Encoding speed

Hi there, Arch (x86_64) has been my home for a while now. Satisfied as I am, I regularly try out other distro's. I've just tried out Xubuntu 7.10 & was quite surprised when ripping a DVD with mencoder, it was almost 3x faster than Arch.
Any idea how this could be so much quicker? I've made custom builds of mplayer, lame & xvidcore for Arch, resulting in absolutely no improvement. It'd be nice to get this performance where I'm more comfortable, after all, Arch kicks 455 at everything else.
Screenshots showing issued command, time, psnr, size, etc:
http://xs.to/xs.php?h=xs220&d=07443&f=arch.jpg
http://xs.to/xs.php?h=xs220&d=07443&f=xubuntu.jpg

retsaw wrote:Are you using the same version of mplayer on Arch and Xubuntu? Do you get the same encoding speed difference when encoding to different codecs?.
Both Arch & ubuntu use the same version of mplayer (0.9rc1), though Ubuntu splits mencoder into a separate package:
http://packages.ubuntu.com/gutsy/graphics/mplayer
http://packages.ubuntu.com/gutsy/graphics/mencoder
I hadn't tried other codecs, so here's the command & time results using lavc:
Mencoder options:
time mencoder dvd://1 -chapter 2-2 \
-oac mp3lame -lameopts abr:br=128 \
-ovc lavc -lavcopts vbitrate=2880:mbd=2:trell:v4mv \
-vop scale -zoom -xy 720 -o wotw_5.avi
Arch:
real 2m42.298s
user 2m37.750s
sys 0m0.837s
Xubuntu:
real 2m48.272s
user 2m43.518s
sys 0m0.612s
That's more like it. I guess something's not quite right with xvid. As xvid works (only slowly in comparison to Xubuntu), what can I/we do to improve it?
PS, skottish, 'enc' is just where I've kept my mencoder options for these tests, I know the only difference was the file name, it's just habit. I only included it in the screenshot so you could see the issued command & that they were the same. The list's now got quite a bit larger, with a little variation: http://pastebin.archlinux.org/17045

Similar Messages

Premiere Pro CS5.5 encoding speed

Hi,
I noticed that the time for encoding a video in Premiere with GPU acceleration enabled doesn't change according to the shaders clock speed. The time is exactly the same with or without GPU overclocking, is this normal?
Another question: the encoding speed is faster during the first 50% of the process, then decreases a bit. Is this related to some option in the program or it's normal?
Thanks in advance.
Luis

I've made an encoding test where we can see that GPU participates more actively in the first half of the encoding process and therefore I don't understand why you said "The GPU is not involved in encoding, that is a purely CPU matter". From 50% until the end of the process CPU takes the lead.
Luis

Encoder Speed with NI 9215 module and cDAQ-9172

I've a NI cDAQ-9172 with the NI9215 analog input module, and I want to estimate the speed with an optical encoder (OMRON E6A2-CW3C). My approach was, making a comparison with a point in the midle of the high of the signal, ie 4V, making this way a perfect square signal with 1of high. After that I use the Tone Measurements and the Timing and Transition Measurements modules to calculate the frequency of the wave form, and this way I'm able to calculate the RPM speed.
The problem with this solution, is that even with the engine at a constant speed of 746RPM, measured with a stroboscope, Labview give me a high variable values between 730RPM to 760RPM, and I can't solve this problem because I really need an accurate measurement.
Should I buy for example the NI 9401 module to make this measurement? If so, does any one knows how to build a VI to get the encoder speed?
In attach I send the VI that I'm using to calculate the speed.
Thank you for help.
Attachments:
Encoder Speed.vi ‏112 KB

Hi,
The best way to measure a speed through a quadrature encoder is to use the counters that are inside the cDAQ chassis.
With the 9172 chassis, you need a digital I/O module (9401) in order to access the counters.
With the new generation of chassis (9178), you don't need the 9401 because you can connect your encoder signals to the BNC.
Best regards,
Thomas B. | CLAD
National Instruments France
#adMrkt{text-align: center;font-size:11px; font-weight: bold;} #adMrkt a {text-decoration: none;} #adMrkt a:hover{font-size: 9px;} #adMrkt a span{display: none;} #adMrkt a:hover span{display: block;}
>> Inscrivez-vous gratuitement aux Journées Techniques : de l'acquisition de données au contrôle/com...

Encoder speed and orientation on NI 9401 in cDAQ-9172

Hello, I've acquired a ni 9401 module for my cDAQ-9172 and it's now installed in slot five to connect a dual channel encoder. Channel A is connected to pin 14 and channel B to pin 17 and the encoder is powered by 5v dc.
I want to mesure the speed and if possible the orientation of rotation.
To do this I started with the VI in attach, that can count the total of pulses in the encoder.
My question is, how can I change the DAQmx Create Channel to Frequency, and the DAQmx Read to Counter Pulse Freq 1 Chan 1 Samp, to calculate the actual encoder speed and orientation?
Solved!
Go to Solution.
Attachments:
encoder.vi ‏63 KB

Hi Eric.
Once again, thank you very much for the help.
I think I finally did it. The Vi is reading the speed witch I think is right, but there's some oscilation.
I've connected only the signal A from encoder to pin 14 on the NI-9401, and the pin 1 to ground. The signal B is not connected right now because I only have one NI-9401 on a cDAQ-9172, so only one counter.
The VI you just send me, is calculating the speed (frequency), witch I divid by 200 because those are the pulses per revolution of the encoder, but I can see a little of oscillation on the speed measured (see picture in attachment).
For now, my question is if is possible to improve the measurement, and how?
Thank you.
André Quintino.
P.S. - Note that in the picture in attachment, the speed is the one measured by the VI you just send me. The encoder was running at 1524RPM witch are 25,4Hz. Since the speed in the graph is multiplied by 200 (the pulses per revolution of the encoder), this result in a measurement of 5080 witch is the average number by the graph.
Attachments:
Screen shot 2011-01-12 at 00.41.13.png ‏16 KB

What factors affect encoding speed the most?

Hello everyone,
I am actually quite happy with the encoding speed to DVD formats. I was under the impression that encoding was "real-time". Encoding a 1 hour long PAL 48kHz capture from VHS to DVD only took ~30m.
I opened the task manager and noticed that both cores of my Athlon X2 5400 were running flat out during this process. Thus I take it encoding can utilise multiple cores.
My questions are:
If I upgrade to a Quad core CPU what performance increase would I see with the encoding example above? 15m?
What influence does hyperthreading have -if any?
Do you expect that PE8 will support nVidia CUDA at some stage? I notice that the latest Roxio software does.
Will the encoding slow down dramatically when I start working with HD content?
Do any of the Matrox hardware encoders work with PE8? They only seem to support Premiere CS4
I have SATA-II RAID 0, 2GB RAM, AMD Athlon X2 5400+ and XP SP3
Justin.

Export and Rendering speed can also be affected by one's I/O sub-system, i.e. your HDD's (Hard Disk Drives), their speed, amount of free space, the controller type and how you have these HDD's allocated.
With the exception of AVCHD, which Steve mentions, having separate physical HDD's (not partitions - logical drives), will speed things up greatly, as all operations on the computer, access to the OS, the program, the media files and then outputting from the program, will be spread over several HDD's. If you only have one physical HDD, everything has to flow through one "pipeline," and this adds significant time to the processes. Having partitions will slow things down even more, and can even cause hangs, because the OS "thinks" that it has multiple HDD's to use, but only has one. It cannot understand what's taking so long, and grows impatient.
With AVCHD, the CPU is used much more than even the I/O sub-system, and becomes the most important link in the chain.
Good luck, and let us know all about your HDD's, their number, size, speed, controller type, say SATA II, and how you have these allocated.
Hunt

Very Slow Encoding Speeds Ripping CD's & DVD's

I have recently purchased a new Mac mini and I have noticed that encoding speeds when I am using iTunes - Lame & Handbrake are incredibly slow.
When I am ripping a CD with 256 VBR it is taking ten minutes per song!
I have a 2 year old MacBook Pro Laptop and it takes generally 5-6 minutes for a whole CD
Is anyone else experiencing the same issues or do I have a faulty drive?

I would have to render a guess of around 100 degrees. and it didn't seem to affect the burn. I put each movey in the home player and bounced around the tracks, but I wasn't going to sit through 4+ hours of movie watching in one night. I just noticed you have the 17" model, maybe due to the tighter confines of the chassis there may be more heat in the area of your drive, but most CDs/DVDs are pretty resilient. I'm sure if you have a CD player in your car, you can attest that those things can get smoking hot in the right conditions. just be careful also, that if you have a fairly cool room, that you don't place a hot disc on cold surface, cause that could thermaly shock the disc, and render it useless. a couple of years ago we put an iron over a cd and then put it in the freezer. we had to completely defrost it to get all the fragments out. a cold table probably wouldn't shatter it, but it would warp the disc and there isn't anything you can do about that.

Encoding Speed with Media Encoder Issue?

I have a 5 hour and 36 minute timeline in Premiere Pro CS5 which is composed of mixed assets- some at 29.97, some at 30 FPS, etc. The timeline is exported to Media Encoder and set for Match Source Attributes Highest Quality (Quality 5.0, VBR 2 pass, min 3.00, target 6.00, Max 7.00 MBPS). NTSC 720 X 480 (0.9091), 29.97 FPS, PCM 48hz, 16 BIT stereo. Format is MPEG2 DVD.
Estimated file size is 18138 MB. The file takes approximately 36 hours to encode. Is this normal?
Premiere Pro CS5
Media Encoder CS5
Windows 7 64 BIT
2.4 GHZ
Intel Quad 4
700 GB free space on hard drive
6 GB RAM

Is this normal?
Probably yes. Aside from the normal conversion with 2 pass VBR, which even under ideal conditions could require up to 3 times the duration of the footage (in your case this makes about 15 hours), you are causing additional operations to conform the frame rates on the Premiere level of things. Also, since this is handled as a Dynamic Link process, you may be a bit short on RAM, which may slow down things further. And then of course the usual with disk speed and file I/O etc.. So it more or less sounds normal, but you should not let the initial rough guesstimates discourage you. Many times these values will get much lower after a while of transcoding when the encoder has a much clearer picture and statistical data about how long it actualyl takes to encode so and so many frames...
Mylenium

Encoding speed limit in Adobe Media Encoder?

Dear All,
I experienced a strange (for me) behaviour of AME. I made an encoding test with different coding profiles used the same 1GB big MOV file (from 5D Mark II) on two PC-s, and got a very similar results regarding encoding duration. This is impossible. The difference was only 4-6 secs, see below:
PC-1:
-          CPU Q9550 (Quad Core 2.8Ghz),
-          6GB /1333 RAM,
-          Vista 64Bit,
-          GT9600TOP,
-          latest update of Media Encoder.
-          Encoding Time: 16:40
PC-2:
-          CPU I7 - 920,
-          6GB /1600C8 RAM,
-          Vista 64Bit,
-          GT9600TOP,
-          latest update of Media Encoder.
      -          Encoding Time: 16:40
The source file was opened directly from AME. Both of them used only 25-35% of CPU. I enclosed two screenshots about it. Any idea? I repeated these tests with DVD and Windows Media 9 profiles, and got similar results.
ps. I have read problems regarding Canon 5DMII videos with PR4. I did not experience any problems expect the preview speed. With I7, it is almost acceptable, but non of slower (tested Q9550, E6400) CPUs gave acceptable playing quality.

- with HDV to DVD, I can reach 100% CPU, but becuse os resizing the frames. If you convert it to Widows Media, or MPEG2-BlueRay, it will not use 100% CPU.
- I have 3 Disks:
a., Vista and programs, SATA2 750GB Samsung with two partitions, second one is for the final DVD / BL folders.
b., "Sources", 1TB SATA2 Samsung 32MB Cache
c. "Projects" and transcoded files, SATA2 750GB Samsung.
If you check the screenshot I sent first, you can see that there was no bi traffic on the disks.

Compressor encoded speed change clip with echo as motorboat sound

Took me two days to troubleshoot, through trial and error and process of elimination, why Compressor encoded an audio clip that had been speed changed and echo applied to it -- as rapid MOTORBOAT sound! Project is a 5.1 mix. Since STP3 won’t accept speed change clips, after I completed mix in STP I placed speed changed clips on FCP timeline below the exported STP 5.1 mix with the FCP audio tracks for that clip assigned to channels 1 & 2 for the fronts and channels 5 & 6 for the rears. The clip on 5 & 6 was duplicated and identical in every way to the clip on channels 1 & 2. Only the dupe clip on 5 & 6 did Compressor encode as motorboat. The identical clip on 1 & 2 encoded correctly. Both clips were set to play at same time since I wanted the effect to be on fronts and rears simultaneously. I could not get the motorboat sound to play back from either FCP timeline or Compressor preview window. Only after encode and burn to DVD did it show up. Why would Compressor introduce such artifacts on one of the clips and not both – if it’s going to do it at all – which it shouldn’t? Any ideas? I have other such effects throughout the project where I duped a speed changed clip with an effect applied to it and set it to play on fronts and rears simultaneously and Compressor encoded those correctly.

Well, it's a little late, but this is one of the reasons to have your startup drive cloned so you can restore it to a working state if an update, upgrade or install sends things south.
https://discussions.apple.com/docs/DOC-2494
Have you tried resetting your fcp preferences
https://discussions.apple.com/docs/DOC-2491
If this doesn't fix the problem, you could try sending the shot to motion. You could aslo try changing the speed in the motion tab of the viewer.

Can Drive Type Affect Render and Encoding Speed?

The title pretty much says it all.
A client gave me a G-Drive (FAT32) with the video files on it. I'm using a PC (NTFS) with the G-Drive connected via eSTAT.
It seems to be taking a very long time (nearly four hours) to encode a 48-minute video that has no color corrections, no FX, no nuthun. I'm encoding to MP4 360p.
Is this normal? Is the G-Drive part of the slowdown?

Jay,
An eSATA drive is equally fast as an internal SATA drive, but just as with internal disks, there are large differences in drive performance, for instance the latest Seagate Barracuda ST2000DM disks achieve transfer rates of 150 - 190 MB/s, but older generation disks may achieve only 80 MB/s. So the question is what disk is mounted in that G-drive, that can tell a lot about its speed.
The second thing is that currently it is formatted as FAT32. While this is certainly not as good as NTFS, you may have no choice but to leave it as it is. If that external drive is sometimes attached to a MAC it must be FAT32 formatted, because MAC does not support NTFS without going through serious hoops.
Third, the export size is dependent only on duration and bitrate. Framesize is irrelevant. Size = (number of frames) x (bitrate per frame).
Last, the Q6600 is a pretty slow CPU today. Most of the Q6600 systems in the Benchmark Results are around 10 - 20 times slower than a fast system. If your system takes around 240 minutes to encode a 48 minute timeline, imagine a 10 times faster system requires only 24 minutes. That is two times faster than real time.
I'm not surprised by your results.

Improve encoding speed for WMV output?

Hi,
Firstly, apologies if I am posting this in the incorrect area of the forums. I'm not entirely sure it's a hardware issue, I think it might be codec related.
I am trying to find a way to reduce encoding times for a (on average) 4 minute clip using Premiere/AME CS5
My input format is either XDCAM HD or IMX50, my output formate is always wmv, 720x 576 (PAL) interlaced, currently single pass encoding (because of the time taken) using the Windows Media Encoder (wmvencod.dll).
There are no effects, nor transistions in my sequence. 2 tracks of video - 1 actual video, the second a logo overlay. 1 stereo pair for the audio.
Hardware is a single Intel Xeon, quad core 2.5Ghz processor, currently with 10Gb memory. I have tried doubling up on the processors and also increasing the memory up to 16Gb but the improvements are marginal. Switching hyperthreading off also provides a marginal improvement in encoding time.
I am running this on a server class 1u box. Currently it only has the standard vga server graphics card in it. I have mixed feelings about installing a graphics card as I have read conflicting information about whether it will provide an improvement in my encoding time.
I don't think MPE is relevant, given that I'm not doing any effects processing.Will the CUDA architecture provide a significant improvement?
I can only find 2 graphics cards that I think will fit in the box and these are the Quadro 4000 and Quadro FX 3800 that retail at around £650 - £800 + so it's a significant punt if they aren't going to provide an improvement.
In essence, I have a template (well a couple of) project that I add the source file to each time that gets dropped into AME for encoding. For HD source encoding takes about 4 times real time. For SD it's about real time.
If I chose a different output format, with the same input format eg h264 or DV (avi) then the encoding time is reduced to roughly 1/3rd of what I am experiencing at the moment. That gives me the impression that the wmv encoder is the bottle neck
Are there alternatives to using the standard Microsoft WMV encoder? If so, how do I go about getting AME to use them, given that prior to installing Media Player I didn't even have wmv as an export option.
Any thoughts on this would be appreciated.

I couldn't tell you whether or not it uses the same dlls as the Microsoft program... that's getting deeper under the hood than I usually go! :-) If it is, then you're correct in that going to an intermediate and using WME isn't going to speed up your process.
For more info on hardware, go to www.ppbm5.com and look at the test results page... see what folks were using who got the higher results. IIRC, your specific processor (even in its dual form) is a rather poor performer with this kind of stuff, and the Core i7 line is much better.
Out of my own curiosity, I'm running a test on my system to see if I can reproduce the similarly long encode times with WMV versus h.264. I've got a 4:23 AVCHD clip which I'm rendering first using h.264 MainConcept to PAL DV Widescreen HQ (the only reason I'm using PAL here isbecause that's what you're using, and I wondered if it would go differently than NTSC... I've never used PAL before). The full encoding time for that in h.264 was about 1:45, so a little better than double real time. The same clip, rendered with WMV using PAL Widescreen source to HQ Download is taking about 3:15, so pretty much double the h.264 clip's time, but still better than realtime, and nowhere near the results you're getting.
I don't remember what you said your HD setup was like, but make sure you've got multiple hard drives, preferably a RAID for your data disk. I learned that having a good disk configuration can really make a lot of difference, all other things being equal!

Kernel 2.6.32 will boost x264 encode speed BIG TIME for multicore CPUs

It's due to a tweaked scheduler in the 2.6.32 codebase. I measured a 53 % speed boost comparing 2.6.31-5 to 2.6.32-rc6 on my quad. See my post over at the handbrake forum here for details and see this page too.
The test encode was a 60 fps 720p (1280x720) encode using the "High Profile" on handbrake snapshot 2907 (which you can get in the AUR here):
Arch x86_64 (kernel 2.6.31.5): avg 15.99 fps
Arch x86_64 (kernel 2.6.32-rc6): avg 24.45 fps
Linux (kernel 2.6.31-bfs): avg 24.49 fps
Win XP 64-bit edition: avg 20.41 fps
Linux is 53 % faster w/ the new kernel and 20 % faster than Windows. Can't wait for 2.6.32 to go final!
Last edited by graysky (2009-12-06 10:57:58)

bangkok_manouel wrote:
MP2E wrote:Epic! I always encode things in x264(though through FFMPEG, though that shouldn't make a difference since it uses the x264 library, should it?) but I'm a user of BFS. Since you are referring to a scheduler change, will this affect me?
you'll need to go back to the mainline CPU sched (CFS)
But this bug was noticed in the first place because BFS kicked CFS's ass with x264, so sticking with BFS is fine
@tomd123: It's a general improvement in the scheduler, which should translate to other workloads as well. x264 got a huge boost, not everything will get such a huge increase in speed.

Encode speed

I was running a few experiments with iMovie encoding and was wondering whether my experience was 'typical'.
I was encoding 1 minute of PAL DV video to h.264 at full quality (25 fps, full resolution, any more important parameters I should mention?) and it took about 4 minute using the single pass encoding. It took around 9 minutes using the multi-pass encoding.
I have seen some people talking about real-time encoding and was wondering if iMovie managed this or if it was just something you got with Final Cut Pro or similar?
(Also, I noticed horrendous tearing in all motion which I presume are interlacing artifacts as a result of some known iMovie bug - is this where I need to use JES deinterlacer to fix this in some way? What exactly do I need to do? It kept giving me a "movie time error #-2015" whenever I tried to open the exported movie)

Seems that the Quicktime H.264 encoder is pretty slow. Found that the open source x264 encoder in Handbrake seems to work much faster.

WOW! Encoding speed up by 19hours! Thanks!

Thanks to Bill, Steve, and John.
My movie only took an hour to encode tonight (vs the 20 hours it was going to take a few days ago).
Always encode your movies to DV-AVIs!
^That should be rule number one!

Hey, speak for yourself Steve. I still seem to be so far behind the curve, that I feel hopeless...
Hunt

IDVD 5 vs iDVD 6: multiprocessors, encoding speed and quality

I complained in another thread that iDVD 6 was taking twice a long as iDVD 5 to encode DVDs. Most everyone who has upgraded from 5 reports slowdowns but the amount varies. To learn more I encoded the identical project (using iDVD 5 menus only) in both versions, both set to best quality. It contained 81 minutes of video, three menus 2 transitions and nothing else.
Video Encoding:
The most important thing I learned is that multiprocessor support has been reduced in iDVD 6. CPU usage during video encoding was:
iDVD 5: 240-270%
iDVD 6: 140-170%
So, iDVD 6 seemingly will not recognize more than 2 processors. This would explain why dual G4 and G5 users would see a slow-down but perhaps not as great a slow-down as Quad users. I certainly hope Apple will restore more complete multiprocessor support in a future update.
However - The final project in iDVD 6 ends up slightly larger than in iDVD 5. My son (a professional video editor) suggested that perhaps iDVD 6 is also encoding to a higher quality, accounting for some of the slowdown and larger size. Based on a purely subjective inspection of the same scenes (from close up) at full screen on a 23 inch monitor I think this is right. iDVD 6 encoding seems to leave fewer artifiacts, less pixillation and blurring and so on than iDVD 5. Depending on the source of your video and where you view your DVDs this difference may, or may not, be important (or even real, maybe I imagined it) and thus worth the extra time. For me, it's not worth the extra time so I'm going back to iDVD 5 for now.
Menu Rendering:
Neither version uses multiple processors for rendering menus so the time difference between the two wasn't as great and both only use 2 processors for encoding the video portion of menus, so there was only a small difference in time between the two version for this part of the project.
Audio Encoding:
The same is true for encoding audio - only 2 processors are used in either version so the time difference doesn't vary much here either.

perhaps iDVD 6 is also encoding to a higher quality, accounting for some of the slowdown and larger size.
That may be true. My last project seemed to fill up a SL DVD, even though it was 1 hour 27 minutes long. Prior to QuickTime 7.1, it even overfilled an SL DVD:
http://discussions.apple.com/thread.jspa?threadID=467718&tstart=60

Encoding speed

Similar Messages

Maybe you are looking for