Bad mousetester graph when videocard is on

[Slender]

In General, I get a strange behavior of the mouse when the driver graphics card on ( screen 1 and 2), sometimes after a reboot I get a smooth graph as usual (picture 3) that, try:
disable MSI mod, remove affinity from the card and the controller, change the drivers etc, nothing helps, after I turn off the graphics card everything becomes OK
spec: 1080ti 9700kf 16ram

[timecard]

Video card and drivers is responsible for drawing the screen/cursor. DWM and explorer may be included in that process depending on a few factors. CSRSS takes the input from the driver and is responsible for dispatching information to the necessary processes, handles input processing for keyboard/mouse. Which device/USB drivers depends on the version of windows, USB2 only, USB3 only, USB2 and USB3 mixed (usually a little worst since it adds DPCs between both drivers even if there aren't both types of devices connected, MS life).

If you disable your nvidia driver (nvlddmkm) on windows it falls back to BasicDisplay.sys/BasicRender.sys to help communicate with the video card and draw the cursor/screen. This is what safemode does or before you install your amd/nvidia graphics driver.

Nvidia itself forces Desktop Window Manager (dwm) and Explorer (Windows Explorer) to use "adaptive" power profiles which can "override the global default" behavior which could cause the gpu to run at a lower speed (power state, PState). This information is hidden unless you manually add the binaries to nvpanel or use Nvidia Profile Inspector. Adjusting the power profiles and "other settings" for dwm/explorer may reduce some variance in your graph when active.

Feel free to give more detail on Windows version you're using or other details, and what you mean by "turning off the graphics".

Screenshot of window scheduler as it relates to mousetester from start to end, process duration by thread priority. (doesn't show call stacks/ready stacks ofc) Keep in mind this is microseconds view as data passes from driver to mousetester.

Lifeinthewindowsscheduler.PNG (37.02 KiB) Viewed 5261 times

[Slender]

Then how can I explain that the rest of the time I get a smooth and even schedule with a working nvidia driver?
I am using an asus z390 tuf motherboard and a 21h1 system.
I used VBIOS EVGA for my video card on which Kboost is active, so I have the maximum stable frequency on my card.
At the moment, I have returned to my native bios and tweaked the frequency using msi afterburner.
im reinstalled driver and on default nvcp settings im get this.
But sometimes after 10 reboots i have normal graph and gameplay.

[Slender]

im use pci-e usb 2.0 controller, he is not have msi mode, but i put my mouse in motherboard controller and give me same result

[TN_fun]

Is the first pic bad? Is this some kind of joke? I don’t understand at all how you can do this on windows 10. It’s perfect.

[Slender]

Is the first pic bad? Is this some kind of joke? I don’t understand at all how you can do this on windows 10. It’s perfect.

yes, it's 10, and in the first screenshot the "perfect launch", otherwise the graph looks like in the 1st and 2nd picture.
I apparently found the reason and it's DWM.

[TN_fun]

I don't know dude, my minimum is 950-1050 when all programs are closed and the power plan is maximum performance. If I open the browser it immediately becomes bad, the delta is 800-1400 on the graph. Therefore, all your pictures are ideal for me, any of them. I sincerely do not understand how you can get such a values and what needs to be done for this.

[Slender]

I don't know dude, my minimum is 950-1050 when all programs are closed and the power plan is maximum performance. If I open the browser it immediately becomes bad, the delta is 800-1400 on the graph. Therefore, all your pictures are ideal for me, any of them. I sincerely do not understand how you can get such a values and what needs to be done for this.

i can try to help you

[Slender]

I don't understand how it works, but after the PC is completely turned off, de-energized, and from the first turn on I get an even schedule, ehh, apparently I will have to join the EMI sect.

[Chief Blur Buster]

I don't understand how it works, but after the PC is completely turned off, de-energized, and from the first turn on I get an even schedule, ehh, apparently I will have to join the EMI sect.

Sometimes it's packet congestion between the multiple USB devices on the same USB host chip.

Imagine USB like an Internet connection, except jitter at the smaller timescales (often in hundreds of microseconds). Imagine "USB Root" like an ISP, with everything below sharing the bandwidth. Now imagine ping jitter of all the competing USB bandwidth users.

Try purchasing and installing PCI Express USB card just only for your mouse and see what happens too. You want a try a dedicated USB host chip for your mouse, preferably with a dedicated PCIe lane separate from everything else (ideally), for purest possible USB mouse operation -- literally a dedicated connection just for your mouse!

But yes, EMI can be an issue as USB does error correction when strong over-the-air interference causes packet loss on a USB cable -- like a nearby transformer cycling on/off near the USB cable. Just like an Internet connection, USB cables are packetized, and there can be packet loss on USB cables with sufficiently strong interference near the cable. USB cables often dangles into a mess of power wires and power bricks behind the desk, a hornet's nest of potential possible EMI. The mouse's built in USB implementation just simply retransmits the packet associated with a poll (late/jittered) until the USB host acknowledges the packet.

While inexperienced people troubleshooting EMI sometimes sometimes resembles a sect, EMI itself is a real problem, all covered by real science. For the easy-to-read stuff, see Electronics Notes Science to understand the real, hard, brutal science behind interference. For the more complex-to-read summaries, see science papers heavily covering EMI affecting computers, such as Institute of Electrical and Electronics Engineers (IEEE).

But first, try to solve your USB jitter/congestion by having a dedicated USB host chip just only for your high-Hz mouse (nothing else in the same USB host tree in Device Manager). A dedicated root hub just only for your mouse. You can try playing USB roulette (external hub plugged onto a distant motherboard USB port for your other USB devices on a different USB host chip), keeping your motherboard USB ports empty except for the mouse. Could be the front panel, could be the separate cluster of USB2 ports versus USB3 ports, could be the USB-C port (use USB-C to USB-A adaptor), etc. They may be grouped to their own dedicated USB chip logic.

That said, there are advantages to the PCIe USB card route if you have spare slots and spare dedicated PCIe lanes. (Keep in mind PCIe lane are also subject to jitter and packet loss too, with attendant error correction overheads for newer versions of PCIe -- but it can be much less than a heavily congested USB root hub, especially with a dedicated PCIe lane).