Slow rendering

[freakedenough]

I wonder if this is really the maximum GPU and CPU usage when rendering a H265 4k60 video. It takes almost a day to render a 2hrs 4k60 video. Why is there not a 100% GPU usage and 100% CPU usage noticable?
It looks like that VideoStudio 2023 is not multi process capable.
Unfortunately, 4k export was unavailable in the Demo, thus I am more than unhappy that I spent 100€ for ending up with the same slowness that OpenShot suffers from.

[Davidk]

Most applications - except games - cannot use the parallelism that exists in todays' processor hardware. Multi-cores and multi threads are entirely possible, but the apps are not usually programmed to use it. More than 3 cores (1 for the OS, 1 for the AV, and 1 for apps) is about as good as it will get. Using an "assistant cpu" in the form of a display adapter can be helpful if the right drivers are installed: editors generally can make good use of those, but any improvement will be about 10-20% of the un-adapted version. However, generally you would only really notice an improvement in performance doing a repetitive task that takes a while - like rendering.

However, you should appreciate just how much of a load the modern hi-resolution formats impose on a rendering task. I've attached a picture image of a slide I used to use in a VS class I tutored some years ago, showing the pixel difference between DVD, FHD and 4K per frame

Rendering builds the final result you want, frame by frame, based on the recipe in the project file. The amount of picture data the app has to process per frame with 4k is nearly 20times that required for DVD, per frame, on top of any other processes you specified in the project file. And depending on where you are in the world, there will be 25x (PAL) or 29.7x (NTSC) that per second. So, yes 4K is going to take quite a while. Just how much of a while will depend on the various components - memory, SSD or HDD, computer clock, display adapter etc - in your computer, and every machine will be different. I'd suggest that you create some small 4k projects - a vsp of say 5minutes - with the sort of editing you normally use, and rendering that will give you some idea of just how long a 2hr project will take.

As a rule of thumb, on my desktop tower machine I ordinarily allow about 1.5 or a bit more times the length of a vsp for any FHD project (a 20 minute project will render in between 30-40 minutes). My laptop is 33% slower, so that extends to nearly an hour on that PC, main reason I don't usually do projects on it. I'd expect a 4k project to take at least twice the project file time to render.

[freakedenough]

I appreciate your comment, unfortunately what you wrote isn't new to me.
Its sad that VS23 is still not programmed to use multithreading and all cores available to speed up rendering. It obviously didn't scale with the hardware in recent 15 years.

[zaphodikus]

I had an NVidia card, the card never made any difference for me to render times when I turned it on. I now have an AMD though. Not sure how much difference it makes, but your render time is not surprising. I get 2 cores "busy-ish" when rendering and that's it. Normally I just do something else while it renders. I assume you are all on SSD or high speed NAS media, that's was my biggest bottleneck on my old machine.

[Davidk]

A big part of the problem is generality of hardware - not many machines will have the multi-core multi-threading capability of top-end cpu's, but every one has at least a single core, or 2. The operating system can manage those, but the tricky program effort to any multi-core application usage performance is likely wasted on everyone except those top-end few, which is really why app programmers don't use it: what they write has work for every user. If the software is installed on a top-end machine, it would have to auto-detect it and seamlessly adjust to that: not easy to do.

The best way to get performance is to focus on the base PC - speed it up for everything and the editor naturally follows. The various ways to do that have been discussed ad-nauseum in these pages, but a fast cpu (meaning, at least 3-4ghz clock, desirably more) plus lots of RAM and a 2nd generation SSD (NVme interface) and a compatible display adapter go a long way towards that end. Note that the display adapter will probably use as much power as the rest of the PC, so you need a power supply that's rated for the task (double the normal tower fit) and because it goes in the expansion bus, is really only suitable for motherboard-based desktop machines.
Another way to improve perceived speed is to change your work practice: break the task into sub-projects and make each one small - say, 20 minutes each - and all using the same rendered parameters. When each one is complete place them all in sequence in a composite render, which because they are all rendered to the same spec, means the final render should be fast and uncomplicated.