MIDI Scrub develops issues with upgrade
I had a nice clean break between projects so I upgraded to X1 Producer from the latest 8.5x Producer. I did not realize there was an a and b upgrade. I did a whole project (about 2.5 min of music for a video in surround). It was stable, I generally liked the enhancements when I got used to them.
I use the East/West libraries, mostly Hollywood Strings, Orchestra Platinum and Symphonic Choirs. I use True Pianos to create a scratchpad track, then mostly hand-draw orchestration in using PRV. I'm 64bit Windows 7 Ultimate on a fast 8 core. I use the onboard soundchip (getting rid of this is high on my wish list) on the intel motherboard. My tracks go as surround wave-files to the video workstation where the final mix takes place.
I decided to upgrade to X1b because the snap bug was frustrating and the consensus was pretty good that more got fixed than newly broken. But I have a problem. A lot of times individual voices disappear in a way that reminds me of note-stealing when I had less powerful hardware. I'm convinced I'm not out of resources, because I had none of these problems on X1 and was doing things of very similar complexity with some of the very same voices.
Here's more detail:
For instance, sometimes I will click on a note in the PRV in a way that should cause the virtual synth to cause the audio engine to generate the note right then and there. I'll get about 300 milliseconds of the note, but that's all. When I release it, sometimes I'll get the release trail, sometimes not.
It is very problematic when I scrub. In X1 using the scrub tool, I'd click (I actually use a wacom tablet) on a chord in the PRV and I'd get all the notes every time. Now when I do the same thing in X1b, I rarely get all the notes, sometimes I get one or two of the notes, and most of the time I get nothing.
I've played with the synth settings on note limits and overload to no avail. Again, because this worked fine in X1, I don't think it is anything like note stealing. I think the audio engine just has problems understanding what its task is in this circumstance and stops generating samples but doesn't think a note off has been received, because it often produces the trails on the actual note off (by either lifting the pen or scrubbing over the end of the note).
When I play the tracks back normally, I very very rarely will get a dropped note. When I export to audio file I don't believe I get dropped notes at all (though I may, and just haven't noticed one).
I see other people have reported things similar to this, but not quite exactly what I am experiencing. I'm able to work this way, but I don't know if getting the snap back was worth losing the surety that the notes would play.
I saw some discussion about changing the buffer size or number of buffers; if I try to change buffer numbers from 2 I get a dialog box on save that tells me to enter a number between 2 and 2. I am running with a lot of latency, which is another thing I discovered between the versions. I could work with far less latency in X1 than X1b--something like 50 ms. With X1b the system requires 188 or more msec to give me a stable output, which suggests that some change was made in the audio engine between the two versions that doesn't favor the flavor of my work environment.
I can do some limited testing, but I'm halfway through a 5 min video that has a deadline, with two more after that, so I am afraid to try and roll back to X1. Can anyone suggest anything, even if it just lets me identify precisely what the problem is?