Ok..

I have a simple project on the timeline at the moment where I'm just playing with BCC beat reactor fxs, of course different fxs, media etc will affect a render time outcome but in reply to your 'times went up 4 or 5x when I zeroed it out'. I just did a few renders with Magix AVC..

5% - 25.82secs
0% - 29.86secs
5% - 24.65secs
0% - 30.30secs
I have it set at 50% because I use Build DR a lot, I have a lot of Ram & don't see any detrimental effect having it that high, so ..
50% - 24.54secs
Back to - 
5% - 24.60secs
0% - 30.49secs
 

---------

Interestingly if you look at my times above (ignoring 50%) - @ 5% the times decreased with each consecutive render but @ 0% the times increased 🤔

@Gid At the system level with 64 GB of RAM, the DRP setting doesn't appear to make a lot of difference with my work at render time. I did notice a difference in the cached memory between the runs from 5% to 50%.

Like any researcher looking for new grants, "more study is necessary". One consideration would be severely constrained memory configurations like 8GB or 16GB.

@john_dennis Thanks, so far rendering is mentioned but I'd like to know more about general playback without building DR, as I wrote 'poster's think increasing Dynamic Ram in Preferences improves VP playback performance', because I read comments like 'I've increased DR in Preferences but the preview still stutters..?' & @RogerS wrote 'It is a preview buffer used at all times' but there's no mention in the instruction manual of that, or I can't find anything.. VegasDerek wrote 'Unless you really understand Dynamic RAM Preview' but then didn't follow that up with an explanation or link,   . Do you or anyone think the amount of DR without building DR Preview makes any difference to playback in the preview window?.

A white paper on how dynamic ram works would be handy.

Yes I think it makes a difference in projects with particular timeline Fx / track blending that are cached in this memory. (You can search the forum for past discussions of DRP.) That's why the VP 16 benchmark required you to zero out the ram to do the playback test. Without that everyone's at full framerate playback.

That said its function (and ram usage in general, especially VRAM) has changed with the new video engine so I think it's necessary to do new tests. For decoding media I don't think dynamic ram preview is a significant factor so anyone with decoding issues should look elsewhere for a solution.

@RogerS Hi,, thanks, I've been on the forum since VP18 (I closed & opened a new account, don't ask.. 😂🤷‍♂️) & have read most of the DR comments, that's why I write this because the majority of answers are 'try this' or try that', It seems that nobody knows exactly what, why or how it works or if it does actually have an affect on general performance.. Although I do get the impression that setting it to 0% can be detrimental...

I find it hard to test any of this because of the specs of my PC, I've never had a need to set it to 0% & when I've tried it I see no difference, saying that .. the renders I just did above beg the question further.,

I have it set to 50% so I can Build DR on a full clip with fxs added, this quite often increases Taskmanager Ram to over 100GB 😱😂 but with 256GB total it makes no difference to my PC experience.

@RogerS Yep some official feedback would be nice.

I doubt official feedback is forthcoming so we have to reverse engineer what we can.

We know it's a preview buffer as creating a preview and then playing somewhere else on the timeline overwrites it.

I don't see much point in testing it as the default of 5% is fine. I make such ram previews so set it to 10%. There aren't any performance benefits that I can tell to going above that so why spend time playing with it?

The 4x slower reference was to the VP 20 benchmark where on my laptop times with VP 20 went from about 5-6 min to more than 20 min when I reduced DRP from 5% to 0%. Column N records dynamic ram amounts. Whether this project is representative of your own is a different question, I just consider it a data point.

@RogerS You added a ? but I'm not sure if that's a rhetorical question, I have it set at 50% so I 'm not limited to how much of an event gets DR built on it, below 50% if I set the range to the full event - Build DR it might stop part way through & not Build DR on the full event, that's just a personal choice & not about performance, it's about being able to play the full clip @ full fps.

These days I think decoding performance is more about pre caching in VRAM. Personally if I have an entire piece of media that's a problem I'll make a proxy, or a shift m pre-render that will persist. DRP I like for transitions and checking timing of layered events.

Feel free to come up with a test procedure and I'm happy to follow it here and report results.