html 5 video performance

[lvm]

Quick and dirty HTML5 video performance test: how much CPU is used on my system (32-bit linux, phenom II X4 975, nvidia gt240 vdpau/libva1) to play this video: https://www.youtube.com/watch?v=v_NRSoVncJQ in 1080p

seamonkey 2.49.1 00:05:38
firefox 56.0 00:04:49
chromium 62.0.3202.75 00:02:29

Twice less efficient than chrome. Sad...

[DanRaisch]

And what do those numbers represent? They look like time figures and not percent CPU usage as you mentioned in your first line.

[therube]

SeaMonkey 2.49.1
XP SP3, Geforce 4, variable, "Auto 480", ~50 % CPU.
XP SP3, Geforce 4, variable, "1080p HD", ~87 % CPU (not viable).

Chrome 38, sandboxed
XP SP3, Geforce 4, variable, "Auto 360", ~12 % CPU.
XP SP3, Geforce 4, variable, "480", ~20 % CPU.
XP SP3, Geforce 4, variable, "1080p HD", ~65 % CPU (viewable).

[lvm]

And what do those numbers represent? They look like time figures and not percent CPU usage as you mentioned in your first line.

CPU usage is usually measured in absolute times but if you are more used to percentages you can divide the time by the video duration (2:17) and optionally by the number of cores (4). I doesn't really matter because only relative values are important - gecko-based browsers use twice the CPU amount compared to chrome. At least on linux, now I tried to do the same thing on 32-bit windows 7 and got rather unexpected results:

firefox 56.0.2 - 0:02:36.404
chrome 62.0.3202.89 - 00:02:14
seamonkey - won't do it. Or rather google won't let it do it - 720p is the highest resolution available. Weird - 1080p worked fine on linux. In 720p performance is also pretty much similar for all browsers - around 50 seconds.

So on windows CPU usages are similar. My hypothesis is that hardware video acceleration doesn't work on linux but works on windows, but it is only a hypothesis. Windows setup is completely different by the way - i5-5300u with intel HD graphics 5500 and directx 11.

[therube]

CPU usage is usually measured in absolute times but if you are more used to percentages you can divide the time by the video duration (2:17)

I'm not following?

> CPU usage is usually measured in absolute times

Are you speaking of the amount of time it takes a task to complete?
(Like what UNIX time or top may show?)

> you can divide the time by the video duration (2:17)

So we're back to CPU time?
And just what time are we measuring?
The time that it takes to play a 2:17 video (which I would think to be 2:17)?
Or are we talking the amount "CPU time" (vs. "user time", or similar) that the video playback requires?
And if so, how are we isolating the video "CPU time" from the rest of the browser process?

[lvm]

Yes, all times are the actual CPU times consumed by the process (or processes in case of chrome) as reported by top or ps or windows task manager, etc, etc.

BTW I see you are using XP - what directX version is installed on you system? (can be checked by running dxdiag) I am starting to suspect that this issue may be related to video acceleration - for some reason gecko browsers use it only on newer windows but not on linux or XP while chrome uses it everywhere. I also seem to remember that when youtube still supported flash player which did use video acceleration on linux its CPU requirements also were significantly lower than that of gecko HTML5 player.

[therube]

SeaMonkey 2.49.1 (Hardware Acceleration is disabled, I believe?)
Win7, Intel HD Graphics 4000 Direct3D11, variable, "Auto 480", ~8 % CPU.
Win7, Intel HD Graphics 4000 Direct3D11, variable, "1080p HD", ~16 % CPU.

FF 56.0.2 (Hardware Acceleration is disabled, I believe?)
Win7, Intel HD Graphics 4000 Direct3D11, variable, "Auto 480", ~8 % CPU.
Win7, Intel HD Graphics 4000 Direct3D11, variable, "1080p HD", ~14 % CPU.

FF has e10s, so giving a %, which varies as it is, is then further split between 4 processes, but thinking in general, its going to be slight less then SeaMonkey.

Chrome 38 also uses multiple processes & CPU looks to be running in the same ranges.

On Win7, & not seeing the substantial differences like I did on XP.

In all cases - in Win7, I see VP9 codec being used.
Didn't think to check what I was getting on XP.
Anyhow, codec used potentially could greatly affect results.

FF & Chrome were running with a single window/tab.
SeaMonkey with 12 windows, &, guessing, 120 tabs?

[Grumpus]

Watching a system monitor of the eight cores on this system it appears that any static statement of use depends on when you catch the core since none of it is static.
Just sitting here watching each core appears to take a turn and they fluctuate consistently between 3.7% and 8% total memory used for the entire system around 640MB.
Hardware acceleration enabled; redone with SeaMonkey.
Video had no issues and didn't blow out the cores.

[therube]

I see you are using XP - what directX version is installed on you system?

DirectX 9.0c (4.09.0000.0904) [presumably from what I call, directx_Jun2010_redist_9.29.1974.exe]