EPIA SP13000 & ML8000: Flash video, audio not right

What works, and doesn't, for you. Be specific, and please include Puppy version.
Post Reply
Message
Author
mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

EPIA SP13000 & ML8000: Flash video, audio not right

#1 Post by mskuma »

Hello

I am trying to understand the reason for a performance-related issue relating to macromedia flash & video/audio on puppy.

I've been trying Puppy 1.0.6 on a couple of Via SP13000 & ML8000 boards (512Mb/DDR400 RAM & HD). I've noticed that some Flash content is sluggish - imagine an on-screen flash button with an accompanying sound effect - when you click the button an audio effect can be heard. I've noticed there is a noticeable lag between clicking & hearing the audio. By comparison, I've installed Win98SE on the same hardware setup & found the lag is virtually non-existent by comparison. So I think hardware can be ruled out. Apart from Flash plug-in code difference (using the same base version on both OSes), a difference I can imagine is the perhaps the installation of specific video & audio chipset drivers for the Win98 case. I wonder what is the situation with Puppy? Via uses AC97 for audio.

About video - the Via uses integrated video. According to this, the ML8000 uses the VT3122 chip known as "CastleRock" & is part of the VIA CLE266 northbridge. For the SP13000, it uses the VT3118 chip with "Unichrome Pro" (A) found on the CN400 (and PM800, PM880 and PN800). I think this is very specific to Via - what would puppy use in this case? If it's generic, could it cause the issue I notice?

I see that Via released source code for "console framebuffer driver" for the graphics chipset (the prev link might be better). Would using this help? If so, how can I try it.

Apparently epios (another distro) is targetting via specifically, and I think ubuntu includes via-specific drivers (at least for graphics). So I guess trying these would distros might answer my question.. before I do, has anyone had any experience with seeing variation with puppy audio/video performance using a specific driver? Or maybe this is due to some kind of interrupts issue?

Any tips to diagnose or resolve this is appreciated. Sorry if this is a newbie lame question.

P.S. from this topic I came across this site which looks promising for via video drivers http://www.openchrome.org
This looks like an interesting read also ("How all driver components fit together").
Last edited by mskuma on Fri 25 Nov 2005, 06:38, edited 1 time in total.

User avatar
aahhaaa
Posts: 341
Joined: Fri 07 Oct 2005, 03:21
Location: Lower Michigan, North America

#2 Post by aahhaaa »

Hi- my Epia is only an M10000, so don't have much direct experience with the probs you are having, but...
there are so many alternative settings in the VIA BIOS, I'd look there first, especially for anything sound related. Are you doing 2 channel or 5.1? Have you changed the default encryption settings? How's the AGP set? How do you have the P&P set? The OS should parallel the BIOS settings, Lotsa things seem to change the priority that the Epia mobo does things.

If you are sure its an OS thing, I'd try another distro and see if the lag is the same. But I'd guess its not the driver, its the 'think about it' prior to activating the driver.

I'm glad to hear from another Epia user, because they are great boards, but in the US have few users that aren't doing carputers or somesuch. The guru center for them seems to be the UK, we don't even have the nano N boards yet.

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#3 Post by mskuma »

Thanks for your reply. You'd think it would be BIOS independent since Win98 is fine, which is what made me think of the driver issue. But I'll try your suggestion - thanks. I'm about to try epiOS to see if it's any different since they use the unichrome driver.

Regards

Michael.

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#4 Post by mskuma »

My quick check shows that epiOS produces a better flash experience (ala my issue reported initially), suggesting drivers are a key issue. I probably have to investigate how to compile in drivers.. this is new territory for me..

User avatar
dvw86
Posts: 636
Joined: Thu 05 May 2005, 00:55
Location: Washington State

#5 Post by dvw86 »

I use a VIA ML8000 and my experience is that Puppy uses the CPU for all of the graphics processing instead of the GPU. I think a different X server may be the answer. I haven't had any luck with MU's Xorg though. My VIA refuses to boot into Puppy after installing it.

User avatar
MU
Posts: 13649
Joined: Wed 24 Aug 2005, 16:52
Location: Karlsruhe, Germany
Contact:

#6 Post by MU »

There also is a via-driver for the kdrive-Xserver:
http://www.murga.org/%7Epuppy/viewtopic.php?t=4082

But I could not try that myself, as I have a SIS-card.

Mark

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#7 Post by mskuma »

Thanks dvw86 & MU for your replies.. hmm different X server? That's interesting.. I'd like to try your suggestion and report. Thank you again.

tempestuous
Posts: 5464
Joined: Fri 10 Jun 2005, 05:12
Location: Australia

#8 Post by tempestuous »

I'm thinking of getting one of these Epia systems, too.
From a little Googling it appears that VESA works, unaccelerated.

Accelerated XF86 drivers are available from the long-standing Unichrome Project, and more recently from Via's own "official" driver package.
From what I can tell, the main driver in both packages is similar or identical, but the packages are distinctly different in implementing MPEG2/DVD hardware support.

Which of these MPEG2 libraries you choose to use will affect how applications like MPlayer, Xine, MythTV, etc. need to be compiled and configured.

I think that recent versions of XFree86 / X.Org include the Via driver, via_drv.o, so perhaps MU's xorg682c.pup package has this driver and accelerated performance can be achieved without any extra bits & pieces necessary ... though without MPEG2 support of course.
There will probably be some tweaking of /etc/X11/xorg.conf necessary. This may help - www.epialinux.org/files/epia_howto/ar01s06.html

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#9 Post by mskuma »

tempestuous, thanks alot for your help again!

MPEG2 acceleration is not required in this instance, so sounds like if the driver has basic 2D graphics acceleration, it will be better already (compared to VESA).

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#10 Post by mskuma »

MU wrote:There also is a via-driver for the kdrive-Xserver
Hi Mark

I am finally getting back to investigating this.. seems like there are 2 approaches here to getting Via support as indicated by you & tempestous. Today I just tried the above approach, and found that using Xvia (with kdrive) on my SP13000 caused a bail out to busybox... maybe this is the same issue dvw86 saw? So I guess that Xvia does not support the later graphics chipset (CN400/UniChrome Pro). I'll probably give it a go on the ML8000A board (non-pro chipset) later. I gather the "xorg682c.pup" referred to by tempestous is different?

If the above doesn't pan out, if anyone is interested to provide some driver code, I'd be glad to test & provide feedback - hopefully we can fold some Via support into Puppy.

BTW I'd be glad to hear of any benchmarking tool for testing graphics performance that anyone happens to know about.

Regards

Michael.

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#11 Post by mskuma »

A minor update & an observation.. I tried the xchips.pup & found there was no bail out to busybox. So at least I could try something different. I've been using a flash benchmarking content, and I can't see any animation performance difference between xvesa (puppy fdo) & vchips (kdrive), perhaps not surprisingly, on my SP13000.

BTW when viewing some flash animation with synchronised audio, I have noticed the lag beween animation & audio cue gets worse over time. That is, when I first run the content (immediately after loading it), the animation & audio cue synchronisation is reasonably ok (not much lag) but using the same content over time, the sync delay gets worse.. this isn't happening on Win98 - maybe this is a flash player issue on linux - or an audio driver issue? Any clues? Anyway I'll try the other suggestion from tempestous.
Last edited by mskuma on Fri 09 Dec 2005, 21:18, edited 1 time in total.

mskuma
Posts: 25
Joined: Sun 13 Nov 2005, 12:39
Location: Australia

#12 Post by mskuma »

I tried xorg682c.pup with the via driver setting & found the same issue - bail out at busybox. I tried it on my other board (ML8000A) & found xorg682c.pup with the via driver setting works, so obviously the via driver is not compatible with the Unichrome Pro driver. Incidentally for the SP13000, I tried other drivers that I thought might be candidates (S3, S3virge, savage) but no go on these either. I've also noticed that Xorg does not always exit (reported previously) properly.

Anyway using xorg with via driver, I still see the sync lag issue, so I am not sure if it's a video driver issue anymore.. I'm kinda puzzled why simply changing to Win98 makes this all go away.. except for the obvious - MM optimised for Windows and not for the Linux plugin or maybe Windows deals with plugins better (from resource POV) - dunno..

So could there be some audio driver issue?

tempestuous
Posts: 5464
Joined: Fri 10 Jun 2005, 05:12
Location: Australia

#13 Post by tempestuous »

I'm keeping my eye on developments for Epia graphic support, too. It seems the Via driver in XOrg 6.8.2 is not fully accelerated - it needs some other support modules. I just read on the viaarena forums that opensource developments (which are now split into 2 forks) make their way back into the XOrg source, but hardware acceleration is only right in the XOrg 6.9.x betas.
You may need to wait for the next stable XOrg ... or compile XOrg 6.9.x ... or compile the bleeding edge version from http://openchrome.org/
The original opensource project, http://unichrome.sourceforge.net/ has become conservative.

Having said all this ... there's no guarantee that an accelerated X server will solve your Flash problem.

Post Reply