Gradual slowdown on batch exports...normal?
-
Sorry this one is a bit long to explain, but here we go!
Recently completed my first large project with Templater where I needed to export around 95,000 short 6-second animations (with just a swapped text layer and image in each. Everything worked beautifully with the exception of one behavior that Im trying to figure out if it’s normal or not.
When the export started, it got through about 500 videos in an hour. Then I’d notice that with each passing hour, it would export a few less and a few less. After like 8 or 9 hours, it would drop dramatically, exporting literally only like 10 an hour.
Here is what I observed in the Render Queue. The exporting process itself was as fast for each and every video (only like maybe 7 seconds or so). But the inbetween stage where it paused to swap out the layers and re-start the export got longer and longer and longer. By hour 8, it would sometime be like 2 mins between exports rather than a couple seconds!
Now obviously one might assume a RAM issue, but the consistent slowing over hours, rather than just reaching a slow speed and flatlining there seems to maybe indicate otherwise? I have 64GB of RAM and AE was displaying about 72% RAM usage the ENTIRE time, varying only by a few percentage points even from the very start.
Is this normal? I know a couple folks on the AE team and told them about it and they thought it was interesting and prooobbably a memory leak someplace but werent sure.
-
I should add that the only solution was to force quit AE, and then as soon as I restarted it and set up the export again, it started flying again, reaching the same problem hours later.
Also should have mentioned Im on an Intel Mac running 12.3 with the latest public release of AE, 22.3 (although I think that was a recent update and I was on the previous version when I had this issue… either 22.2 or 22.0?)
-
Hmmm, that’s quite odd. I haven’t seen this particular issue before, but I’d be willing to bet that it has to do with how Templater handles imported assets in After Effects.
To save time and reduce redundancy, Templater will try and match up a reference to an asset with the currently loaded footage items in the project file. This is a size/time-saving feature so that the same footage doesn’t have to be loaded multiple times, thus creating duplicate footage links and reducing asset efficiency.
In this case, I believe that, as the project runs, more and more assets are being stored in the project file. As the asset list grows, so does the amount of footage that Templater has to check to see if it can use a previously loaded file to complete the versioning process.
The first thing I’d try is to enable the “Remove unused footage after each job” option in the Templater Preferences panel. This should keep the project from ballooning up in size as more and more assets are processed. You might also consider enabling the “Purge all memory caches after each job” option, just in case there is a memory issue that’s causing this problem.
If neither of those options affects the issue, we’ll probably have to look into the circumstances in more detail to track down the cause of the performance decrease.
Let us know how it goes, and we’ll take it from there.
Thanks!
-
@jasontcox Let me add to what Jeff mentioned, that this sounds like the option in the Templater Preferences “Purge all memory caches after each job” might be the best choice to help alleviate the runaway render times. This can add a slight bit of time between each render, but it won’t stack up like it’s currently doing. What’s happening is that AE’s undo history is filling up with hundreds or thousands of actions. When you’re doing smaller batches or longer videos, it’s not as noticeable, but when you’re doing a significant volume of even short, simple videos, the undo levels pile up fast.
-
Thank you both for responding! I worked on this project about 6 weeks ago and I THINK I tried the “Purge All Memory” command but might have to try another test export. Im pretty sure I did NOT try “remove unused footage after each job” so I should definitely try that. Thanks and Ill try and come back here after I have time to do another test!
-
@jasontcox Just checking to see if you were able to solve your issue. Let us know if you get a free moment. Thanks!
-
Hi there!
I’ve got the same issue due now. I’m working with a much smaller amount of comps, but they’re longer – from 1 to 1:30 minute, 3 comp in a row with updated data.
Up to 100+ changing texts.
So I’ve ran it in two different windows machines and both are slowing.
On first machine results went from 10 minutes to 37(in like 3rd time of rendering), on the second one 4:31 to 4:53 just in two renders.
Purging after job and clearing preview cache is on
-
@zhigulingo @jasontcox Did you guys find any solution? I am having the same problem
-
@ermalr We have seen some systems do this, but unfortunately cannot pin down the issue. It seems to happen mostly when the .aep files switch between jobs. Are you using the .aep column or prop in your data to make Templater switch template files? Also, what kind of drive stores the .aep files that Templater processes? Is it a shared or sync drive or a mounted folder? Are you able to observe where AE is hanging the most? Is it rendering phase, gathering data phase? Also, in the log files you can examine the timestamps to see where the slow down occurs on the Templater side.
-
If you continue to have issues with speed and performance over time, I’d recommend getting in touch with us directly at support@dataclay.com so we can get some more specific details on how the project is configured.
Outside of enabling the “Remove unused footage after each job” and “Purge all memory caches after each job” in the Templater Preferences panel, as @ariestav mentioned, it might be necessary to take a look at the templater.log file to see if we can narrow down exactly where the issue is occurring.
Let’s start by enabling those options, and if we still run into issues, go ahead and send us a message, and we’ll see if we can’t make some more specific suggestions based on the project’s operating environment.
Thanks,
Jeff
