gxquickly / angleproject

Automatically exported from code.google.com/p/angleproject
Other
0 stars 0 forks source link

Chrome 26 vs Chrome 41 (screen tearing) #975

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago
Please check out this WebGL benchmark:
http://necromanthus.com/Test/html5/test8.html 
Press "F" for full screen.

Tested on a dual-boot computer (Core 2 Duo E6400, Nvidia GTS 450):
1) Windows XP SP3 with Chrome 26
2) Windows 7 SP1 with Chrome 41

1. 59-60 FPS, very smooth motion, no tearing at all
2. 59-60 FPS, very smooth motion, severe tearing (horizontal lines).

Monitor set at 1600x900 and refresh rate at 60Hz.

In my opinion the issue is related to "requestAnimationFrame".
Chrome 26 is in perfect sync (no matter if the monitor refresh rate varies 
around 60)
Chrome 41 is out of sync (it's probably locked to 60).

all the best

Original issue reported on code.google.com by necroman...@gmail.com on 14 Apr 2015 at 1:43

GoogleCodeExporter commented 9 years ago
You should look into it.
Because it's about D3D9 vs D3D11 (the bug is related to D3D11).

Original comment by necroman...@gmail.com on 18 Apr 2015 at 12:44

GoogleCodeExporter commented 9 years ago
Not able to repro tearing on NV Quadro K600 & Chrome 43. Is this still 
happening for you?

vsync & frame syncing issues might point to something in Chrome's interaction 
with ANGLE, rather than ANGLE itself. CC'ing bajones@ to that end.

Original comment by shannonw...@chromium.org on 28 Apr 2015 at 5:53

GoogleCodeExporter commented 9 years ago
Yes, it still happening for me (and not only).
I've updated the benchmark (more processing power is needed):
http://necromanthus.com/Test/html5/testA.html 
On Chrome 26, 59-60 FPS, very smooth motion, no tearing at all.
On the latest Chrome 42, 59-60 FPS, very smooth motion, but tearing (horizontal 
lines).
Tearing occurs on most of the computers capable to render a frame in less than 
16ms (more than 60Hz).
But only for D3D11.
And I think you're right: frame syncing issues points to something in Chrome's 
interaction with ANGLE.

Original comment by necroman...@gmail.com on 3 May 2015 at 7:22