Copy link to clipboard
Copied
Hi.
So, latest version, updated and all that.
CPU i5-11400F @2.60GHZ
32GB RAM
GPU AMD Radeon RX 6800
I opened an old file, which with my old configuration was running super smooth, it only took about 30 seconds to render a 6 sec. logo opener with some particle effects in it.
Old configuration:
CPU: i5-660
16GB RAM
OnBoard GPU
Right now, CPU is 100% when rendering, drops down to 50, then spikes to 100, then drops down again, so on..
Same file on old configuration was rendering super smooth and didn't abuse the CPU like this.
I updated everything, I pre-comped some complex effects (althought the same effects on old configuration were running super smooth), still having the same problem.
Any ideas anyone?
Thanks.
Copy link to clipboard
Copied
this is expected. only a couple effects use GPU, and multi-frame rendering (MFR) is slower than normal rendering in a number of workflows. GPU usage is essentially nothing, all the time, for every workflow- I don't think it's even used to generate the UI, which is why making sure the renderer ignores a core can speed up performance, by giving headroom so it doesn't just kick your computer in the family silicon every frame.
pre-comping is, for the most part, purely organizational, and is more likely to reduce performance than increase it. note that your new CPU has a lower clock speed, and that your GPU is discrete now. After effects was designed for single core cpu's, and honestly not much else. the further you stray from it, the more you're using patches, updates, and not the core or original code of the program. The people who first coded it are unlikely to still be doing so after 30 years, so in my purely biased and assumption-full concept of it, the people coding it might not fully understand everything going on in the background. It's like taking backstreets instead of the highway. some are faster, but there's more places to hold you up and it's nowhere near as direct.
if the old file or version of AE was prior to 2015 or up to 2017, it likely also had access to multiprocessing, a more advanced, if messier and more incomplete in code, rendering schema. I keep a copy of CS6 kicking around for exactly that reason- I don't need or use the bells and whistles added in the last decade and it can be ported forward if I ever do.
Copy link to clipboard
Copied
Sow hat you are basically saying, is that my new configuration does very poor job at increasing performance of the overall application, and that no matter what PC I build, this will be the same across all platforms..
If this is the case, Adobe needs to seriously think about scraping it all and rewriting it from start.
Copy link to clipboard
Copied
as clips or whole podcasts, these are pretty good podcasts on the subject. it's more complicated and has quite a nuance, because everyone I personally know who use adobe products has gone through the same thing. clips because they get to the important bit and don't spam a big inline video here.
https://www.youtube.com/clip/UgkxRKVxSZbTuOH6u4tlJ5DmP5B49vbyh2SF
https://www.youtube.com/clip/Ugkxn5_DhGxpTsdlrHuhxuCw9WYXnEsiolQ9
https://www.youtube.com/clip/Ugkxsjp5XfIgDJOuBZSGv0rmPECI9eJifOxY
https://www.youtube.com/clip/UgkxqoQ4eS3V5d1njuXDnjHEgNvjcaVbKJsp
https://youtube.com/clip/UgkxlyjT0jkCZU8fwukW607OM6kfwWp2j1qy
and trust me, you're not alone in that thought. 😉
https://community.adobe.com/t5/creative-cloud-services-discussions/cc-and-the-ui-is-too-slow-every-r...
I went through exactly what the corridor crew did 9 months in advance. I think, at this point, everyone using AE has this moment sooner or later.
even if adobe is a big soulless company, it hurts that they don't listen. I'm a human to the best of my knowledge and I care. if that means anything. it's part of why I'm still here.
Copy link to clipboard
Copied
Thanks, will look at these later tonight.
What I found out about my file though, is me being super-stupid. There was a pre-comp with a ton of particles, and I realized this as I was observing the rendering process. Once the particles were rendered and faded out completely, the process would speed up. So it's just me being dumb [removed by moderator], I guess.
Copy link to clipboard
Copied
still, it's odd that such a thing is even an issue. if a game, which doesn't need to do raytracing, can render a million particles 60 times a second, why can't a program whose entire reason for existing is to render things? some might quote "precision" or so on, but is there a reason those optimizations, that work in 3D space, aren't suitable for 2D as well? we're seeing things like unreal engine pulling off "the mandalorian" in near realtime- but why should it be slower when the passes are precomputed by the camera that took the video? Some things, yes, like tracking, and rotoscoping, and so on make sense, but at the same time a single video with alpha? we've had algorithms and hardware to handle that better for decades.
Copy link to clipboard
Copied
I'm sure a time will come where Adobe will want to stay relevant and will decide to scrap AE and rewrite it.
What UE did was a masterpiece, and in real time??? I was very amazed by that. Their engine is so chad. My guess on why they are still not doing anything to improve the performance of AE is because of money, or they are still simply thinking inside the box. I did a little research these past few days, it's not that hard to create a rendering engine that will utilize the GPU in a way that Twinmotion does. I mean, it sucks the life out of the GPU, but it's so much faster. I rendered a 3D video in real-time with this new configuration in less than 30 minutes, and it had complex camera movements and physics and what-not. I guess we can only hope to see this in our lifetime lol.
Copy link to clipboard
Copied
I doubt it. They're in the business of making money. they'll be hesitant to do something which isn't proven, as their lives are basically on the line. They'll be breaking the law if the shareholders find them in contempt of "not working in their best interest".
Besides, look at autodesk- they did the same thing years prior, and on the same timescale, too, they started sucking up smaller companies, their programs got old, obsolete, and bloated, and after a decade and a half were consigned to "industry standard by name only"... except adobe never became industry standard beyond photoshop and acrobat- they didn't make any difinitive plays to become a media giant outside of small businesses and single use, and still have basically no place in any government institution or highly technical field.
so I see them going the way of autodesk- letting decades old programs fizzle out, huddling around a dying fire in a blizzard, confident they'll never need more firewood again because the ones who made the wood pile in the spring were so masterful. If they do nothing, after all, they're never in danger of angering the shareholders.
Copy link to clipboard
Copied
If you use a desktop type PC, you change the graphic card to NVIDIA one.
Copy link to clipboard
Copied
What's the difference though, care to explain?
Copy link to clipboard
Copied
Nvidia GPU's are simply more common at the moment among consumers, so if code has to be optimized in a way some hardware won't like, they'll prefer to optimize for the one more of their users have, and Nvidia (well, all GPU manufacturers I know) have a dedicated line of optional studio drivers that interface with production programs differently, though I've never noticed a difference in adobe software. maybe it's just stability?
Copy link to clipboard
Copied
It doesn't make sense though. From what I've read online and what @nubnubbud said, AE optimizes GPU very little (some special effects and small things like that), and all the other work is CPU.
I don't see how an acceleration software by Nvidia would change that, as the core of the software is programmed to utlize CPU over GPU.
Copy link to clipboard
Copied
precisely. the difference will be minimal at best, through there will be one- it's because the optimizations probably pertain not to the processing (I'll bet my pants they're using standard openCl) but that it's optimized to hand files over faster, without a translation layer. basically it's like if you're handing a bucket of data back and forth, but because you don't know AMD as well, you just hand them a generic bucket, which they fumble with 'cause there's no handle. It's almost the same speed, but over a thousand passes back and forth, it becomes noticeable. If the packets were larger and fewer, the bucket type would matter less, etc.