Windows Driver Foundation using more CPU and makes games stutter

[edwardpm89]

• I have searched open and closed issues for duplicates
• I am submitting a bug report for existing functionality that does not work as intended
• I have read the project documentation and not found a solution to my problem there
• This isn't a feature request or a discussion topic

---

Bug description

I've been using BthPS3/DShidmini for many months now without any issue, V3 version. A couple days ago I noticed my games started stuttering after a few minutes of play and troubleshooting what was happening I found the Windows Driver Foundation User-mode on the task manager using a lot more cpu than it should after a while. I'm guessing it was a windows 10 update but who knows. The only way to "fix" it is to kill the process but it comes back a few minutes later again. Happening on both Bluetooth and USB modes.

Already tried reinstalling both DShidmini and BthPS3 drivers and it still happens. It also "fixes" itself whenever I plug/unplug the usb cord, I'm sure it's just because it restarts the WUDF process.

Tested on SXS and it's working correctly, issue seems to be only on XInput

Screenshots

SystemInformer readings with stuttering:

Reference as soon as it restarts(no stutter, all normal):

Machine info

CPU-Architecture: x64 AMD
Windows version: Windows 10 22H2
Software/driver version(s): BthPS3: 2.17.0 / DsHidMini: 3.5.1
Using xinput HID mode

[minichris]

I'm also getting this with Windows 10 22H2

[zment4]

Just noticed this too. Feels like a leak of something, but no idea where and how and what.

As a workaround was thinking of automating the WUDFHost.exe killing every minute or so.

EDIT: Tested with latest v2 drivers too, and it seems to exhibit the same kind of behaviour, except a lot more mitigated - in v2 the CPU used in WUDFHost.exe occasionally drops down to near 0%, hasn't risen over 10% yet, and the stutter (if it's there at all) isn't as noticeable. Yes, the stutter is as noticeable. Almost want to point the finger to Windows doing windowsy things, but at the same time there perhaps is a potential workaround to mitigate it.

[edwardpm89]

Just noticed this too. Feels like a leak of something, but no idea where and how and what.

As a workaround was thinking of automating the WUDFHost.exe killing every minute or so.

EDIT: Tested with latest v2 drivers too, and it seems to exhibit the same kind of behaviour, except a lot more mitigated - in v2 the CPU used in WUDFHost.exe occasionally drops down to near 0%, hasn't risen over 10% yet, and the stutter (if it's there at all) isn't as noticeable. Yes, the stutter is as noticeable. Almost want to point the finger to Windows doing windowsy things, but at the same time there perhaps is a potential workaround to mitigate it.

My fix is using SXS, it's working fine on every game I've tried for now.

[kot7k]

This happened to me too. I was going nuts killing Windows driver foundation until i found out It was the connected PS3 controller. Started to happen recently so i guess It was a Windows update.

[TsundereSupreme]

Also getting the same issue on Windows 10 22H2. Stuttering begins even when a game isn't open, and persists after I close a game. Unplugging the controller's USB cord solves the issue, though of course it returns once it's plugged back in.

[AgusMarcos1510]

same here, i dont know when it started doing this, it was working fine 2 months ago from what I remember, i turn the controller on, fps drops, a while later audio starts to stutter. so i guess it was a windows update, lets see what happens in the meantime, its kinda a good thing that im not the only one with this issue

[AgusMarcos1510]

i followed the tip of using it as sxs, running games through steam and enabling play station controller support in controller settings and its getting detected as an xbox controller, so thats the fix for now

[nefarius]

Thanks to user Vik on Discord I have some diagnostic data to poke around in now.

I suspect this is due to a fairly recent Windows Update since the offending Microsoft driver xinputhid.sys is the high CPU consumer:

Which has been updated recently:

Looking at the driver traces we appear to get hammered with repeated rumble off packets that cause the processing overhead:

When I get rumble via XI I do enqueue and complete immediately since the DS3 is susceptible to getting packet-DDoS'ed. This opens a gap where the upper driver can basically send us hundreds of requests a second without any throttle which explains high CPU.

I have an up-to-date Windows 11 machine I can try to reproduce it with. I am quite sure that either the behavior of this driver or some component sitting above it has changed and wasn't as spam-heavy in the past which flew under the radar of the processing design decisions.

There's always something... If I can repro it it should be a fairly easy fix 🤞

[minichris]

If it helps, it happens to me a lot playing Forza Horizon 6

[nefarius]

If it helps, it happens to me a lot playing Forza Horizon 6

Are you getting/feeling vibration? Would help if affected folks join the Discord so I don't have to repeat everything everywhere 😛

[nefarius]

Talking about vibration, I remember having to disable it completely because in some games, the camera would start twitching sideways (nfs the run and most wanted 2012 for example)

That's unrelated, pls stay on topic.

[nefarius]

Made a help article for everyone who can contribute data: https://docs.nefarius.at/projects/DsHidMini/v3/Debugging-the-drivers/