Closed Bug 571386 Opened 14 years ago Closed 6 years ago

WebM skips many frames

Categories

(Core :: Audio/Video: Playback, defect)

Product:
Core
Component:
Audio/Video: Playback
Platform:
x86
Linux
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED INACTIVE

People

(Reporter: donrhummy, Unassigned)

References

(
URL
)

Details

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.3a4webm) Gecko/20100518 MozillaDeveloperPreview/3.7a4webm
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.3a4webm) Gecko/20100518 MozillaDeveloperPreview/3.7a4webm

This is really feedback but it is technically a bug for the webm alpha. I just wanted to help out the developers so firefox's webm implementation can get better:

1. The video skips a LOT of frames, making it a pretty poor video experience
2. The image quality is phenomenal! The quality is better than what I get with flash
3. The audio is good and might be a tad off from the video but it's hard to tell as it skips so many frames

Reproducible: Always

Steps to Reproduce:
1. Watch a webm video in firefox on linux
2.
3.
Actual Results:  
Laggy, skipping video.

Expected Results:  
Smooth video showing all frames.
Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.3a5) Gecko/20100610 MozillaDeveloperPreview/3.7a5

Video plays fine for me with 3.7a5build1 on Ubuntu 10.04 64-bit.  Could you please attempt to reproduce this on the same build?  Thanks.
(In reply to comment #1)
> Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.3a5) Gecko/20100610
> MozillaDeveloperPreview/3.7a5
> 
> Video plays fine for me with 3.7a5build1 on Ubuntu 10.04 64-bit.  Could you
> please attempt to reproduce this on the same build?  Thanks.

Unfortunately, I don't have 64-bit. My comparison was with running it in flash (and also running theora video in firefox 3.6). I don't get any skipping or lag in flash (nor with any theora videos).

Out of curiosity, what processor and how much RAM do you have?
Component: General → Video/Audio
Product: Firefox → Core
QA Contact: general → video.audio
Version: unspecified → Trunk
donrhummy@yahoo.com, what CPU/video card/video driver do you have?

If you can, please try this and let us know if playback is any smoother.  When you're about to watch a WebM video on YouTube, open "View Page Info" in the browser context menu, select the "Media" tab, find the direct URL of the video and copy/paste that URL into the location bar.  That'll open the video directly and removes YouTube's default video scaling from the performance equation.
(In reply to comment #3)
> donrhummy@yahoo.com, what CPU/video card/video driver do you have?
> 
> If you can, please try this and let us know if playback is any smoother.  When
> you're about to watch a WebM video on YouTube, open "View Page Info" in the
> browser context menu, select the "Media" tab, find the direct URL of the video
> and copy/paste that URL into the location bar.  That'll open the video directly
> and removes YouTube's default video scaling from the performance equation.

It is SIGNIFICANTLY better. However it's still not 100%. There's still some lag or occasional frame freezes for a fraction of a second and the sound is just slighty off every now and then. Still not as good as flash but if it was like that normally, I could live with it as a first iteration.

I have not yet tried it but I've heard there are no problems with Opera's webm. I'll try it and let you know the results/comparison.

As for comp. info:
AMD Sempron(tm) Processor 3100+
Speed:  1,808.17 MHz

Total memory (RAM):  1.4 GiB
Free swap:  2.0 GiB

nVidia Corporation G98 [GeForce 8400 GS
Are you by chance using 720p instead of the default 360p setting?
(In reply to comment #5)
> Are you by chance using 720p instead of the default 360p setting?

Nope.
Well, I have the same problem on 64 bit Linux, but only with 720p. Flash can handle 720p with ease though.
I'm no longer having this problem after the last few days of updates. I'm using 64-bit Linux.
(In reply to comment #8)
> I'm no longer having this problem after the last few days of updates. I'm using
> 64-bit Linux.

Great! I'll have to try the latest build and see if it fixed this.
(In reply to comment #8)
> I'm no longer having this problem after the last few days of updates. I'm using
> 64-bit Linux.

We enabled optimized assembly on 64bit Linux a few days ago, that is probably why you're getting better performance now.
donrhummy: can you test this issue again with last night's nightly? We landed some changes recently which should help with this.
(In reply to comment #11)
> donrhummy: can you test this issue again with last night's nightly? We landed
> some changes recently which should help with this.

I would except for this issue: https://bugzilla.mozilla.org/show_bug.cgi?id=611651
(Using Firefox 4 Beta 7 deleted all my cookies/website settings)

I just added back all my logins for sites and I really don't want to have to do it again. I'll see if I can find another computer to try it on but losing all my logins is just a bit too much of a hassle right now. Sorry. :( (NOTE: I'm assuming 3.6.x still doesn't work with WebM, correct?)
create a Backup of your Firefox profile....
(In reply to comment #12)
> (In reply to comment #11)
> > donrhummy: can you test this issue again with last night's nightly? We landed
> > some changes recently which should help with this.
> 
> I would except for this issue:
> https://bugzilla.mozilla.org/show_bug.cgi?id=611651
> (Using Firefox 4 Beta 7 deleted all my cookies/website settings)
> 
> I just added back all my logins for sites and I really don't want to have to do
> it again. I'll see if I can find another computer to try it on but losing all
> my logins is just a bit too much of a hassle right now. Sorry. :( (NOTE: I'm
> assuming 3.6.x still doesn't work with WebM, correct?)

You can also start the browser with the -P option, which will start the profile manager and give you the option to create a new profile.

We'll assume this is fixed until we get feedback to say otherwise.
Status: UNCONFIRMED → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Status: RESOLVED → VERIFIED
> You can also start the browser with the -P option, which will start the profile
> manager and give you the option to create a new profile.
> 
> We'll assume this is fixed until we get feedback to say otherwise.

I just tried this with Firefox 4 beta 11 and WebM is performingly poorly again! It skips a lot of frames and is very jerky, even at 360p. 

Using this:
Linux OpenSUSE
Processor (CPU):   AMD Sempron 3100+, 1,808.35 MHz
Total memory (RAM):  1.4 GiB
Free memory:  333.2 MiB (+ 648.3 MiB Caches)
Free swap:  2.0 GiB
GPU: nVidia GeForce 8400 GS
With Firefox 4.0.1 (and the specs listed above) still getting very poor WebM playback relative to Flash playback.
Status: VERIFIED → UNCONFIRMED
Resolution: FIXED → ---
Can you please paste the graphics section of about:support into the bug?
This is all it has:
GPU Accelerated Windows 0/1
Using Firefox 7.0.1, WebM/HTML5 video is awful.

Test Video: http://www.youtube.com/watch?v=a6cNdhOKwi0

1. With Flash (480p, full screen) - runs perfectly, smoothly

2. With HTML5/WebM (480p, smallest) - runs well

3. With HTML5/WebM (480p, expanded to 30% larger) - runs very poorly, audio AND video skips

4. With HTML5/WebM (480p, full screen) - barely runs, audio AND video skips

As you can see, full-screen flash runs better than even the smallest version of WebM!

This is my about:support graphics section:

Adapter Description: NVIDIA Corporation -- GeForce 8400 GS/PCI/SSE2/3DNOW!
Driver Version: 3.3.0 NVIDIA 275.21
WebGL Renderer: false
GPU Accelerated Windows: 0/2
@donrhummy, can you compare behaviour in Firefox 10.0a1 Nightly to your results in comment 19?
(In reply to Anthony Hughes, Mozilla QA (irc: ashughes) from comment #20)
> @donrhummy, can you compare behaviour in Firefox 10.0a1 Nightly to your
> results in comment 19?

Sure, can you post a link to that nightly? Thanks.
(In reply to donrhummy from comment #21)
> Sure, can you post a link to that nightly? Thanks.

http://nightly.mozilla.org/
(In reply to Anthony Hughes, Mozilla QA (irc: ashughes) from comment #20)
> @donrhummy, can you compare behaviour in Firefox 10.0a1 Nightly to your
> results in comment 19?

OK, I tried it and found it to be basically the same. *Maybe* a slight difference in the expanded view (still only OK) but the full-screen view was even worse than before.

Are there some settings in about:config you want me to try?
For the various tests in comment 19, how much CPU is Firefox using for each?  It sounds like there's not enough processing power available when scaling is involved.  Are you explicitly choosing the 480p version, and does the 360p version run any better?  How big is the display/window for fullscreen?  Have you tested Chrome with the same steps as comment 19? How does it fare?
(In reply to Matthew Gregan [:kinetik] from comment #24)
> For the various tests in comment 19, how much CPU is Firefox using for each?
> It sounds like there's not enough processing power available when scaling is
> involved.  Are you explicitly choosing the 480p version, and does the 360p
> version run any better?  How big is the display/window for fullscreen?  Have
> you tested Chrome with the same steps as comment 19? How does it fare?

I just tried Chrome with that same URL (and I left Firefox running as well - forgot to shut it down, so technivally there was less available RAM for Chrome) and:

1. Chrome ran it perfectly at 480p and full screen

2. When Chrome does full screen on an HTML5 video, there's no window decoration. All you see is the video (Firefox's HTML5 fullscreen has the window borders and titlebar, etc)

(Chrome definitely ran HTML5 for that video, I double checked by right-clicking and getting the HTML5 context menu popup)

So the problem is NOT CPU, GPU or RAM related, because for Chrome, it not only ran it perfectly but I also had Firefox open as well. When I ran Firefox 10a in my previous test, nothing else was open.
Hm, actually, testing with Chrome isn't as useful as I thought since they switched VP8 decoders from libvpx to ffvp8 around 15, so that will be one reason for Chrome to be faster.  I'm not sure that would explain such a big difference in behaviour.  It'd be useful to know the CPU utilization numbers for Firefox vs Chrome on your machine with this testcase.
Another difference (on my machine) is that Chrome is using HW acceleration on pages containing <video>, whereas Firefox is not.  I discovered this by enabling the FPS counter in about:flags (which is only displayed with HW accelerated content).  If the same is true on your machine, that'll explain the majority of the performance difference--scaling video content on the CPU is fairly slow.
And if Chrome is using HW acceleration on your machine, I'd be interested to know whether the playback performance is still acceptable through the range of tests in comment 19 if you run Chrome with the argument "--disable-accelerated-video".
My old IBM X60s (1.66Ghz Core Duo) is slow enough (and doesn't get HW acceleration in Chrome) to reproduce this if I limit the browser and Xorg to a single core.  480p expanded mode is aggressively frame skipping (maybe bug 688363) in Firefox, but plays smoothly in Chrome.  Both the browser and Xorg process are using significantly more CPU in Firefox.

I'll investigate further when I have time.

(In reply to donrhummy from comment #25)
> 2. When Chrome does full screen on an HTML5 video, there's no window
> decoration. All you see is the video (Firefox's HTML5 fullscreen has the
> window borders and titlebar, etc)

That'll be fixed by bug 545812.
Status: UNCONFIRMED → NEW
Ever confirmed: true
(In reply to Matthew Gregan [:kinetik] from comment #29)
> My old IBM X60s (1.66Ghz Core Duo) is slow enough (and doesn't get HW
> acceleration in Chrome) to reproduce this if I limit the browser and Xorg to
> a single core.  480p expanded mode is aggressively frame skipping (maybe bug
> 688363) in Firefox, but plays smoothly in Chrome.  Both the browser and Xorg
> process are using significantly more CPU in Firefox.
> 
> I'll investigate further when I have time.

Sounds great, let me know if you need me to do anymore testing.
Any progress on this? HTML5 videos are still quite poor in Firefox.
Firefox 9, they're still only OK. Fullscreen HTML5 video (the one at Firefox 9's new page after install) is still very jerky. Any testing I can do to help?
(In reply to Matthew Gregan [:kinetik] from comment #29)
> That'll be fixed by bug 545812.

Anything more to be done here? Fullscreen HTML5 video is still choppy (as per comment 32).
Component: Audio/Video → Audio/Video: Playback
Mass closing do to inactivity.
Feel free to re-open if still needed.
Status: NEW → RESOLVED
Closed: 14 years ago → 6 years ago
Resolution: --- → INACTIVE
You need to log in before you can comment on or make changes to this bug.