The 2018 roadmap at
https://community.kde.org/Kdenlive/Deve ... ng_Roadmap
has support for "Fast render (multicore)" as a mid-term, post refactoring goal and support for "GPU processing" as a long term goal.
I want to get a new machine by purchase or build one myself yet again. I am just coming back to the current hardware scene after about 7 years.

I am considering a high end Ryzen CPU like a 9 3900X and low-medium GPU like the Nvidia 1650 Super (with a cost ratio of about 3:1).
Are those good choices for near term like 2020-2022?
That is, can I expect Kdenlive to take advantage well of all those cores?

How soon will use of GPUs be enabled again for rendering?
Which ones will be the target for development, e.g. Nvidia 1650 Super or an RTX board?
And/or AMD/Intel with certain properties?

Are there other options like RAM and SSD speed that have to be in the tradeoff cost-effectiveness mix?
Is there a cheaper combination that is more cost-effective like Ryzen 7 with cheap GPU?

i think that there is not a list of better specifications/confugurations for Kdenlive right now... the project is under development and seems that a lot of things are managed with single threads.. GPU acceleration for now is switched off due to compatbility issues.
the only thing i can suggest is to have a good CPU (lot of cores/threads and 3.5+ ghz speed) because this is the only thing that is surely used during rendering (final rendering...) i switched from a an Athlon x4 to a 6+6 xeon and i noticed no changes on interface speed.
during the rendering (1080p) on my system the memory usage is 6Gb (of 32gb)

any investiment on GPU and RAM (right now) is only for other applications (games?).

i also not noticed variations using different SSDs (1080p) ...

for my opinion the Ryzen 5/7 are the best CPUs on the market (1800 ... 2800... 3800...) maybe just not spent a lot of money on motherboard .. i noticed no changes at all swapping from A320 to B450 chipsets (not on Kdenlive)
i think that my next pc will be a Ryzen 7 1800x (very cheaper used) with M2 Disk. (and this just for kdenlive usage)

Thank you, bartolini, for your suggestion.

Can you tell me if setting up a profile as mentioned in the exchange between J Regenold and Constantin Aanicai in the discussion at
https://kdenlive.org/en/2019/11/kdenlive-19-12-beta/
can actually cause a GPU to be involved in rendering now?
That assumes an Nvidia GPU card capable of using NVENC.

On the other hand, it is at least tricky as in this post by a person who participated briefly.
viewtopic.php?f=265&t=153218

Is there value in trying to figure out how to use such profiles and using NVENC now?
Or is it better to just have cores on a fast CPU until developers return to the GPU issue for rendering?
I don't know enough and have no experience trying NVENC and do not have an Nvidia card.
I am guessing that you prefer to wait.

Nvenc for now is able to ruin quality...
viewtopic.php?f=272&t=164576&p=428525&hilit=gpu#p428525

Thank you again, bartoloni,

So NVENC in a profile does affect results, but not the way we want, unfortunately, for now.
I hope Merlimau is right in his estimate that "We hope we can implement GPU accelerated rendering during this year."

I will give my new setup thought and see about learning about the use of profiles which I did not use last year in my first experiences with Kdenlive.

I'm throwing 24 threads at Kdenlive and get - after the rendering crashed - a message that assigning more than 16 threads is not recommended.

So much for CPU muscle.

Thanks for that report of inadequacy of the current code.
Let's hope Merlimau is correct and it will be addressed soon.
Maybe suggest it should be able to use Threadripper CPU's many threads?
I hope there is a progress report or even a place to kick in donations for developer time....

An update after I started testing the AppImage of 19.12.3: Render times seem fine, even with GPU acceleration switched off (as the AppImage version recommended). Htop reports CPU loads of 600..700 %, so Kdenlive is definitely using more cores than the setup default of "2 concurrent threads" suggests.

I get the impression that multi-core support of Kdenlive is not so bad after all.

However, just like pwrcul I would be happy if I could use some of my GPU's power, too.

That is very good news.

A good target CPU might be Threadripper 3990X with 64 cores and 128 threads though at a cost of almost US $4K it is out of reach of most individuals and many organizations. It is real silicon available now. It may take a good while before exceeding that is a practical option.
https://en.wikipedia.org/wiki/Ryzen#CPU ... sse_(Ryzen)_and_Castle_Peak_(Ryzen_Threadripper)

[Sorry for off-topic]

I would be curious to see a comparison between my roughly 10 year old CPU design and a recent Threadripper. When I bought the used rig a few years ago (before AMD launched its current blockbusters) it seemed like a good choice for encoding: two sockets, each holding a 6 core Xeon X5690. With hyperthreading that gives me 24 threads - a lot back then and still more than many applications can handle.

Still got away for a fraction of the cost of an X3990X. I am planning to upgrade to one of the new AMDs though. Intel won't see a dime from me again, not after Meltdown and Spectre and how they handled that.

When I bought my workstation I was aware that most FOSS programs couldn't use GPU power as much as Windows based programs. That also applies to RAW converters and graphics programs like GIMP. That's why I chose to invest in CPU power and not a powerful graphics card. I only had to exchange the included Quadro 4000 because it couldn't feed my new 4k screen. Good riddance, Nvidia!

Current render times for your comparison:
A 7:30 minute clip of 1080p50 source material rendered to MP4 H264 at highest settings (Video 24 Audio 256) takes approximately (and consistently) 15 minutes and results in a 200 MB file. I can live with that.

hello everybody.
I'm a daily kdenlive user for work (news editing) and personal use (youtube channel), and i love kdenlive.

i always use 18.* version 'cause i do not like the new multitrack idea that "force" separated audio and video tracks, 19.* is faster in rendering time, but is slower in scrubbing the timeline AND lacks of filter end effects (auto mask for example).
So i use 18 for complex edits and 19 for quick fire and forget ones

Now the "cons":
In 2020 is unbealivable that a PC rig with super power RTX hardware have an idle gpu during rendering time.

I can understand that coding cuda's stuff is a pain, so i can understand that real time effects are addressed to the CPU, but during final render, the idea that we can not pass frames to the nvenc part of the gpu is crazy.

Do we (we in the sense of main kdenlive coders) not like to bind to the nvidia hardware? Ok use openCL and let kdenlive be a more flexyble piece of code.

"cit: OpenCL support is included in the latest NVIDIA GPU drivers, available at www.nvidia.com/drivers In addition to OpenCL, NVIDIA ... Download - Linux/Mac ."
"cit. ‘amdgpu’ is an open source Linux kernel module for AMD graphics. It is available in stock kernels starting with Linux 4.2. OpenCL support can be added by installing libraries from binary AMDGPU-PRO driver download, provided by AMD. "

OLD 18.* could be tricked to use nvenc by passing the correct parameter to the encoder, now the same trick is not working in the 19.* so i have to sleep the night to watch the final render of a 20 minutes 4k edit

cheers

Thanks, Viper, for chiming in with your ideas and experience and to all the previous comments.

Update; I did order the parts, built the machine, and have emerged from hardware and software hell (worst was getting MS Win10 Home to run on it--eventually succeeding on 4th try in VirtualBox with a DVD burned with K3b in Neon).

We are under lockdown in beautiful Santa Cruz near the University.

The new box is a Ryzen 9 3900X with Asus GeForce RTX 2070 Super Overclocked 8G EVO GDDR6 Dual-Fan Edition for GPU.

It may be a good while until I am up to doing real work at which point I will want to climb the learning curve more on Kdenlive and perhaps contribute from my experience.

For now, I will aim to make Zoom work well for family communication, etc.

Meanwhile, be safe and aim to live through the public health disaster brought on by starving public health capability for decades here and many places and the now unbelievable, even surreal, sociopathic non-leadership at the top here in the States (cutting off $$ for the World Health Org. today to blame shift).

I apologize if I am not harsh enough about it. It's astounding. Tens of thousands of lives lost here and counting and how many will die in Sub-Saharan Africa, rural India, refugee camps in the Middle East, etc. where health resources are so pathetic, etc.

If anyone is up to it it is a fitting subject for a dystopian horror video, but a real challenge it it keeps getting more awful.

Here in Santa Cruz things are relatively better because the SF Bay Area Health Officers, our County and our State all acted faster than New York, etc. But we just had a shuttle driver on campus die.

Stay safe. Same to the developers.

We are going to take a big economic hit globally as well....

I look forward to returning to the topic and the further interaction about the software.... Our efforts are important.

Thanks.