Tracktion is a great audio sequencer, with an extensive user base on PC (the Mac versions never worked terribly well, by all forum accounts).
Here are some performance tips I've put together over the years:
The exclamation mark shows when tracktion isn't able to get the audio data from your disk to the mixer fast enough.
Upping the cache size under tracktion to 150MB helps when mixing a large number of audio tracks (obviously this will do nothing for VSTi tracks).
Tracktion 1.6 has a very CPU-efficient algorithm for internal resampling of audio files while mixing,
unfortunately they dumped it in Tracktion 2 and replaced it with one which doubles the CPU-usage of resampling when you try and
mix audio at sample rates other than those of the given audio file.
So, if using T2 or T3, you're better off to be mixing at the sample rates of your audio files and to not mix sample rates.
Tracktion 2 & 3 have very slightly better overall sound quality (particularly with 64-bit track mixing turned on in options) and are easier to do many editing tasks in, however due to the above notes in point 2 it is more difficult to save cpu by mixing at lower sample rates (for audio - not for VST instruments obviously).
If you're trying to mix something but are getting heavy CPU usage, take the project into T 1.6 and drop your working sample rate to 44k. Suddenly your CPU-usage is more than halved. It's easy to do the basic mixing in this way then step up to higher samplerates in T2 later for perfectionist tweaking.
Since the freeze feature in tracktion is pretty much broken & too slow to use (should be per-track, should store freeze files in project directory, etc, etc, etc) don't bother with it.
When you're getting overwhelmed with CPU, select one or more tracks that you don't need to work on and render them into the project, then mute those tracks and move on. Render at 32-bit, and the sample-rate that you're working with, as 32-bit is what the internal mixer works at anyway,
and if you use a sample-rate other than the one you're working with, Point 2. from above kicks in.
Okay, so this one isn't 100% tracktion-specific, but the plug in question works really well under tracktion - anyway, If you have the time and the money and an extra machine you're not using (or using intermittently), the FX-teleport plugin from http://www.fx-max.com/fxt/ can save a heck of a lot of CPU and time.
If you're pushing it you can save a (very) small amount of CPU by deleting the VU-meter filters and of course any other unnecessary plugins in your tracks.
Not that everybody doesn't do this already, but the tracktion built-in plugins (in T2 at least - not in T1) don't work as CPU-efficiently at higher samplerates than they do at lower. Note I'm not meaning the mackie ones like the side-chaining compressor etc.
Also, obviously find out which of your other plugins are CPU-hogs and replace them with onces that ain't.
... and stick in at the front of one of your disks. Disk access is 35% faster at the start of a hard drive than it is at the end - HD's read outer to inner, outer rotates same speed but has more circumference, therefore more data, hence better transfer speed.
If it's small enough, make it FAT8 or FAT16 - 16k or larger cluster sizes are best. This will speed up access and make tracktion bootup plugin scans a lot faster.
Course, you could just turn off the plugin scanning feature if you wanted to.
In modern ensembles there's often a lot of audible separation desired between the many tracks. You achieve better separation if you use separate effects on each track - even if they're the same effect with the same settings - rather than routing all the tracks requiring that given effect through one 'send' effect channel.
However if you don't mind (or are in fact going for) a more "grouped" sound between a bunch of instruments, you can achieve this and also save CPU, by routing them into one effects channel.
Take my word for it, or try it yourself- I don't mind-
In a more 'traditional' sequencer you would do this with a 'send' effect channel-
while in tracktion you achieve this by setting up a new track with the filter(s) (effect(s)) required, then changing the output of the tracks you want 'effected' to this new track.
Tracktion 3 has a number of bugs which are fairly well documented on the KVR forums, but two performance bugs are workaround-able:
The logging bug - Appears to be a developer debug feature that was left on by mistake - T3 logs any and all activity, creating superfluous CPU disk access (to the log file). Workaround by making tracktion's log file read-only.
The VSTCache issue - T3 appears to implement a linked-list of sorts to access plugin's, increasing the access times for plugins at the end of the list exponentially. Workaround by putting all plugins in a single directory and renaming the dll's of the ones you want fastest access to with a "z" at the front (VSTcache is reverse-alphabetical linked list). More details.
Just to be clear, neither of these bugs exist in T1 or T2.
IF you are sure that your plugins do not denormalise based on the incoming signal (as opposed to denormalising internally, which cannot be prevented by tracktion's built-in anti-denormalisation feature), turning off Tracktion's "Add low-level noise to avoid denormalisation" option (plugin's menu under settings)
can give you a minor performance boost. For whatever reason, the algorithm which generates the low-level noise is somewhat CPU-intensive once you have a large number of tracks.
This can be tested by setting your ASIO drivers/tracktion to 96000, creating an edit in tracktion with 20 separate tracks of 24bit 96khz audio playing concurrently, then checking the CPU usage while the anti-denormalisation option is turned on, then off.
On my 3ghz Pentium 4 computer Tracktion's CPU usage (using the Windows Task Manager to check) drops from an average of 10% to an average of 5% with anti-denormalisation turned off.
Obviously this creates a problem when dealing with large numbers of tracks - 20% of CPU usage could easily be consumed by the anti-denormalisation feature in an 80-track project (on my 3ghz machine).
Though of course you might want to experiment with an existing edit, just in case you have plugins which denormalise easily.
From experimenting with this on a Core 2 Duo 2GHz processor, the difference was more like 1-2% CPU (not worth worrying about).
See my brief audio partitioning guide here.
All advice is given without guarantee- use your brain - and if anything explodes, see a doctor (don't talk to me).
If you've got anything to add, my contact email can be found here.
-matt
Back to the index