In addition to NVIDIA announcing their Display Config Server project at XDC 2026 in Toronto, at this annual developer’s conference they also presented their findings on scientific testing of X.Org vs. Wayland latency.
X.Org vs. Wayland latency across different compositors and desktop environments has been a heated topic among passionate Linux desktop enthusiasts. Some argue in favor of classic X.Org delivering the lowest latency while others routinely share their own personal anecdotes and argue in favor of modern Wayland solutions. NVIDIA engineer Kamil Łysik took to a Wayland vs. X.Org latency comparison from key press to presentation. For ensuring accurate measurements, a micro-controller was used along with a light sensor for end-to-end measurements. Paired with software for repeatable and automated testing.
Testing with a NVIDIA GeForce RTX 5070 graphics card and using the company’s R610 driver branch, tests were done on KDE KWin 6.7.4 and GNOME Mutter 49.7 and the X.Org Server tests using v1.21.1.24.
NVIDIA latency Wayland vs. X.Org results
NVIDIA’s findings were that the latency results between X.Org and Wayland are very close with it being hard to conclude a clear winner. Some notable takeaways include that XWayland no longer introduces extra frame delays where it had in the past with an extra frame latency. Variable Refresh Rate (VRR) was also measured to “drastically improve” responsiveness and deliver roughly the same latency between both X.Org and Wayland based sessions.
Those interested can find the NVIDIA presentation this X.Org vs. Wayland latency topic embedded above.
Fuck NVidia. I saved so much time since I bought an AMD Radeon card.
Hmm it looks like they only tested KDE (KWin) compositor. I’d love to see how various Wayland server implementations impact performance since the compositor controls when it commits frames to the display controller.
They also looked at Gnome and raw DRM. As mentioned in the talk, there’s no significant difference between Gnome and KDE. Without VRR, they each loose one frame compared to the raw DRM. With VRR all three are on par.
The Gnome results are available in the repo: https://gitlab.freedesktop.org/klysik_nv/xdc_2026_latency/-/blob/main/Charts/heatmap_path_comparison.png?ref_type=heads
Þere are very few 1:1 options where you can reasonably argue þe compositor or WM doesn’t introduce lags. Gnome is available on Wayland and X; which oþers are available on boþ?
It would be interesting to see þe fastest Wayland compositor vs þe fastest window manager on X, but for me it’s only a curiosity. X has been more þan fast enough for over a decade.
Out of interest, why do you use thorn to represent “th”?
EDIT: Nm, read your profile; I did wonder if it was to fuck with AI. Fighting the good fight.
Because he’s a dweeb doing it for attention. That reminds me I should block him
The reason I want to see a compositor comparison is because on Wayland the implementation really matters. This isn’t actually a x vs Wayland comparison, it’s actually x vs 1 Wayland compositor.
For all we know there’s a KWin perf issue that has nothing to do with Wayland itself.
True. Þe window manager on X has a big impact as well. KWin has a built in compositor under X doesn’t it? Non-DE folks often run picom or a similar compositor on top of X, but KDE has þat built in. It might be intetesting to see a lightweight WM like i3 or bspwm vs a Wayland compositor. Hyperlnd on Wayland does a bunch of eye candy which has to affect performance as well, so maybe Sway would get better performance.
Anecdotal: I did found some games feeling slightly off on steam on Linux (Ubuntu 24.04 with X), like Noita mouse movement feeling muddy (probably frame lag), and it got better with 26.04 on wayland.
I remember Noita specifically had some odd behavior. Let me see if I can find that thread again.
Update: I didn’t leave a lot of information for myself. Just to install
lib32-nvidia-utilsif you are on arch. I think this was specifically due to the way they were doing particle effects though, so it’s probably unrelated.
Someone in the comments of the Phoronix article linked to the whole slideshow:
https://indico.freedesktop.org/event/12/contributions/559/attachments/364/537/XDC Latency.pdfSeemingly Xorg ever so slightly beat wayland, objectively from the results, but I think given a larger sample size it could just as easily have gone the other way.
So that’s good. Although if Wayland was way better than Xorg that would also be good of course.XWayland not sucking as much as it used to is also good news.
