100
Measuring input latency on Linux: X11 vs Wayland, VRR, and DXVK - Marco Nett (marco-nett.de)
submitted 3 months ago by to c/linux@programming.dev
25 comments fedilink hide all child comments
 

Really cool research. Overall, X11 still wins compared to Wayland(kwin) but only by 0.14 - 0.22 ms. The really bad performance hit is from XWayland which added 3.13 ms.

you are viewing a single comment's thread
view the rest of the comments
[–] 29 points 3 months ago (8 children)

Wtf no 100ms is absolutely unacceptable. Anything over 30 ms becomes easily noticeable. You can feel the difference between preformance dlss (30ms added latency) and 60 on quality. 10ms difference is whatever but its one part of a stack and it can compound if every part of the stack neglects latency optimisation

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] 10 points 3 months ago (1 child)

    Whenever US players find themselves on one of the AU hosted World of Warcraft servers with 200ms latency they can't stop complaining about how awful the game is to play.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 6 points 3 months ago

    When I find myself on servers like that in any game I pretend I'm training for operating a Martian Rover and see it as an extra challenge. I just hope the ping in consistent.

  • permalink
  • fedilink
  • source
  • parent
  • [+] -18 points 3 months ago (5 children)

    It takes you ~100-400ms just to blink... No, you have a placebo effect on latency. You have been trained by marketing to "feel" the difference.

    I think if anything, a response time with frame rate, BFI, and other technologies can be perceived as faster or slower. I remember watching my first 120hz TV for a movie... 24fps x 5 = 120hz so I worked out to be really smooth. 24 x 2.5 = 60hz so it wasn't really "good looking" perspectively for an uneven fps/refresh rate, even if the 60hz monitor/TV had better specs, the way the 24 fits the 120 made appearance better.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 13 points 3 months ago (2 children)

    Blinking has no relationship with input latency. Its not speculation people who play games universally agree that latency is noticeable. Put some on a game at a lan then put them in an online match at 140ms and you won't find a single person who doesnt notice.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] -5 points 3 months ago* (1 child)

    I'm not saying it does. I'm just saying your argument of 30ms being noticeable is null as blinking is ~10x that.

    Anyway your new argument is online competitive play where Ethernet might have 30ms server RTT vs someone with 140ms wireless, sure, bc even if your reaction time is 100ms better than the other guy, he still has a 10ms advantage over you, minimum.

    Anyway, now we're all over the place. Look at the figure given in the article that gives a higher level of Total latency

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 11 points 3 months ago

    End to end system latency on a good system is around 15-30ms. I'm saying that I can absolutely tell the difference between end to end system latency of 15-30ms and 60-75ms. Its pointed out and noticed constantly by players and reviewers its one of the biggest strikes against dlss and other framegen. The 100ms network example was an extreme one to highlight that its noticeable even though its "only a blink of an eye".

    If any part of the stack increases input latency it quickly degrades the experience. Lucky 10ms on an already very efficient system does not make Wayland a deal breaker but if that was 30ms difference it would be.

  • permalink
  • fedilink
  • source
  • parent