Horrible VI performance when recording in 2019.10!

[TNM]

this post was incorrect..

[JFreak]

Before I get excited please tell me why on earth would I want to record 20 VI when I only have two hands to use them

[musicman691]

Originally Posted by JFreak

Before I get excited please tell me why on earth would I want to record 20 VI when I only have two hands to use them

Exactly.
I often have a dozen or more vi's running at the same time but only recording with one at a time. All others are running with already recorded MIDI. Combine this with freeze/commit and there shouldn't be a problem. The closest I ever had a problem is when I was working on my own version of Tubular Bells when I had 50+ tracks about half of which were unrendered vi's which were a combo of samples based (Kontakt, EW Play) & computational based instruments (aka softsynths like Arturia Synclavier). Overall cpu usage as measured in PT 2019.6 never went above 50%. Period.



What helps is I use low-rent plugins to do the recording and then replace them with whatever synth/sampler I want to use. Recording I can get away with 128 and even 256 samples but when I get to mixing I max out at 1024.

[Carl Lie]

Whenever I install a new version of PT I delete my preferences and run disk utility.

Maybe it will help.

Carl

[TNM]

Originally Posted by JFreak

Before I get excited please tell me why on earth would I want to record 20 VI when I only have two hands to use them

I guess you don't do orchestral..

I often have 20 kontakts activated at a time and all on different patches for performances.. sometimes more than 20.

What on earth does that have to do with testing the performance of pro tools anyway? Who cares whether not many people arm many VI's at once?

The poorest results were in playback, that means, NON armed tracks.. PT performed OK on armed tracks at 128 buffer.

On 32 and 64 buffer i have issues with ONE vi in a brand new empty project. I can not play one with a keyboard.. What, you don't arm ONE vi or you think that's weird too? I can not play a single kontakt for example at 32 or 64 buffer in PT and never have been able to, ever, for almost 4 years now. In Logic and Cubase it's plain sailing to do so.

Sound like you missed a lot of points I made.

To the musicman, my projects often consist of 128 VI tracks.

People just don't realise the sort of track count detailed orchestral work requires.. Sure, if I am just doing some pop or something, then 32 tracks is often enough. Just depends on the project.

Check out DAW Bench.. It's called benchmarking.. it does not mean that everyone is going to need the amount of VI voices being done in the tests, or over 1000 effect plugins.

[TNM]

Oh and YES, I have used 32+ sylenths in a track that was made purely of sylenths.. drums and all.. of course cause it's a stereo synth, each drum had it's own instance which used up a lot of those.
And yep, the drum patches were all armed at once when I played the drums.. The reason they were all on separate instances was because I wanted to be able to have total control at the mixer level for each and every drum sound, including FX.
You hit record, and all the midi places itself perfectly on separate tracks therefore separate mixer tracks.

I also barely ever use the multi outputs of synths.. I use stereo instances (or mono where desired) because everyone knows that with Kontakt, for example, you get way better performance not using the multi timbral/multi out mode.

Anyway, I have installed Mojave and PT 2019 on the imac pro, and just like the macbook, it can't deal with buffers under 128 and VI's.. This is with the apollos rather than the in built apple driver.. In Logic I can play ANY synth on the imac pro at 32 buffer, even at 96 kilohertz (which is like 0.8 of a millisecond output) and of course 44/48K. In Pro tools I have to set it to 256 buffer at high sample rates.

Every time I ask or talk about this, and ask what happens when people arm a single VI track at 64 or below buffer, NO one actually answers that part, or take a pic or anything to show how their system is performing.. They just ignore it and throw comments like it must be a system error or something. Yeah, both my computers are faulty, and i am sure the new laptop coming will be too. BTW, the imac pro was formatted and mojave was installed fresh and then pro tools immediately and then the latest ilok drivers, Apollo drivers, etc.

So will anyone PLEASE show me how their mac is performing at 32 or 64 buffer arming just ONE VI track in a brand new, empty project? So i KNOW for sure if it's my computers or not?
It will take you 10 minutes total to do, and i'll pay you for your time, whatever your normal rate is. I asked this last time and no one would do it.

Cause they know I am right, and there is something seriously wrong with core audio and PT at low buffer. The only person who did answer me once was at GS, we exchanged chat and videos and he showed his mac pro giving exactly the same issue, but windows being much much better when using low buffers. But that was three *years* ago and I was hoping 2019 would have addressed something.

No one has actually provided ANY valuable info whatsoever, ANY productive talk, NOTHING that's on topic.. all they are doing is criticising and basically saying who cares... Only southside music at least said he wasn't having issues but then did not answer further when I asked if he tested PT 2019 with sierra on a macbook.. The whole point was whether to risk a working imac pro and install the newest versions or if i was going to get WORSE performance.
THANKS all for the typical patronising comments and not one bit of actual on topic assistance.
Anyway, I took the risk and installed it, so now I have to open some projects that maxed out the previous version/high sierra, and see if they play back OK. That will answer me myself.

[musicman691]

TNM:
Thing is you're not actually recording 128 vi tracks at once. You may be playing 128 vi tracks at once but then you don't need to run at low buffers. Max out the buffers and be done with it. You only need to run at low buffers if/when recording.

People have provided realistic info for you but due to cognitive dissonance you don't or can't or won't see that. I'd trust what Christopher (Southside) or Janne say any day of the week.

Have you actually tried freeze or commit? I'll bet you haven't.

[TNM]

Ok, I am bumping this because I have done some serious real world testing, and going to give it a shot at getting it back on track.. I myself am to blame as I had a very short fuse in this topic, so for that I apologise.

Now, what I have done are not "20 record armed VI's" and so on.. I have tested real world projects that actually exist of Vi *PLAYBACK* and also monitoring external synths through PT"s mixer at low buffer..

what I have found is, that 2019.10 has about 25% lower performance overall than PT12, which, at 128 buffer, exceeded Logic's performance.. Now there is no DAW that can even get close to Logic for efficiency, well, except, sort of, Reaper, but it uses a lot more real CPU in the process.. that said it came closest to matching track counts and VI polyphony..

So.. forget recording, forget 32 and 64 buffer.. I am talking directly in comparison with Pt 2 years ago on a 128 buffer at 44 or 48K and Vi tracks that are playing back already recorded midi, not VI tracks that are armed for record.. *playback* tracks which reside on PT's internal 1024 sample hybrid buffer.

All I can do is send a report to support and hope for the best that they'll fix it..
I absolutely can not play back heavy songs on my older macbook that I wrote in sierra/PT12, in HS PT 2019.10..

With PT 12, as long as no tracks were record armed, you could take PT's meter to about 90% solid, with occasional peaks to 100.

Now, it starts crapping out at 70% on the same computer and has crazy peaks to 100% in one jump from 70%.

[Darryl Ramm]

Originally Posted by TNM

With PT 12, as long as no tracks were record armed, you could take PT's meter to about 90% solid, with occasional peaks to 100.

Now, it starts crapping out at 70% on the same computer and has crazy peaks to 100% in one jump from 70%.

Uh metering changed significantly in Pro Tools 2018. Earlier problems seemed to include falsely high CPU meter values... which might be what you were seeing. But what does some synthetic meter matter? What matters is can you find a workflow that works for what you need. If you are having problems I would *not* focus on these meter differences.

Personally I would never want not want to go back to issues with Pro Tools 12 and (general not VI necessarily) plugin performance.

[TNM]

Originally Posted by Darryl Ramm

Uh metering changed significantly in Pro Tools 2018. Earlier problems seemed to include falsely high CPU meter values... which might be what you were seeing. But what does some synthetic meter matter? What matters is can you find a workflow that works for what you need. If you are having problems I would *not* focus on these meter differences.

Personally I would never want not want to go back to issues with Pro Tools 12 and (general not VI necessarily) plugin performance.

Um.. PT 12 and 2018 could use more than 85% of the total REAL cpu power available in the computer (for example, I could get it to 680% or even 700% out of a possible 800% on an 8 thread machine), and now on the same machine I can get no more than around 545% of real cpu usage consistently. I literally have projects here that are so heavy with plugins and PT is playing them back without stopping on the sierra/PT 12.8.4 combo and the real cpu is as above..

I had no idea there was a faulty performance meter but that doesn't change anything at all, and you've completely misread my comment and focused entirely on the performance meter aspect only. That comment being, that PT 2019 can run about 25 to 30% less plugins (or less VI overall polyphony) than Pt12 and the early builds of PT 2018 could. That's the entire point I am making and what to take out of this. So I am talking about real world, real performance.. as I said, projects in previous PT versions that truly maxed out this cpu that played perfectly, will instantly have a dropout and refuse to playback in 2019.
Of course, they WILL work on my imac pro using PT 2019 as that's twice as powerful as the laptop they were initially created on.. but the same logic applies.. If i were to create a song on my imac pro that maxed out a previous version of PT but could be played back, PT 2019 would not be able to do it.

There is currently zero slowdown with PT2019 on this old laptop when PT is crapping out.. what I mean by that is.. zero sluggishness or unresponsiveness.. it feels like there is still so much headroom.. With PT12 i could use so much of my real cpu that things became sluggish.. I am telling you I even saw 720% out of 800% once.. that is unheard of for a DAW to be able to use that high of a percentage of a multi core cpu.. but it STILL played back every time i pressed play. Anyway, the fact PT2019 is overloading when the system is still super fast and responsive is also an indicator that it just can't tap into the cpu power that previous versions could.

I never had issues with third party plugins other than the midi reload bug which is still there in 2019, so I have no idea what you are on about to be honest.. I did have some stability issues with Avid's own plugins in some releases at high SR's but that was all fixed quickly. I am telling you, I am 100 percent certain, that PT 2019 has had a drastic reduction in how much of your system resources it can take advantage of, vs earlier versions...

You can even skip all that text and just use the punchline.. projects that used say even just 600% of my 8 thread cpu that played without slowdown or without a hitch in PT12/2018, will NOT at ALL play in 2019.. It can't do it. I don't know how much clearer I can make it.. I'm going to do some tests on bootcamp to see if there's been any changes there too on the windows version.. but right now, the performance loss on mac is massive.

I think i even have video somewhere of PT using over 700% of cpu but I'd have to find it.. as I was so impressed by it's performance. it's low latency performance was always terrible for me personally (by that I mean 32 or 64, 128 and up was always great), but once all your tracks were recorded and you were at the mixing stage and there were no tracks armed for monitoring or recording, PT 12 was the most efficient DAW I have ever used in my life, eclipsing even Logic.. It could get more VI polyphony and more FX instances when using the same plugins and use more of the entire cpu overall.

I have no idea how you just chose to focus on the meter differences (which I did say, true, as I thought it was related), and think thats all I focused on.. It's like you didn't read the meat of what i said at all.. that previous projects can no longer play back.

For this entire reply I have been talking in real cpu terms.. activity monitor/istat.

I already have preliminary results that PT 2018 on high sierra greatly outperforms PT 2019 on the same imac pro on Mojave.. once again about 25% less plugin instances capability overall.. so.. that's two macs with two different OS now.. I can't test the 16" cause it only uses Catalina..
However I am getting some reports filtering in from GS users that they can get way less plugins now in 2019 than they could before also..
I can absolutely guarantee you that the performance cap has drastically dropped in 2019. It's still way better than Studio One 4.5, and about right on Par with Cubase 9/10 and Asio guard.

The highest two performing DAW's on the mac overall, regardless of what internal buffering they do or don't use, are Logic followed by Creeper (sorry I just hate it). Reaper of course is it's real name. However Logic can get around 10% better overall performance, when real cpu usage is around 85% of the cpu. Reaper's meter is the only DAW i know of with a real cpu meter, and the anticipative buffering approach it takes uses a ton of CPU.. in fact.. Reaper can get to around 97% real cpu usage on playback.. I have never seen a DAW do such a thing.. So reaper was at 750% usage in the 8 thread, but running 10% less plugins than Logic at 600%. For laptop users that should be taken into account for heat and noise.. That said, I am overall very impressed by what Justin has managed to squeeze out of Reaper as far as performance goes.

In order of overall best performing/highest plugin count DAW on mac, just 2 years ago it used to be:

PT/Logic/Cubendo/DP/Bitwig/Ableton/Reaper/S1
and is now:
Logic/Reaper/Bitwig/DP/PT - Cubendo equal /then Ableton/and of course S1 last again.

The best performing DAW by a LANDSLIDE that doesn't use any underlying "cheat" mechanism or hybrid buffer is bitwig.. if i set bitwig to 128 buffer, that's it, it's 128 buffer from A to Z.. record armed or playback tracks don't make a difference.
It can match PT 2019's performance when PT is using a 1024 playback buffer.. so that pretty much says everything also.. That shows you how poorly PT is performing on playback when it's using an 1024 buffer.
If they dropped the hybrid buffer or something, to say 512 or even 256, then I can understand it! So if there's anything you can tell me that they changed purposely that may account for this, please do.

Avid coders have contacted me many times over email and everything they told me would be fixed by end of 2017 is worse than it was.. I have said it before but I don't think the current coders know HOW to fix things.

PS I don't know of any individuals other than the pro daw builders that compare CPU's to one another, who has ever done DAW vs DAW performance comparisons on the SAME hardware as much as I have.. I have done at least 1000 hours of it in the last decade.. I have 100's of dedicated streamlined DAW efficiency test projects..
But this is even a bit more different because this time PT would not play back real projects that it previously could (and easily).. it wasn't one of my standardised DSP tests.

That said, a standardised DSP test is the easiest way to point it out.. I am currently in the process, in collaboration with a few users at gearslutz covering the DAWS I don't own, designing standardised DAW performance testing projects using entirely freeware plugins.. It will enable anyone to simply load the project and compare performance capabilities of different DAWs on their computer(s).. Think of it as DAWBench but for DAW's rather than CPUs! All one will have to do is load the project.. if it can play back, they keep duplicating tracks till they get the number of tracks the DAW can do.. If the default project is too heavy for their machine, they delete tracks till it can play back and get the number that way.

there will be 4 tests:
VI playback polyphony
VI record armed low buffer polyphony
Audio track record armed with reverb monitor busses track count
Audio track non armed playback count with a dedicated realistic FX layout on each track (EQ/Comp/one other effect I haven't decided yet).
I'll have to find a very heavy cpu effect to include on each track so as to not exceed maximum track counts in DAWs that have it (ie PT).
The reason I brought up this new upcoming benchmark suite is that it will also be entirely suitable for comparing the performance between different versions of the same DAW on the same computer. Anyway it's a lot of work so it will still be some time, but I hope to have it completely done by Christmas for all known major and not so major Daws, and the priority will be to initially focus on OS X Daws. (Logic/PT/Cubendo/S1/Reaper/Tracktion/Live/Bitwig/DP/FLStudio/MixBus/Reason).