Synchron-ized Dimension Strings: CPU and system advice

[thomasoehler]

  Hello,

I was wondering about the demands of the (by the way wonderful) synchroni-zed Dimension Strings on CPUs. I would like to experiment with them in the context of complex all-out divisi and possible layering, but I quickly ecounter limitations.

on a 2018 mac-mini 3.2 GHz 6-Core i7 Coffee Lake Mac with 32g of Ram, (512 I/O buffer size in Logic) I can run a maximum of about 9 different simultaneous voices (with no other libraries loaded). Beyond that number, it starts stuttering as the CPU load is too high. I can push that to about 13 voices if I use the light convolution option.

- Is that in line with what would be expected with such a system?

- I am possibly thinking of getting a dedicated slave system to run only the Synchron-ized Dimension Strings. What kind of specs would I need to run about 15 simultaneous voices smoothly without worrying about CPU overload (with full-on convolution)?

Any advice greatly appreciated! Thanks!

[bbelius]

Hi Thomas,

Make sure you run the Dimension Strings from an SSD and increase the preload size higher then the default value that is set by the player. Also make sure the SSD is not fomatted with FAT/exFAT.

Best, Ben

[richhickey]

@thomasoehler said:

I was wondering about the demands of the (by the way wonderful) synchroni-zed Dimension Strings on CPUs. I would like to experiment with them in the context of complex all-out divisi and possible layering, but I quickly ecounter limitations.

Any advice greatly appreciated! Thanks!

It matters quite a bit how you are organizing your use of synchron player(s) (SP) and dimension string players.

I presume when you say 9 'voices' you are talking about dimension instruments/players? VIs typically use the term voice for simultaneously playing sample, and if you look at the voice count at the bottom of SP you'll see dozens of voices used even for single note lines.

Presuming you are talking about dimension players, you have a number of choices re: e.g. which preset (e.g. all players vs desk etc). By far the easiest way to work with SynDS is with the 'all players' patch, even if you only want to use a subset. This is because you can quickly create dimension tree edits and mappings for all players rather than have to replicate your work as you had to do in VIPro, and quickly try different player combos.

But by far the hardest thing for a single SP to do is run all 8 DS players.  

I'll presume you have an SSD, and you are running SPs in VEP (from your tags), if not then the DAW performance is another factor.

What you want to do is:

a) (Continue to) use the 'all players' patch for convenience in editing, selecting players etc. It is, IMO, the single biggest value prop of synchronized DS.

b) Use more/many SPs - VEP is much better at multitasking than is a single synchron player.

c) In any particular SP, activate (see below) only a subset of the players, e.g. no more than 4. You can setup another SP listening to the same channel in VEP for the other 4. 

d) In any particular SP, make sure you are not using CPU for resources you don't need (i.e. when you only use a subset of the all players)

This last part is subtle and here are two tips:

1) If you are using MIR and not the built in synchron IRS, don't merely disable them - remove them. Removing IRs releases the CPU

2) Don't just mute any players you are not using (or the reverb), deactivate them by clicking on their name at the bottom of the channel in the mixer. The name will go grey and italic. Deactivating releases the CPU burden in a way that muting does not.

In this way you always have the convenience of having all players at hand (vs having only a subset in the tree/mixer).

I had a lot of dropouts with synchronized DS until I adopted these practices after which I've had no dropout problems (using VEP on the same box).

Good luck!

[thomasoehler]

Dear Richhickey,

Thank you SO much for your detailed advice. I was indeed going about it the wrong way, loading "all pIayers" patches like a brute in each SP and keeping them all on in the mixer...

I have followed it very carefully, and indeed it has improved performance; I now have a setup with 5 instances (1 per "part", Violin I, Violin II, etc) containing 9 SP each; with 3 SP for each mode of play (normal, Sul Tasto, Sordino), which are divided as such: All players, Group 1, Group 2. I therefore have the possibility to use either all players or divide each section in 2, up to 10 different simultaneous string parts. Even with other instances doing their thing (various other Synchron percussion and keyboards), I experience no dropout. Or very few if I start piling up simultaneous instruments - which may not happen very often in a real composing situation; I can always use the "light convolution" option in the SP for works in progress in the worse case scenario.

So thank you so much for your sterling advice!

I would have another question: I thought I would originally organise things "per desk", and therefore have (for example for violin I) 4 SP with only the corresponding players loaded/active in the mixer, which means 2 players per SP. I would then use a summing track in Logic to control the 4 SP at once when needing a tutti of the 8 players. It worked as intended, but it did not sound as good as the "all players" patches. Did I make another foolish mistake, or as I suspect, the "all players" patch is different (and nicer) than a mere combination of the 4 desks? Or maybe I need to "re-mix" the summing of the 4 tracks to re-balance things when using a tutti? This solution would be more elegant and optimise CPU and memory load even further; but I am not sure if it is achievable.

Again, I am so grateful for insights! I recently endeavoured to up my game with the quality of mockups, but my tiny composer's mind sometimes struggles with the complexity of the technical side of things...

Thanks a lot

[richhickey]

@thomasoehler said:

I would have another question: I thought I would originally organise things "per desk", and therefore have (for example for violin I) 4 SP with only the corresponding players loaded/active in the mixer, which means 2 players per SP. I would then use a summing track in Logic to control the 4 SP at once when needing a tutti of the 8 players. It worked as intended, but it did not sound as good as the "all players" patches. Did I make another foolish mistake, or as I suspect, the "all players" patch is different (and nicer) than a mere combination of the 4 desks? Or maybe I need to "re-mix" the summing of the 4 tracks to re-balance things when using a tutti? This solution would be more elegant and optimise CPU and memory load even further; but I am not sure if it is achievable.

Again, I am so grateful for insights! I recently endeavoured to up my game with the quality of mockups, but my tiny composer's mind sometimes struggles with the complexity of the technical side of things...

Thanks a lot

You're quite welcome. I've used the same summing track techinique in Logic to target multiple instances of SynDS and it worked and sounded fine. Maybe something else is amiss? Make sure you are constructing the group with distinct players (1+2, 3+4, 5+6 etc) and not getting 1+2, 1+2, 1+2, else you could have phasing.

Also note, since you obviously have DS II and DS III as well, that once you use "all players" and no longer the desks you can reclaim the first dimension to select between modes of play in a single SP by copying and pasting the Sul Tasto and Sordino "all players" trees into a single preset.

It will look like this:

Syn DS combo preset

[thomasoehler]

Dear Richhickey,

I just wanted to say a quick thank you about your sterling advice. I did once again follow your idea of joining the modes of play in the same SP tree, and it worked a charm - elegang and satisfying setup!

The Dimension Strings require quite a bit of polishing and programming to sound right, but the potential is great.

Again, many thanks,

[Dewdman42]

great thread!