stable 5400 MHz). Beyond frequency × time, the X, Y, and counts graphs are really revealing too — they expose batching and jitter patterns that the frequency plot alone can hide.

My diagnosis: what many people call "input lag" is actually erratic mouse polling. Sometimes the system batches inputs, causing both delays *and* batching artifacts with interpolation — and it's even worse because the rate oscillates.

This lines up with the fact that NVIDIA FrameView reports low input lag, and that AMD's Frame Latency Meter (FLM) injects fake frames during benchmarking and the game moves really fast under it. I strongly believe many "high input lag" complaints are actually *input* latency (polling-side) rather than frametime or PC-side input lag.

Next steps: Test each fix individually to see what actually moves the needle in MouseTester — including testing *without* the ground connection. Also compare against a PCIe card that supports 8000 Hz polling (ASM3142 - waiting for it to arrive).

Additional findings: The interpolated MouseTester noise have a wave oscillating at 16ms (before and after the fix). Initially I thought it was related to the 60Hz of my electrical network, but then I tested it on the car's V2L (which is 50Hz) at both PC and Monitor and it stayed at 16ms. Then I suspected the monitor refresh rate (it was 240Hz), so I dropped it to 24/30Hz — no change. And despite the lack of precision, it's so regular that it really looks like exactly 16ms, not 16.67ms.
The PCIe card won't help you achieve lower CPU overhead from drivers, regardless of which drivers you use.
I think you shouldn't rely on MouseTester in the slightest for your troubleshooting journey.

My personal theory is that the vast majority of users which claim “electricity related lag” usually have issues in one of the categories or a combination of some:
• Networking
• Hardware stability (either faulty HW, such as the CPU, DRAM, GPU and/or other general instability)
• Psychology
• Physiology

Re: Car V2L = Clean eletricity?

Posted: 11 May 2026, 16:02
by andrelip
kyube wrote:
11 May 2026, 15:06
andrelip wrote:
11 May 2026, 12:21
Tried on the car and the noise signature in the mouse tester was really different. So 100% sure eletricit affects the components. In V2L the car is an isolated island, so ground do not work properly.
Just to be certain – have you used the “Interval vs Time” setting?
It's extremely difficult to assess driver behavior consistency using the mouse... it can lead to faulty conclusions.
I'm very certain that it's likely not even possible to assess any kind of EMC-related issue with this software.
All MouseTester does is measure .etl data, namely bInterval.. or ISR Interval.
I personally haven't been able to reliably test that, for instance, grounding impacts graphs in any way possible.
andrelip wrote:
11 May 2026, 12:21
I tried to isolate as much as I could:

- Put everything on the exact same outlet to share ground potential.
- The PC is not connected by cable to anything other than the monitor.
- Tested potential between the PCIe brackets, case screws, and everything — all show continuity and the same potential.
- Moved from Ethernet to Wi-Fi 7 on the 6GHz band.
- Switched to a fiber optic HDMI cable (though there's still continuity between the connectors, so a minimal ground loop may occur — only ~0.01V difference).
- Modified MouseTester with Claude Code to avoid garbage collector pauses, list hooks on the USB devices, and identify other devices sharing the hub.
- Also modified MouseTester to send fake mouse input from an async thread, so I can isolate failed inputs from failed input *capturing*. It ran perfectly smooth with the injected inputs, which means any oscillation I see is actually coming from the real input path — not the measurement.
- Disabled RGB, Bluetooth, and other devices sharing the USB hub.

Final test: MouseTester now fluctuates between 999.98 Hz and 1000.02 Hz. The game feels butter smooth and micro-corrections feel perfect. I also disabled idle states to rule out power saving (stable 5400 MHz). Beyond frequency × time, the X, Y, and counts graphs are really revealing too — they expose batching and jitter patterns that the frequency plot alone can hide.

My diagnosis: what many people call "input lag" is actually erratic mouse polling. Sometimes the system batches inputs, causing both delays *and* batching artifacts with interpolation — and it's even worse because the rate oscillates.

This lines up with the fact that NVIDIA FrameView reports low input lag, and that AMD's Frame Latency Meter (FLM) injects fake frames during benchmarking and the game moves really fast under it. I strongly believe many "high input lag" complaints are actually *input* latency (polling-side) rather than frametime or PC-side input lag.

Next steps: Test each fix individually to see what actually moves the needle in MouseTester — including testing *without* the ground connection. Also compare against a PCIe card that supports 8000 Hz polling (ASM3142 - waiting for it to arrive).

Additional findings: The interpolated MouseTester noise have a wave oscillating at 16ms (before and after the fix). Initially I thought it was related to the 60Hz of my electrical network, but then I tested it on the car's V2L (which is 50Hz) at both PC and Monitor and it stayed at 16ms. Then I suspected the monitor refresh rate (it was 240Hz), so I dropped it to 24/30Hz — no change. And despite the lack of precision, it's so regular that it really looks like exactly 16ms, not 16.67ms.
The PCIe card won't help you achieve lower CPU overhead from drivers, regardless of which drivers you use.
I think you shouldn't rely on MouseTester in the slightest for your troubleshooting journey.

My personal theory is that the vast majority of users which claim “electricity related lag” usually have issues in one of the categories or a combination of some:
• Networking
• Hardware stability (either faulty HW, such as the CPU, DRAM, GPU and/or other general instability)
• Psychology
• Physiology
“Frequency × time” graph.

FLR injects fake inputs: quick left and right movements at a high polling rate. The game feels fast and responsive, and it reports low end-to-end* latency, at least from input reception to rendering.

If you try to do the same with a mouse, when the polling rate is not perfect, it sometimes groups inputs in a way that cancels them out, making it feel “smoothed/interpolated/heavy”. These small left-right flicks feel very responsive when MouseTester reports low distortion.

I tried it with a cheap Renesas USB controller, and it capped at 500 Hz. Forcing 1000 Hz registers a fixed outlier every 16 ms, which creates distortion in the other graphs. People with high-polling-rate mice report that the ASM3142 is stable at least up to 4 kHz.

I tested it in MouseTester by saturating the input with fast circular movements. The interpolation becomes really obvious when you look at the movement history, especially with small corrections when “frequency × input” drifts.

I don’t rule out stability or power-saving issues, especially because disabling idle states seems necessary for the desired stability. However, disabling turbo boost, XMP, fixing cores, and other tweaks do not seem to eliminate it.

Maybe there is a system routine running at a 16 ms interval?

Anyway, it seems to affect only the input registering / interrupt, because with fake input it's perfect.
My fork is here: https://github.com/andrelip/MouseTester

Re: Car V2L = Clean eletricity?

Posted: 11 May 2026, 19:55
by kyube
andrelip wrote:
11 May 2026, 16:02
“Frequency × time” graph.
Image

This might make you re-consider.
andrelip wrote:
11 May 2026, 16:02
I tried it with a cheap Renesas USB controller, and it capped at 500 Hz. Forcing 1000 Hz registers a fixed outlier every 16 ms, which creates distortion in the other graphs. People with high-polling-rate mice report that the ASM3142 is stable at least up to 4 kHz.
Yes... I've made a thread in regards to how certain third party USB XHCI cards behave.
See my signature.

Re: Car V2L = Clean eletricity?

Posted: 12 May 2026, 15:11
by andrelip
kyube wrote:
11 May 2026, 19:55
andrelip wrote:
11 May 2026, 16:02
“Frequency × time” graph.
Image

This might make you re-consider.
andrelip wrote:
11 May 2026, 16:02
I tried it with a cheap Renesas USB controller, and it capped at 500 Hz. Forcing 1000 Hz registers a fixed outlier every 16 ms, which creates distortion in the other graphs. People with high-polling-rate mice report that the ASM3142 is stable at least up to 4 kHz.
Yes... I've made a thread in regards to how certain third party USB XHCI cards behave.
See my signature.
Really nice. I was off the forum for years because I never had any issues until I moved to this old house. I think I’m right now where you were when you first posted it. You even mentioned grounding in that post. Unfortunately, the images have expired.

Now, with proper grounding and some type of galvanic isolation (0.1v between the optical hdmi), it’s really good again. I’m just trying to isolate the root cause and maybe verify whether there’s still room for optimization. I did other thinks like removing mouse hooks such as the Logitech filters seemed to improve it for me, especially after replacing the wireless and mouse drivers with the generic Microsoft ones. I had to fully remove the driver because it was still being injected by device ID.

In my fork, I fixed the grouping issue in OxyPlot that displayed it as 2ms even when the interval was erratic. The way the "frequency × interval" graph works is not through any type of aggregation, such as an average or moving average. It simply checks the delta between the last input and the current one, then calculates 1000 / delta. In version 1.7.2, the result was rounded.

Re: Car V2L = Clean eletricity?

Posted: 12 May 2026, 17:12
by Slender
there is no version of mt better than 1.5.3.
If you're talking to kuobe, you're talking to a very poorly trained neural network model.

Re: Car V2L = Clean eletricity?

Posted: 16 May 2026, 19:12
by Mr1991
It’s not in your head, it sounds exactly like 90+% of people are experiencing on here, yes it is power/electric related.

Re: Car V2L = Clean eletricity?

Posted: 17 May 2026, 07:33
by kyube
andrelip wrote:
12 May 2026, 15:11
Really nice. I was off the forum for years because I never had any issues until I moved to this old house. I think I’m right now where you were when you first posted it. You even mentioned grounding in that post. Unfortunately, the images have expired.
I'm somewhat cooking up a follow-up to the entire discussion, as I've purchased quite a few USB XHCI PCIe cards (and even EHCI) in order to find the one with the least CPU overhead.
The entire "grounding helps polling rate stability" theory is moot IMO. I wasn't able to notice a difference in my plots whatsoever, nor a difference in feel.
andrelip wrote:
12 May 2026, 15:11
Now, with proper grounding and some type of galvanic isolation (0.1v between the optical hdmi), it’s really good again.
I’m just trying to isolate the root cause and maybe verify whether there’s still room for optimization.
I did other thinks like removing mouse hooks such as the Logitech filters seemed to improve it for me, especially after replacing the wireless and mouse drivers with the generic Microsoft ones. I had to fully remove the driver because it was still being injected by device ID.
Difficult to differentiate DPC/ISR driver performance from EMI-related topics, as you've introduced many variables into the mix.
I still think most topics in regards to this are HW instability / defect related.
andrelip wrote:
12 May 2026, 15:11
In my fork, I fixed the grouping issue in OxyPlot that displayed it as 2ms even when the interval was erratic. The way the "frequency × interval" graph works is not through any type of aggregation, such as an average or moving average. It simply checks the delta between the last input and the current one, then calculates 1000 / delta. In version 1.7.2, the result was rounded.
Neat to know, thanks!