Media Encoder Stops Encoding with the New CS5.5

[Neverends Productions]

I just upgraded to CS5.5 a yesterday and having some issues encoding an 1 1/2 program for a DVD. The program doesn't seem to crash per se, just stops encoding. The task manager says it's running. I tried to encode the same sequence, but shorter and have the in and out points go through the part where it always stops and it worked fine. It's just when I try the entire sequence it stops. Anyone have any experience like this with CS5.5 just coming out?

Thank you,

Doug

[Mark Mapes]

Hi All,

We recently isolated a repro case where encoding hangs with some regularity, and we'd like to know whether it covers the scenarios in which any of you have encountered hangs.

• PPRO sequence contains a frame sequence of very high resolution stills (greater than 1920x1080. Our test project uses stills of 5184x3456.)
• At least two or three such sequences are queued up in AME. It doesn't matter whether the first one is encoding when the subsequent ones are added or if you wait until they're all in the queue before starting to encode. However, the bug does not occur if you queue up only one job at a time, wait for it to complete, then add the next one.
• And, consistent with most reports on this thread, GPU acceleration is enabled.

In the above scenario, the first job encodes properly, but other jobs in the queue fail. Failure can take two forms:

1. AME hangs indefinitely on the last frame before the super-HD frame sequence; if that frame sequence is at the very beginning of the sequence, then it hangs before you see any frame in the preview thumbnail.
2. Encoding proceeds past that point, but the resulting video repeats the last frame before the super-HD content.

This failure case is consistent with reports in this thread that a given job might work one time but fail the next. If, that is, the instance that fails was queued up right behind another such job.

Two workarounds: wait until Job X is done before exporting Job Y; or go ahead and export Job Y but set its status to Skip in AME through the Edit menu or the context menu.

This case is not consistent with reports that AME hangs at different points in the job, unless of course you're talking about either different sequences with the super-HD frame sequence at different positions, or a given sequence exported with different Work Area range or In/Out points.

Please report whether this repro case fits your experience or not. We're very eager to know if we've nailed at least one of the scenarios that's been plaguing some of you.

[Paul_Duffey]

The times with my renders the only still layers I had was PSD layers the size of the screen (1920x1080), and it would hang in AME in the middle of encoding sequences, not at the end.

[Mark Mapes]

Thanks, Paul.

Just to clarify, I'm not referring to the hang occurring at the end of the sequence but on the last frame of regular SD or HD content before the super-HD content. If your sequences do not contain super-HD content, then the repro case we've isolated does not apply.

[PeterLaJolla]

Hi,

I just want to report that I am also experiencing random, intermittent freezing issues when exporting a sequence to queue up in AME CS5.5. I initially had CUDA GPU acceleration turned on.  I've been trying to encode for the last three days. More than half of my files have successfully encoded. However, lately, I've noticed that the freezing has become more frequent, to the point that I'm tired of restarting the queue and getting fed up. I will now try encoding the remainder of my sequences with GPU acceleration turned off. I hope that works.

FYI, here's my info when I was trying to encode with GPU acceleration on:

Video recording device: Canon HV20

Video recording mode: HDV 1080/60i (1920x1080)

Audio recoding mode: 2 channel, 384kbps, 48 kHz

Video capture software: HDVSplit 0.77 beta

Video clip captured files: AVCHD Video, m2t files

Sequence info: 29.97 fps, 1920x1080, 48khz stero,

Sequence lengths: between 14 and 37 mins

Sequence video layers: only one (imported video clips)

Sequence effects: Cross fade only between clips, no other effects or filters

Export settings:

   Multiplexer: MP4, iPod compatibilty

   NTSC: 1280x720, 29.97 fps, de-interlaced

   AAC: 64 kbps, 48Hz, Stero

   VBR: 2Pass, Target 5.00, Max 7.00 Mbps

   Use maximum render quality checked, Use Frame Blending checked

   Filters: none

CPU temperature: about 50.0 degs C

GPU temperature: about 51.0 - 53.0 degs C

Software: Adobe Premiere Pro Version 5.5.2

Software: Adobe Media Encoder Version 5.5.1.12 (64-bit)

System OS: Windows 7 Ultimate SP1, 64-bit

System Processor: : Intel Core i7-3930K CPU @ 3.2Ghz

Installed RAM: 16.0 GB (Corsair CMT16GX3M4X2133C9 DDR3 PC3-17066 2133MHz)

Graphics Card: EVGA GeForce GTX 580 3072 MB GDDR5

Graphics Driver: Version 296.10, (Graphics clock 772 MHz, Processor clock 1544 Mhz)

Graphics BIOS version: 70.10.48.00.82

Motherboard: Asus P9X79 Deluxe

I hope you guys are able to find the problem. The whole reason why I bought the GTX 580 was for the CUDA acceleration, and now its worthless.

--Peter

[Paul_Duffey]

For this one I'd try a couple drivers back on the card. The listed installed driver on my EVGA GTX 570 is 8.17.12.8026, but I know I downloaded the driver form EVGA's site around 5 months ago, so it's not at new as the latest one you posted which is a 2012. If you try a couple previous versions and that doesnt work, then I'd try to render directly out of premiere rather than queuing in the adobe media encoder. That's worked for a lot of people and with you just haveing a couple lengthy clips rather than lots of short clips, that might be the best solution for you anyway.

[PeterLaJolla]

Hi, Paul.

Thanks for the tips. I did as you suggested and just ran the render manually for each sequence out of Premiere, rather than queuing up in AME. I did this for five sequences without any problems, with GPU acceleration turned on.

I then turned GPU acceleration off and ran the remaining four sequences by queuing up in AME. I let the queue render overnight. They all rendered without problems.

The GPU acceleration made rendering dramatically faster. A clip which would take about 2 hours 20 mins to render with GPU acceleration off took about 12 mins to render with GPU acceleration on.

I did not backdate the graphics card driver.

[Paul_Duffey]

Glad that worked atleast,. I bet if you tried some backdated drivers for your video card that might work too. If you do try let us know the results.

[BigArmDigital]

Mark, this is EXACTLY what's going on in my case!

In my case I was queing up duplicates of the exact same sequence (titles, with the same background images but different PNG text tracks) and AME would run two sequences just fine and choke on the third sequence every single time. I had 16 different variations and no matter which I started with it would lock up at the start of the third sequence of the queue.

These sequences are 14 seconds long, 1920x1080p, but I was using a moving background track that was a 2324x1307 JPG file.

Turning off MPE and they queued up just fine (but oh so slow!).

Replacing the 2324x1307 JPG background image with the same (still moving) image at 1920x1080 and the queue worked just fine!

(of course there is a slight quality difference because I had a slight scale/zoom-in on the background image, but changing that image to a 1920x1080 image and no issues at all!).

If you need my system specs, please let me know. I've had this problem ongoing for six months but never tracked it down because it didn't always happen. I just got in the habit of only queing two sequences at a time.

In my case it is definitely the use of a scaling JPG that was larger than 1920x1080 that is causing the freeze.

[DirtyLowPull]

my AME failures of duplicated sequences almost always contained

subsequences of still frames, but they were always no larger than 1920x1080

- a previous version of PPro would fail to encode larger images so we built

a resize to a max of 1920x1080 prior to import into our workflow and on 5.5

we stil do that. Still AME would lock up randomly - on first or nth

original queued export - or later on while on one of the duplicated items

in AME - the duplicated are nearly always made in groups - queue a few

different prproj's to AME, then group duplicate all in the queue a few

times switching format and preset - I expect it to fail at some point with

MPE on.

[Stan Jones]

Dual core and 6 GB Ram is what google shows.

[jbiggs-nyc]

replaced the 2x1gb with 2x2gb.

[BigRedFish]

Hi,

This had been plaguing me for months. Down-sampling super HD stills in my project has resolved this issue for me.

I had several images (over 4k across) that I was using as slide overlays and AME would hang on exporting just one of these (not even multiple copies) using CUDA acceleration on my Dell Precision laptop with an Nvidia Quadro FX 3800M graphics card. I would also get Nvidia messages that my hardware wasn't good enough (Premiere Pro 5.5 would then crash) while I was editing these stills in the timeline - e.g. putting fade transitions onto them.

Changing the stills to be 1920 x 1080 last night has fixed both the export problem and the editing crashes and AME got to the end of the export no problem.

Hope that helps.

Kaleel

[nreedus]

I installed addition memory (8G) hoping this would solve the problem of freezing during export from AME. Unfortunately the problem persists with no clear path for a solution. I switched to PP from FCP and I like the system a lot. But if I can't export my 90 minute doc it is useless. Anybody, help, please.

[h2ofun]

Have you tried exporting straight from PP?  I gave up on AME a long time ago.  All my exports from PP work just fine.

[nreedus]

Exporting now from PP. 10 hours. Looking for post house to finish. Thanks.

[BigRedFish]

Are you editing video with large stills in the project? I down-sized my oversized stills to 1080p and it works fine now.

[tjsCreative1]

I've solved this problem for my system and it was easy.

I'm embarrassed to admit that I was cheating the system using a non-approved EVGA GTX550Ti and editing the approved list from a posting I had found on the internet.  Most of the time, everything worked well and the Mercury Render Engine raced through stuff.  On the more complex projects or with big stills, Premiere might occasionally crash and then on a couple projects, Media Encoder would just stop at random and couldn't get through the project.

This past week i had an issue on a fairly complex project and decided that the issue was costing me money, so i went out and bought a new EVGA GTX580 with 3 GB video ram.  The same project breezed through with no issues and as hard as we have tried, we can't even make problems any more.

I've learned that the Adobe and Nvidia experts may still be the real experts out there and I should just listen to them.  I paid a lot more in lost time and problems than I would have if I'd have just bought the right card in the first place.  I hope this same solution works for everyone.

[BigRedFish]

tjsCreative1 that's great and fantastic that it's resolved your problem.

For my own setup, I have a desktop machine with a non-approved card in, but my laptop is where I do most editing and exporting because I do a lot of work at home. It's a Dell Precision M6500 with an Nvidia Quadro FX 3800M in it, which is on the approved list. However, AME was still exhibiting the same hanging issues on several projects (and during rendering operations) and the thing it was tripping over was the over-sized stills.

[tjsCreative1]

Oversize stills have bothered Premiere Pro for me for years, way before Mercury.  Sorry to hear you have problem with approved card.

I can say my Asus notebook i finally had to give up on.  Its Nvidia graphics card always overheated and reset, and Agearnotebooks said there was nothing wrong with it.  Now i use a Macbook Pro 17 with Windows 7.  I'm not impressed with MacOSx - it's like working with handcuffs for an old-fashioned Unix/Pc guy.  My next notebook will be a real PC again with Nvidia and i hope it works right.

[nreedus]

I do have quite a few stills but it doesn't freeze on a section with stills. Trying an export directly from PP now. I want to hand this baby over to someone else. Frustrated, tired.

[zikade]

By any chance - did you import it from FCP or did you create a fresh project? You may also try to disable CUDA for exporting - if enabled.

[BigArmDigital]

@nreedus: When it's stills that cause my issues it doesn't freeze in the section with the stills, it just locks up at some point, usually not related to where the stills are in the sequence. Resizing the stills down to 1080 (or close) always fixes the issue for me. (I have 16GB of RAM and an approved nVidia card and still have issues if my stills are too big.)

Any beta testers of CS6 know if the new AME fixes the large stills issues?

[Mark Mapes]

I'm not a beta tester, but I can report on the status of the hang involving sequences that contain very large still images. Unfortunately, we did not figure out until very recently that the bug occurs only when two such sequences are queued up back-to-back in AME. As a result, although we had the bug on the books, we were missing one of the keys to reproducing it reliably, and at one point when we could not repro the hang we wrongly concluded that it was fixed. By the time we discovered that the bug was still present and nailed down the critical detail, it was too late to fix it for the next release. Our developers are working on a fix now.

Incidentally, although the bug manifests in AME, it actually results from an issue on the Premiere side. That means that if the fix makes it into a patch--and let me emphasize the "if" because I can make no promises--then it will be in an update to PPRO, not AME.

I know I posted this workaround before, but it's worth repeating: if your PPRO sequence(s) contain stills >1920x1080, avoid queuing multiple jobs. Either wait until Job X is done before exporting Job Y, or go ahead and export Job Y but then set its status to Skip through the Edit menu or context menu. And, of course, as you pointed out, BigArm, scaling the source assets also works.

[PeterLaJolla]

That may reproduce the bug many of the reported cases, but in my case I had no stills whatsoever in any of my sequences. (See my post March 26, 2012).

[staheph1]

I am having the same problem with a project with stills >1920x1080, but only one job at a time.
Just exporting one by one is not the solution. Encoding stops randomly, no error message or "not responding"

Same occures if exported via AME or PP directly.
After 2 hours of idling I abort the encoding. Message from AME: "Encoding is running, do you really want to quit?" (CPU load 2%, RAM at 2.9GB)

scaling the stills is not an option since I want to zoom in without loosing resolution.

Windows 7 Ultimate Service Pack 1 64bit
Intel Core 2 Quad CPU Q9550 @ 2.83 Ghz
8GB RAM (RAM usage never goes beyond 4.5GB, eventough lots of free RAM is listed, 1.5GB reserved for other applications than Adobe)

Production Premium: Premiere Pro CS 5.5

Project:
Lengh 4.5 Minutes
Timelapse of 12000 stills (38GB)

at 2592*1944 2.8MB each.

Sound with .mp3 file

Export:

Video: NTCS, 1280x720 or 1920*1080 29.97fps, Quality 4, Progressive

Audio: MPEG, 384Kbit/s, 84kHz, 16bit, Stereo
Bitrate: VBR, 1-Pass, Min. 4.00, Target 10.00, Max. 18.50 Mbit/s