| Summary: | [nvidia] Poor performances on second monitor with on-demand prime profile | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | Arnaud <arnaud.vergnet> |
| Component: | multi-screen | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED UPSTREAM | ||
| Severity: | normal | CC: | aleixpol, andre.vmatos, nate, notmart, stefan.hoffmeister, xaver.hugl |
| Priority: | NOR | ||
| Version First Reported In: | 5.24.0 | ||
| Target Milestone: | --- | ||
| Platform: | Other | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
|
Description
Arnaud
2022-02-12 23:00:32 UTC
I think it would be very helpful to describe the physical connector setup you are running. Case in point: On X11, I have a setup where * Intel iGPU (only) controls internal display and HDMI * Nvidia dGPU (only) controls the USB-C output path (i.e. DisplayPort Alternate Mode et al) * Nvidia is in PRIME offload mode; Intel is primary Now connect an external 4K screen to the Nvidia output path, i.e. USB-C / DisplayPort. Now the *Xorg* process starts consuming between 25% and 40% of one CPU on an otherwise totally idle system. Remove the 4K screen from the Nvidia output path, attach it to the HDMI port (Intel) Now Xorg is totally fine. Some sleuthing with perf suggests that all that CPU is burnt on getting the current system time (gettimeofday / clock_gettime) via vdso and kernel calls, with this originating from the Nvidia driver (510). My computer only has a single HDMI port. From what I could test, it seems this port is connected to the Nvidia dGPU because I cannot make the second monitor work with only the Intel iGPU powered on. On Xorg, Xorg and its driver modules are responsible for multi gpu support. Please report this to NVidia |