Slow/stuttering playback on "decent" system

[velderia]

I came here wondering if people are having the same problem as me and just... wow. o_o

I hate mkv files like a crying baby jesus hates the devil.

I don't know much about this... Encoding/rendering stuff y'all speak of and it's making my head hurt.

Right now, I realized pausing for longer than 5 minutes, then playing again, helps to play it more smoothly. Seems like a temp fix for now.

[arenine]

Originally Posted by velderia

I came here wondering if people are having the same problem as me and just... wow. o_o

I hate mkv files like a crying baby jesus hates the devil.

I don't know much about this... Encoding/rendering stuff y'all speak of and it's making my head hurt.

Right now, I realized pausing for longer than 5 minutes, then playing again, helps to play it more smoothly. Seems like a temp fix for now.

If you use Media Player Classic, you can just press Ctrl + Left Arrow and it will jump back a few second and resynchronize. How far it jumps back can be adjusted in the settings.

[npcomplete]

Originally Posted by Tran225

i really want my computer to last me for another good 5 years, since u say i have good enough GHZ (3 GHZ) do u think a new video card would help? i dont plan on doing any intense gaming, mostly anime thats why i wanna stick with this computer

short answer: no
long answer: maybe, depends on card, software, encode

yes have tried Nire's Spice and Wolf II file size was approx 348k-ish .mkv about the size of gg's bakemonogatari also and it had audio desync with picture too

When it desyncs, does the video stutter or lag? Or does the video play ok and the audio just seems to eventually drift?

My blanket suggestion right now would just be to:
- install latest Nvidia driver
- uninstall old CCCP
- install latest beta CCCP (as of today is 2009-08-24)
http://www.cccp-project.net/beta/
- re-register filters and reset settings in the CCCP settings tool
- use MCP-HC that comes with CCCP
- in MPC-HC -> Options -> Playback/Output -> Overlay Mixer

[SeijiSensei]

Originally Posted by Tran225

can someone let me know if my video card is good enough for HIGH QUALITY mkv animes? if not ima just get another one for my 7 year old computer @ black friday sale

A new nVidia card might do the trick. Read some of the links above about CUDA, if you're using Windows, or google for VDPAU if you're using Linux.

[Tran225]

Originally Posted by npcomplete

short answer: no
long answer: maybe, depends on card, software, encode


When it desyncs, does the video stutter or lag? Or does the video play ok and the audio just seems to eventually drift?

My blanket suggestion right now would just be to:
- install latest Nvidia driver
- uninstall old CCCP
- install latest beta CCCP (as of today is 2009-08-24)
http://www.cccp-project.net/beta/
- re-register filters and reset settings in the CCCP settings tool
- use MCP-HC that comes with CCCP
- in MPC-HC -> Options -> Playback/Output -> Overlay Mixer

there is not lag spikes or anything as i am watching it
it eventually drifts away like u said
i recently reinstall into a Beta CCC, Panyone know wat my setting for ffdshow audio setting should be if i am using surround sounds (logictech g51 - http://www.kustompcs.co.uk/acatalog/4138.jpg?

[npcomplete]

I think it should be:
- "Mixer" enabled
- "3/0/2 - 5 channels"
- "LFE" enabled

Have you tried setting it to:
- "Mixer" enabled
- "2/0/0 - stereo"
- "LFE" disabled

and seeing if the audio still drifts?

[Vagrant0]

Originally Posted by velderia

I came here wondering if people are having the same problem as me and just... wow. o_o

I hate mkv files like a crying baby jesus hates the devil.

I don't know much about this... Encoding/rendering stuff y'all speak of and it's making my head hurt.

Right now, I realized pausing for longer than 5 minutes, then playing again, helps to play it more smoothly. Seems like a temp fix for now.

I ended up asking specifically how to re-encode to a lesser format. After a bit of debate and rehashing information already here, a simple, concise, straight forward answer was given.

Originally Posted by TheFluff

Let us review:


Then, here is the retard's guide to using ffmpeg:
1) http://lmgtfy.com/?q=ffmpeg+windows+builds&l=1
2) click the big obvious "ffmpeg-latest" download link
3) make a folder called ffmpeg somewhere, preferably directly under your C:/ drive, and unzip the contents of the package you just downloaded to it
4) go to the start menu, hit run, type in cmd, press enter
5) navigate to the folder where you have the animu you want to transcode (*)
6) type in

"C:/ffmpeg/ffmpeg.exe" -i "animu episode 01.mkv" -vcodec mpeg4 -qscale 2 -acodec libmp3lame -scodec copy -ab 128k "transcoded animu episode 01.mkv"

and press enter (you obviously need to replace the filenames and path to ffmpeg with the ones that apply to you first)
7) now you get to wait; when it's done you'll have a new mkv that is exactly like the old one except it has 128kbps CBR mp3 audio and an xvid-like video.
8) THE END

(*) If you are unfamiliar with basic commandline usage, you do this by saying

X:
cd "my animu folder/seriesname"

or somesuch, where X: is the drive letter, and the stuff after the cd is the folder name you want to go to. cd .. takes you up one level, and cd \ takes you to the root of the current drive. Google or use the windows helpfiles for more info.


PROTIPS:

• As dark shikari said, you can add -s 852x480 or another resolution of your choice before the -ab option if you want to downscale to a smaller resolution (or upscale, but that'd be really pointless).
• If you want higher audio quality you can either change -ab 128k to -ab 192k, or just change -acodec libmp3lame to -acodec copy and remove the -ab 128k option.
• If you're too lazy to type out "C:/ffmpeg/ffmpeg.exe" every time you can add the ffmpeg folder to your PATH environment variable and just say "ffmpeg" instead; you do this by right-clicking my computer -> properties -> advanced -> environment variables -> edit the PATH one and add ;C:\ffmpeg at the end (the semicolon is important).
• If you want to transcode an entire series at once you can make a batch file that does it automagically for you (google windows batch file, or search the windows help file).

I ended up just making a shortcut in the folder where ffmpeg is, changing the target to have all the flags and such, and just dropping the episode into the folder, re-naming it, running ffmpeg, and then renaming the output back to what it should be once done. So far it's been working well enough, and doesn't require anything higher than a 6th grade edumacation.

Tried that pausing stuff as well, waiting 5 minutes for 2 minutes of sync didn't seem worthwhile in my case. Re-encoding was the only real option for me.

[Miles Teg]

Playing around with the different computer at home

mplayer -vo null -nosound -benchmark [filename]

This command line deactive the sound, don't show the video (post-processing is done), ASS activated in the config file.

File : [SS]_Kemono_no_Souja_Erin_-_22_(1280x720_h264)_[5805E6C3].mkv

Time needed for the processor to decompress the video as fast as possible.

All time in seconds (the lower the better).

[npcomplete]

Thanks for the benchmarks!

This is with your E6750 under clocked, right? Is the X2 the older K8 based one?

Also, is it possible to build mplayer (ffmpeg) without SSEx? I'm curious to see how much that makes a difference.

[Miles Teg]

Yes I have underclocked the E6750, the X2 is a "Brisbane" from the 65nm SOI line so one of the latest K8.

Looks like it's possible to deactive all the SSE optimization so I will try that (but only with the E6750@2.67Ghz )

EDIT : With MMX/SSE/SSE2 270s Without 425s

[npcomplete]

Originally Posted by Miles Teg

Yes I have underclocked the E6750, the X2 is a "Brisbane" from the 65nm SOI line so one of the latest K8.

Looks like it's possible to deactive all the SSE optimization so I will try that (but only with the E6750@2.67Ghz )

EDIT : With MMX/SSE/SSE2 270s Without 425s

Interesting. I was curious because I've read that the K8 see very little to no improvement with SIMD instr, while the Intel cpus see a large improvement. The newer K10s see some but are still not as strong with SIMD as Intel.

[Miles Teg]

Originally Posted by npcomplete

Interesting. I was curious because I've read that the K8 see very little to no improvement with SIMD instr, while the Intel cpus see a large improvement.

You have make me curious too

And the results are

C2D With MMX/SSE/SSE2 270s Without 425s
A64X2 With MMX/SSE/SSE2 408s Without 637s

Both CPU are 1.6 times slower

[velderia]

Originally Posted by arenine

If you use Media Player Classic, you can just press Ctrl + Left Arrow and it will jump back a few second and resynchronize. How far it jumps back can be adjusted in the settings.

Oh thanks, that actually helped quite a lot.

[npcomplete]

Originally Posted by Miles Teg

You have make me curious too

And the results are

C2D With MMX/SSE/SSE2 270s Without 425s
A64X2 With MMX/SSE/SSE2 408s Without 637s

Both CPU are 1.6 times slower

Hah! I guess not then

I got that impression about K8 from info such as:
http://www.mersenneforum.org/showpos...40&postcount=9
http://www.mersenneforum.org/showpos...1&postcount=11

Some slides on optimization for opteron even say, that for double precision one should use x87 since peak throughput are nearly the same as for SSE2 with the plus of having more instructions (log2, sin, cos ...).

Athlon+AMD64 can do FADD+FMUL+FLD/FST per cycle or SSE2 (AMD64 only) with 1 FMUL and 1 FADD. Unfortunately an ADDPD can't be run in parallel to a MULPD because each of them needs 2 decoder ports of the 3. But an ADDPS could be issued since it needs one port. (2 ports are needed to issue double FP operations for each half).

I don't know how much of that applies to decoding h.264, which uses integer right?

http://www.mersenneforum.org/showpos...03&postcount=9
http://mersenneforum.org/showthread....5&pagenumber=7

and some bechmarks running prime95 from:
http://www.mersenneforum.org/showthread.php?t=475
http://www.mersenneforum.org/showthread.php?t=1947

- Pentium 4 @ 2.3Ghz, SSE2 ON = 108.204ms
- Pentium 4 @ 2.3Ghz, SSE2 OFF = 298.231 ms
- Athlon XP (K7) @ 2.2Ghz, NO SSE2 = 153.008 ms
- Athlon 64 (K8) @ 2Ghz, SSE2 ON = 159.141 ms

[Dark Shikari]

The [video] decoders don't use floating point math whatsoever.

[chikorita157]

Well, I got a Mac Mini G4 off ebay recently (1.33 GHz, Late 2005 model)... Surprisingly enough, it lags when you play H264 on it, even in mplayer. Apple said that it should be able to support H264 video with a resolution of 852x480 at 24 frames per second.

For 852x480 (480p) video at 24 frames per second:
QuickTime 7 for Mac OS X:

* 1.25 GHz PowerMac G4 or faster Macintosh computer
* At least 128MB of RAM
* 64MB or greater video card

I know the performance of the PowerPC G4 is supposed to be the same as a Pentium 4 non-HT at 1.8+ GHz. Maybe it's because of the graphics card which is only 32 MB or H264 being too intensive for a 4 year old computer, which is hard to believe. Aside from that, XVid/DivX video works as expected (it also worked on PowerPC G3 Macs without any hiccups).

This pretty much makes you wonder why Apple made the switch to Intel...

If you want to know the full specs, it's a Mac Mini (Late 2005) with a PowerPC G4 processor at 1.33 GHz, 512 MB of RAM and a ATi Radeon 9200 32 MB graphics card.

Edit: 852x480 resolution or lower Fansubs work fine with the G4, but anything higher causes choppy playback.

[Miles Teg]

The PowerPC G4 don't have SSE unit, and I don't think the H264 decoder is optimized for Altivec, so you probably have one part of the problem.

A pentium 4 at 1.8Ghz will have problem with H264, 720p video so if the G4 is equivalent to this processor then it probably simply lack horsepower

Mplayer has some VO for ATI card, you should look for that, that could help.

[chikorita157]

I highly doubt that the PowerPC G5 would have a problem with H264 since the G5 is a lot more powerful than the G4, at the same time produces alot of heat.

I really don't like the Pentium 4 since it's a very inefficient processor and produces alot of heat (mainly because Intel is more focused on making the clock speed higher than making a efficient and less power hungry processor.) I have a Inspiron 9100 which doesn't really work anymore and it had problems with overheating because the vents seems to always get blocked by a thick layer of dust and it used the Prescott processor, which is the most inefficient processor in the Pentium 4 family, even if it have SSE3. I would rather use a AMD or PPC processor than the Pentium 4 Prescott. Also, the battery life sucked on the Inspiron 9100 because of the power hungry processor and GPU (it only lated 1 hour and 30 minutes) and you couldn't plug it into a power inverter since it uses too much power.

Remember, megahertz doesn't mean everything... there are other factors that effect performance. Thank god Intel kicked the habit of having more megahertz/gigahertz with their Core processors.

PowerPC G4 will play H264, but no HD resolution or 768p video... and it will work best if you use Quartz/Quicktime in Mplayer OS X Extended so you can offload some of the processing using Quartz Extreme via GPU, it brings down the CPU usage significantly. I'm pretty sure the 2.0 GHz PowerPC G4 upgraded PowerMacs can play 720p with no problem since they have even faster processors.

[Dark Shikari]

Originally Posted by Miles Teg

The PowerPC G4 don't have SSE unit, and I don't think the H264 decoder is optimized for Altivec

if (has_altivec()) {
        c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec;
        c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
        c->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_altivec;
        c->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_altivec;
        c->h264_idct_add = ff_h264_idct_add_altivec;
        c->h264_idct_add8 = ff_h264_idct_add8_altivec;
        c->h264_idct_add16 = ff_h264_idct_add16_altivec;
        c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
        c->h264_idct_dc_add= h264_idct_dc_add_altivec;
        c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
        c->h264_idct8_add = ff_h264_idct8_add_altivec;
        c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
        c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
        c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;

#define dspfunc(PFX, IDX, NUM) \
        c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
        c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
        c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
        c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
        c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
        c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
        c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
        c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
        c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
        c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
        c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
        c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
        c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
        c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
        c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
        c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec

        dspfunc(put_h264_qpel, 0, 16);
        dspfunc(avg_h264_qpel, 0, 16);
#undef dspfunc

        c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels16x16_altivec;
        c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels16x8_altivec;
        c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels8x16_altivec;
        c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels8x8_altivec;
        c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels8x4_altivec;
        c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels16x16_altivec;
        c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels16x8_altivec;
        c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels8x16_altivec;
        c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels8x8_altivec;
        c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels8x4_altivec;
    }

Hmm, not optimized at all...

[Miles Teg]

New computer at home so some new test

The computer is composed of :
ASUSTeK Pundit P4-P5N9300 (NVIDIA GeForce 9300).
Pentium Dual Core E5300 (2.6Ghz).
2Gb of RAM (256Mb for the video card)

First the chipset support VDPAU I have tried some 720p files and no problems.
Second the processor is decoding H264 as fast as my C2D E6750 (2.67Ghz) looks like the L2 cache and the FSB don't play a big role for decoding video.

All the test done on an Ubuntu 9.10 (x86-64).