If you've read though other threads you'll see that the load depends almost entirely on the specific kinds of edits you've made, what format your source clips are encoded in and what profile you're producing to. For example, using color presets or LUTs can only be done with the CPU and cannot be offloaded to the GPU.

Follow JL_JL's suggestion of using the Skateboard.mp4 clip in this post to get a clearer utilization view.

YouTube/optodata

DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°

Quote If you've read though other threads you'll see that the load depends almost entirely on the specific kinds of edits you've made, what format your source clips are encoded in and what profile you're producing to. For example, using color presets or LUTs can only be done with the CPU and cannot be offloaded to the GPU.

Follow JL_JL's suggestion of using the Skateboard.mp4 clip in this post to get a clearer utilization view.

Thank you. I will try that clip and see whats going on.

I've done some tests and it shows that as you say putting the skateboard clip in the timeline does start to use the GPU but I could not workout what is so special about this clip?

I then found the killer in this was Fix / Enhance. Certain things will take full use of the CPU and deduct this from the GPU use. Video denoise alone will take utilization from 40% down to 9%. Add some colour fixes and your down to 4%

This has shown me the nature of how this works, correct me if Im wrong but it seems the CPU and GPU cant be optimized to work together simultaneously. If the GPU does 30% the CPU will only do 70% which tells me either this is how all apps use this tech or it needs re-thinking as optimal could be for both CPU and GPU to both work at their max abilities.

The harsh reality is this then. Use PD features to make your video how it should look and sound and wait in my video project 1:26, or turn all fix and enhance features off and wait 5:10.

Question is is there a feature than can pre-render things or speed this process up?

Thanks.

Quote I've done some tests and it shows that as you say putting the skateboard clip in the timeline does start to use the GPU but I could not workout what is so special about this clip?

I then found the killer in this was Fix / Enhance. Certain things will take full use of the CPU and deduct this from the GPU use. Video denoise alone will take utilization from 40% down to 9%. Add some colour fixes and your down to 4%

This has shown me the nature of how this works, correct me if Im wrong but it seems the CPU and GPU cant be optimized to work together simultaneously. If the GPU does 30% the CPU will only do 70% which tells me either this is how all apps use this tech or it needs re-thinking as optimal could be for both CPU and GPU to both work at their max abilities.

The harsh reality is this then. Use PD features to make your video how it should look and sound and wait in my video project 1:26, or turn all fix and enhance features off and wait 5:10.

Question is is there a feature than can pre-render things or speed this process up?

Thanks.

Nothing special about the clip. As has been mentioned numerous times in the forum pages, PD's architecture has some activities that are CPU based so performance is timeline editing content dependent, it's not like a massively shared equal access to GPU/CPU with computation load split.

When you use say color correction, this frame by frame preparation is done by the CPU prior to the GPU doing the encoding. The GPU can't encode at a higher throughput, more load, as it's waiting for the frames to be color corrected and passed off to it for encoding. Get a faster CPU and GPU load will go up, have a very slow CPU and GPU load will be very low.

You can preproduce all segments of the video and then pull these into a final project and do no additional editing and use SVRT for the final production. This will be very fast as long as SVRT restrictions are adhered to as no video encoding is done, just a copy assembly. One must pay the fiddler at some point though, either in preproduction or final production.

Jeff

The GPU is very, very limited in what it can do with it's CUDA (3D) cores. And a 2080Ti will blow trough those limited tasks very quick, bottleneking on the CPU tasks. The effects that are GPU capable are marked as such. See in the attached picture the effects that have a nvidia logo in the lower left corner.


More, the actual final encoding of the video (that would take a lot of time if done by CPU only), if done in H264 or H265 formats, is hardware accelerated (processed) by a small ASIC core inside of the GPU (calle NVENC in nvidia's case). That special core, different from the 3D CUDA cores, is the same no matter what GPU you have (in the same family/series). For example GTX 1660 will have the same ASIC core as the 2080Ti, no speed difference between them:

https://developer.nvidia.com/video-encode-decode-gpu-support-matrix


PS: On the laptop systems the situation is even more dire, the Optimus technology is not able to work properly with the NVENC core, defaulting on the integrated Intel GPU encoding core.

Quote If you've read though other threads you'll see that the load depends almost entirely on the specific kinds of edits you've made, what format your source clips are encoded in and what profile you're producing to. For example, using color presets or LUTs can only be done with the CPU and cannot be offloaded to the GPU.

Follow JL_JL's suggestion of using the Skateboard.mp4 clip in this post to get a clearer utilization view.

This is an issue that I regularly have to explain to my YouTube viewers.I'm going to add a link to this thread in GPU video as I think it sums things up quite well. Subscribe to PowerDirector University on YouTube.

Subscribe to PDU Mobile on YouTube.

Thanks your all your replies, much appreciated.

The CPU is certainly a limitation or bottleneck here. I cant see me using SRT as Im always going to be doing some editing of some kind and it will need rendering either Pre or Post so in the long run it wont save time but add additional steps. My CPU is a Intel 5930k all cores running at 4.3ghz.

So to sum up grab any video clip, dont add anything from Fix/Enhance, and apply the general steps to allow it to use the GPU. It will at this point use the GPU until you add something from the Fix/Enhance. The more CPU intensive that task is, the less the GPU will be used as the GPU is waiting for the CPU process to be rendered first, this is where the time adds up. The slower your CPU is the less your GPU will get utilized as the CPU can be the bottleneck in the system.

How to find the balance? Do basic colour corrections in PD, shoot the best footage you can with the best audio quality and lighting.

While you wont get much GPU utilization as in my case it took my rendering from 1hr 24min down to 12min.

Thanks.

Quote ...shoot the best footage you can with the best audio quality and lighting.

IHMO thorough preparation during the production phase is a critical component of a pain free post-production experience. The more work you put in during production the less dependent you will be on the CPU to render enhancements/effects.