DS HAP vs QT HAP in Windows - some findings

jtsteph

Hello,

I have been meaning to do a proper test of HAP DS vs HAP QT in windows for some time now. I did some extensive tests years ago and had determined that QT was the way to go. I recently redid some cursory tests and think that I fell into confirmation bias. I thought I should undertake a more thorough test aligned with my current project. There is a link below to a ZIP of my test framework that is a lite, mini version of a project I am working on.

TESTING FRAMEWORK

- 4 clips, 640X480 (I have a decade+ of material in this raster)

- The clips are combined in pairs with an ADD actor into two projectors (in my show I switch between ADD and a third party FFGL ADD ALPHA CHANEL)

- Each clip has two versions: a HAP AVI and a HAP QT

- Stage is set to 1280X720

- Each projector ZOOMS to 138%

- One clip is running in interactive mode with a sine wave generator determining time position

- Actors have been minimized to minimize gui updates.

- The 'a' key switches between the QT and the DS AVI. "play natively" is on in the preferences

- I tried both 60fps 4X and 30fps 8X in the preferences.

FINDINGS

- I have always, in the past, run my projects at 60FPS and 4x Per Frame. I had poor results with this in my test. The project cycle count would dip down to 60 and then back up and then back down again in unpredictable ways. When I switched to 30x8 in the preferences, Isadora held consistent cycle counts. My rational in the past was that I would be fine to dip under 60fps but not under 30fps. Why are the cycle counts more consistent with 30 8X vs 60 4X?

- the HAP AVIs were quicker - solid 240 after things settled when switching from the Quicktimes to AVIs

- The HAP QTs were a little slower - 230-238. Not anything dramatic, but measurably slower. Multiply this with live audio inputs and more effects, I'm certain that this will make a difference.

- The HAP AVI in interactive mode behaved perfectly fine. In the past this was not at all the case. I assumed it to not be the case again now, but I was incorrect.

- Testing machine specs: W10 Pro, 8C/4ghz, 64GB ram, GTX1080, NVMe Storage @1500Meg/s

- I will repeat these tests on my laptop off a single SSD (below) in the coming days.

In any event, I am sharing my test project and media here. The footage is Copyright(c) Justin Stephenson 2018 and is only to be used for testing purposes.

I would be curious if anyone has insight into the frame rate/cycles relationship in the preferences.

https://www.dropbox.com/s/htx1...

DusX

Thank-you for sharing your findings. Out tests have also shown that Hap Avi performs better than Hap QT.
Its nice to see that you have similar findings.

Regarding your question about the difference between 30fps + 8x cycles and 60fps + 4x cycles
the biggest factor may be your display setup. Please take a look at this article https://support.troikatronix.c..

If your video cards supports it, you may want to try to enable triple buffering, this may have an effect.
If you do test it, please let use know how that affects your system.

jtsteph

@DusX I just ran a quick test of the buffering. Double buffering @60fpsX4 is as reliable (60fps w 240 cycles solid) as single buffering @30fpsX8. I do not see an option for triple buffering, though for this project a 3 frame delay would not be ideal - isadora's fabulous lack of latency for audio reactive visuals is key to this for this work. Two frames @60fps is still quite pleasing. Thanks for the tip.

Best,

- J

DusX

@jtsteph

Perfect.
Mark had explained to me why additional buffering may help, so its great to hear that its working for you.
I will do some testing in Isadora 3 with the same setup. I expect with some of the improvements that have been made that performance will be even better.