2% CPU Utilization Premiere Pro CC 2017

[ShadowWolf810]

Hey guys, 

 

I usually use Avid and DaVinci for my editing, but do have Creative Cloud. Was helping a friend with edits on his project, and when it came time to render I figured we could save some time by opening up the project on my system and rendering there, instead of on his 5 year old Macbook. 

 

Specs: 5960x, GTX 1080 TI, 64 GB Ram, 960 Pro Boot drive, etc. 

 

Point being, even Premiere's notoriously bad multi-threaded optimization shouldn't cause the same project to take 1.5 hours on my machine, and 40 minutes on an old Macbook. Even though we didn't take the time to move all the source clips to the 960 Pro, it was all stored on a decent external USB 3.0 drive. It was a rough draft of his project, so there's no color correction or stabilization, like 3 dissolves etc and it's all H.264 footage from a DSLR being rendered back to H.264, otherwise known as possibly the easiest render job possible. 

 

Have done some digging but only was able to come up with people saying there's nothing that could be done about it. But when Youtuber's do render tests on similarly spec'd systems they're getting close to real time render's in the same sort of scenarios. 

 

Is there some sort of setting I'm missing buried somewhere? Any thoughts would be appreciated! 

[emosun]

if you have a 1080 ti why are you cpu rendering in premiere?

[ShadowWolf810]

Just now, emosun said:

if you have a 1080 ti why are you cpu rendering in premiere?

Can you even do a full render out as GPU only? To my knowledge its just GPU acceleration. But I don't know Premiere very well since I never use it, wasn't gonna waste time getting everything back online from an AAF in something else to render. 

 

Regardless, CPU should not be at 2% on all cores from the very beginning of the render, even if GPU render would have been faster. 

[emosun]

2 minutes ago, ShadowWolf810 said:

Regardless, CPU should not be at 2% on all cores from the very beginning of the render, even if GPU render would have been faster. 

well idk what the situation is there or specifically what you were rendering at what settings.

but essentially the gpu (even if the cpu was working normally) will be 10 fold faster than the cpu. I know that even my 1060 outruns my xeons in the final render. I usually only use the xeon when editing as they seem quicker with on the fly tasks

[ShadowWolf810]

1 minute ago, emosun said:

well idk what the situation is there or specifically what you were rendering at what settings.

but essentially the gpu (even if the cpu was working normally) will be 10 fold faster than the cpu. I know that even my 1060 outruns my xeons in the final render. I usually only use the xeon when editing as they seem quicker with on the fly tasks

Hmm interesting, I might try out some test renders then with my own footage, I don't have his project anymore since he took his drive home. It was very simple footage, source was H.264 from DSLR at 1080p, render settings were standard H.264 1080p, 10 Target and 12 for Max Bit rate. Was a 15 minute cut that stabilized at 40 minutes on the render on his machine. 

 

I guess the consensus is that the CPU isn't working normally though? Like I understand not getting 100% usage on all cores (although that's what I get every time rendering out of Davinci 14) but it seems like at least 30-50% should be guaranteed.

[emosun]

44 minutes ago, ShadowWolf810 said:

I guess the consensus is that the CPU isn't working normally though? Like I understand not getting 100% usage on all cores (although that's what I get every time rendering out of Davinci 14) but it seems like at least 30-50% should be guaranteed.

i think it's important to note that not every project will use a lot of cpu power to render it

For example if you loaded the project down with unsharp masks and medians and frame blending then you'd definitely see 100% usage

but in your case the cpu wasn't really needed and something else was preventing the project from rendering frames quickly.

[ShadowWolf810]

39 minutes ago, emosun said:

i think it's important to note that not every project will use a lot of cpu power to render it

For example if you loaded the project down with unsharp masks and medians and frame blending then you'd definitely see 100% usage

but in your case the cpu wasn't really needed and something else was preventing the project from rendering frames quickly.

Ok but if GPU rendering isn't enabled then the CPU has to be rendering the frames, and especially if there are none of those highly CPU intensive effects applied in the timeline then there's really no reason for it to run so slowly. 

 

I've been looking into trying to do the GPU Rendering like you said, only thing I'm seeing is Youtube tutorials of people modifying text documents, but not actual settings in Premiere itself, is this what you're referring to?

[emosun]

you go file>project settings>general>video rendering and playback>and select the gpu accelerated cuda option

 

might have to save then close/reopen the project just to be sure it's using it

[ShadowWolf810]

11 minutes ago, emosun said:

you go file>project settings>general.video rendering and playback>and select the gpu accelerated cuda option

Oh yup, I'm blind. I have a tiny bit of experience with CS6 and the General settings looked different. Ok well here's the curious thing. 

 

The CUDA acceleration was already enabled, which should mean that when I was testing on my friends project, it was also enabled since I haven't messed with it. 

 

I tried my own footage, this time a 25 minute H.264, rendered out to an H.264 just like in the previous project and immediately got 80% CPU, and render time of 9 min. 

 

Turned off CUDA acceleration, saved the project and closed Premiere, re-opened it. Rendered the same clip again with it on "Mercury Playback Engine Software Only" and got the exact same results for render time and CPU utilization. 

 

So its weird because A, we got what I would have expected in the other project with just changing footage implying that something was affecting it in his project. And also B, we didn't see any improvement with CUDA acceleration and you claimed it'd 10x the speed