Thank you for sharing your experiences! Super helpful!
I thought about sensible tests, and I’ve made a VDMX project here:
It’s about 8 GB. I explain why below. This is dumbed down version of my typical performance project that uses four cams in a multitude of layers:
Front masked overlay, 720p HAP Alpha
Front creatures 2, 720p HAP Alpha
Front creatures 1, 720p HAP Alpha
Artist cameras montage, 1080p, because sometimes zoomed in
-Artist cam 1, Blackmagic Atem Mini 1080p
-Artist cam 2, Logitech C920 webcam 720p (or lower if needed)
-Artist cam 3, NDI network cam 360p
-Artist cam 4, NDI network cam 360p
Back creatures, 720p HAP Alpha
Background, 720p HAP Alpha
Insertions for screens in background, 720p HAP
The dropbox has a simplifed version of this, including test HAP media, total size 8 GB. The media is just multiple transparent videos filled with noise with titles and timecode running. The reason for 8 GB media is because the stresstest is to force VDMX to read unique individual files with lots of information, and alpha layers to see them all. The project on average loads six of these in many scenes of different montages. Press any random key to load a “scene”. It looks like a ridiculous mess in this project but the point is to test responsiveness and CPU cost.
For sure keep the video files on internal or external SSD disk. The project will autofind them. Don’t panic when it claims hundreds missing,its just the same files used multiple times. Add your Facetime cam to the input boxes upper right part (the input sources are available in the workspace)
Without cameras, or with only the internal Facetime camera, my 2016 MBPR system can load all these pretty much instantaneous. You can tell, by seeing the scene load without anything coming in “late”, and you should see timecodes are pretty much always in sync. For me VDMX uses 80-100% CPU, VTDecoderXPCService about the same, of a totalt available 400% CPU (4 cores). VDMX output framerate is 60 fps when running a scene, when loading a scene, 30-45 fps. This is all good.
Performance with cameras:
Then, my problems start when I add cameras. When this project uses four cameras, the CPU use increases massively, and the latency of loading of scenes takes a serious dip. Then, I can very very visually see each scene being “drawn”, it takes sometimes up to a second for all clips, layers and positions to load, and the illusion of a “cut” is broken, the audience can see the scene being built. I also have problems with clips drifting out of time, masks then being out of sync, if multiple layers are built with synced movement. VDMX goes up to just under 300%, and the rest is completely taken by VTDecoderXPCService, so system at 400% constantly. Framerate down to between 15 and 25 fps. So, doable, but under duress and I have frequent crashes. This is not good.
I also have zero issues with camera frame rates when using one camera. But when using four cameras, frame rates drift, and particularly USB 3 cams seem to lag randomly when sharing bus, even with just dumb UBS devices like a mice, that uses pretty much ZERO bandwidth.
Improvment to this loading and weird framerate drifting is what I hope that a new 2020 MBPR model with the 5600 graphics card will help me - probably because it is just generally “stronger” in total, but i’m not sure!!! Maybe the issue is actually somewhere else!
So if you have multiple cameras you can somehow connect, and see how the project responds THEN, that would be super helpful.
I am probably going to go for the computer anyway, because of the general gains I will get from a model that is 4 years newer than my old one. I will for sure get a lot of boost when prepping clips and rendering the videos before performance. But it is such a immense investment for me, I kind of seek some extra data to emotionally convince myself that its going to be worth it!
Thanks if you read this far and have time to test this!