I ran into an interesting problem yesterday!
 
My main DAW is an Intel Q6600 @ 2.4GHz with 8GB of RAM, running Win7 x/64. While I do sometimes run up against resource problems, I can usually get away with increasing latency to mix, I haven't had to freeze a track in ages.
 
So yesterday I imported a MIDI only project for a class I am taking. The MIDI only uses 273 MIDI tracks. There is only MIDI data in a handful, and I had not yet loaded any instruments, but I wiped out my physical memory. The good news? Well it did not crash, but I had to wait quite a while for the swapping to take place.
 
I have, in the past, had projects that used up to 60 MIDI tracks, and I never had to pay attention to how much memory was used, and I suppose now I need to go back and load some of them to see.
 
In the meantime, are there settings I need to tweak to support this many MIDI tracks? And why do MIDI tracks use so much memory if they are not connected to instruments?
 
edit: for grins I loaded the same MIDI file into Studio One and it required only about 3.5GB - which is why I am pretty sure I'm doing something dumb in Sonar...
 
Thanks!

Can you post the MIDI file (273 tracks seems a lot!) for others to test?

I'm not sure - I will check with the instructor. I created several test cases yesterday, and I can post those while I wait for an answer.

So ... With a file size of 3.5 gb this can't be just midi data, can it? There are no sample libraries that are being loaded? If you save midi only in Studio One what is your file size?

let me provide a more detailed view of the situation...
 
I ran a test tonight, loading Sonar till it became unresponsive, unstable, or crashed. The following represents the minimum and maximum memory loads and they correlate quite well with the starting condition, which varied a lot more than I expected. I ran the test five times.
 
I rebooted the computer, memory used was (min) 1.26 GB, (max) 1.98 GB
I launched Sonar, memory used was (min) 1.48 GB, (max) 2.47 GB
then I started adding empty MIDI tracks:
 
25 tracks    1.67 GB    2.63 GB
50 tracks    1.85 GB    2.82 GB
100 tracks    2.21 GB    3.20 GB
200 tracks    2.97 GB    3.95 GB
300 tracks    3.74 GB    4.69 GB
400 tracks    4.50 GB    5.38 GB
500 tracks    5.26 GB    6.19 GB
600 tracks    6.03 GB    6.96 GB
700 tracks    6.80 GB    7.73 GB
718 tracks    6.93 GB    7.86 GB (became unresponsive)
800 tracks    7.42 GB    
900 tracks    7.66 GB (became unresponsive)
 
In comparison, the test project (which I mis-typed earlier) contains 718 tracks!
If I import the MIDI file into an empty Sonar project it uses 7.36 GB (min) and crashed at the max case
If I drag the file into Sonar with no project loaded it uses 7.18GB (min) and 7.88 GB (max) at which point it is very (*VERY) sluggish.
 
I'm not sure what other information might be in the test file, other than a few odd bytes of actual MIDI data.

So the question has become, for me, why can Studio One manage a large template, but Sonar can't. I still think there must be a setting I need to change.

I am going to submit this to the bakers, along with the actual test file. I have not heard back from my instructor yet as to whether or not I can post it here.
 
Thanks for any and all thoughts!!

KPerry
Can you post the MIDI file (273 tracks seems a lot!) for others to test?

I have asked permission to post the track here - will do so if they say it is ok.

As far as 273 tracks being a lot, it's not really, and it is about 1/3 of the actual file, I have no idea how I mis-typed, but the track count in the test file is 718 MIDI tracks.

This sort of template is used when you have a bunch of large-ish sample libraries and you do not want to spend hours loading the template so you instantiate the sample players (Kontakt, VSL, Play, Engine...) and load them and leave them loaded. Depending on the size of your template you may also move some of the sample players off to other machines (often referred to as slave machines). The template for the example project above runs on seven slaves.
 
If you have the requirement, and the money, it is a very nice way to work... sorta. Several years ago many of us had to keep a slave machine around just for GigaStudio, and it was a real treat to finally get back to a single machine. But now it looks like I am headed back to multiple machines. On the plus side, tools like VE Pro and Bidule make life a lot easier this time around!

I don't know where your memory usage numbers are coming from. I just loaded an orchestral template with 83 tracks, 16 buses, and 9 instances of VEP: <350 mb memory usage. Just out of curiosity, I created a project with 800 midi tracks.Total Sonar memory footprint: 204mb. So ... About .2mb memory usage per midi track(an empty Sonar instance is ~ 50mb). You have something that is sucking-up memory in your template and it is _not_ midi.

bentleyousley
I don't know where your memory usage numbers are coming from. I just loaded an orchestral template with 83 tracks, 16 buses, and 9 instances of VEP: <350 mb memory usage. Just out of curiosity, I created a project with 800 midi tracks.Total Sonar memory footprint: 204mb. So ... About 2mb memory usage per midi track(an empty Sonar instance is ~ 50mb). You have something that is sucking-up memory in your template and it is _not_ midi.

MIDI should never get into the GB area. I have yet to encounter a MB size MIDI file. Also to the OP don't import MIDI files load them. MIDI files are a native file format for Sonar.
 

Bill can you say what is in those files?

wst3

KPerry
Can you post the MIDI file (273 tracks seems a lot!) for others to test?

This sort of template is used when you have a bunch of large-ish sample libraries and you do not want to spend hours loading the template so you instantiate the sample players (Kontakt, VSL, Play, Engine...) and load them and leave them loaded. Depending on the size of your template you may also move some of the sample players off to other machines (often referred to as slave machines). The template for the example project above runs on seven slaves.

This suggests it's *not* MIDI only but includes soft-synths, so I'm not surprised the memory usage is so high!