Page 1 of 2

Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 1:51 am
by Elektroakoustika
Hey MotuNation,

So I was sitting at my very large orchestral template, and I sat there wondering what I could do in order to save CPU and performance in DP. Right now things are running just fine, but I'd like to try to find out any way to make my template run more efficiently. Let me run down the template for you:

- Mac Pro 8-core 2.26 Ghz with 16GB of ram with 4 1TB WD Caviar Black Drives.
- DP is on the boot drive
- All String libraries are on the second drive
- All Woodwinds and Piano libraries are on the third drive (This is also where Omnisphere is)
- All Brass and Percussion libraries are on the fourth drive
- All virtual instruments (with the exception of Ivory) are running outside of DP in VE Pro servers.

Digital Performer 7.2 at a buffer of 512 running through my firewire 828mk3.

- 3 Instances of 64 bit VE Pro each running a Kontakt's 4.1 Player of LA Scoring Strings. I have all the divisi patches for the Leg/Speed Port LT patches running as well as full section mixes of all the other articulations. This totals up to about 44 separate tracks leading into DP. I've tweaked Kontakt's disc streaming settings to keep memory as free as possible without dropping notes. Inside DP, all three tracks lead into a AUX track which has 1 Masterworks EQ and 2 Proverb effects on it.
- 1 instance of 64 bit VE Pro running 2 instances of Omnisphere (just choirs) for a total of 16 tracks. Inside DP this leads to 1 Masterworks EQ and 2 Proverb effects on it.
- 1 Instance of 64 bit VE Pro running 15 VSL Woodwind tracks with 4 articulations on each track. Inside DP this leads to an AUX connected with the VSL Brass with 1 Masterworks EQ and 2 Proverb effects on it.
- 1 instance of 64 bit VE Pro running 5 VSL Brass tracks ranging from 4 to 6 articulations on each track. Inside DP this leads to an AUX connected with the VSL Woodwinds with 1 Masterworks EQ and 2 Proverb effects on it.
- 3 Instances of 32 bit VE Pro running EastWest EWQLSO in PLAY. Each VE Pro instance has it's own instance of PLAY. One instance is a heavy brass instance. One is short woodwinds. And the last instance is harps and short articulation strings. Inside DP, EACH of these sections have 1 MW EQ and 1 Proverb instance on them.
- 4 Instances of 64-bit VE Pro running Kontakt 4.1 Player and True Strike 1. No reverbs on these.
- 2 Instances of 32 bit VE Pro running Full Version of the Kontakt 3.5 each. Each Kontakt 3.5 instance has instruments from ProjectSam's True Strike Percussion library. Inside DP, only one of the tracks has a single Proverb effect on them.
- Inside DP, Synthogy Ivory's Steinway is loaded with one Proverb effect on it.

All of this is running in my super template of around 120 or so MIDI tracks inside DP. I disconnect the VE Pro instances when I'm not using any instruments that they contain.

Currently I'm using about 13 GB of my 16 GB of ram with this template.

The main reason I have SO many proverb instances is because I'm using one instance for early reflection, and the second instance is the hall. Also, since I'm using libraries from many different companies, I have to EQ them differently, as well as position them within the reverb to get a more homogeneous sound.

So does anyone have any ideas of how I could make this more efficient (especially on the side of CPU)? I know I could raise the buffer, but like I said, the computer is able to handle everything now, but who knows what would happen when I start expanding the template, adding more tracks, adding different effects. And I also want to avoid pushing the computer too much that I cause a crash.

Even though I'm not looking for any hardware upgrades at this moment, if anyone has any ideas in that department I would be greatly appreciative!

Thanks! :D

-ea

ps - random additional question - the Mac Pro I have does not have a Firewire 400 port on it. Only Firewire 800 ports. I use a 400-800 cable that came with an external hard drive as my audio cable between my MOTU 828 mk3 and the Mac Pro. Do you think I should buy a new cable that is more designed for audio/etc? If so, what cable should I get?

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 5:57 am
by AnthonyS
Elektroakoustika wrote:Hey MotuNation,

So I was sitting at my very large orchestral template, and I sat there wondering what I could do in order to save CPU and performance in DP. Right now things are running just fine, but I'd like to try to find out any way to make my template run more efficiently. Let me run down the template for you:

- Mac Pro 8-core 2.26 Ghz with 16GB of ram with 4 1TB WD Caviar Black Drives.
- DP is on the boot drive
- All String libraries are on the second drive
- All Woodwinds and Piano libraries are on the third drive (This is also where Omnisphere is)
- All Brass and Percussion libraries are on the fourth drive

Digital Performer 7.2 at a buffer of 256 running through my firewire 828mk3.

- 3 Instances of 64 bit VE Pro each running a Kontakt's 4.1 Player of LA Scoring Strings. I have all the divisi patches for the Leg/Speed Port LT patches running as well as full section mixes of all the other articulations. This totals up to about 44 separate tracks leading into DP. I've tweaked Kontakt's disc streaming settings to keep memory as free as possible without dropping notes. Inside DP, all three tracks lead into a AUX track which has 1 Masterworks EQ and 2 Proverb effects on it.
- 1 instance of 64 bit VE Pro running 2 instances of Omnisphere (just choirs) for a total of 16 tracks. Inside DP this leads to 1 Masterworks EQ and 2 Proverb effects on it.
- 1 Instance of 64 bit VE Pro running 15 VSL Woodwind tracks with 4 articulations on each track. Inside DP this leads to an AUX connected with the VSL Brass with 1 Masterworks EQ and 2 Proverb effects on it.
- 1 instance of 64 bit VE Pro running 5 VSL Brass tracks ranging from 4 to 6 articulations on each track. Inside DP this leads to an AUX connected with the VSL Woodwinds with 1 Masterworks EQ and 2 Proverb effects on it.
- 3 Instances of 32 bit VE Pro running EastWest EWQLSO in PLAY. Each VE Pro instance has it's own instance of PLAY. One instance is a heavy brass instance. One is percussion. And the last instance is harps and short articulation strings. Inside DP, EACH of these sections have 1 MW EQ and 1 Proverb instance on them.
- 2 Instances of 32 bit VE Pro running Full Version of the Kontakt 3.5 each. Each Kontakt 3.5 instance has instruments from ProjectSam's True Strike Percussion library. Inside DP, only one of the tracks has a single Proverb effect on them.
- Inside DP, Synthogy Ivory's Steinway is loaded with one Proverb effect on it.

All of this is running in my super template of around 100 or so MIDI tracks inside DP. I disconnect the VE Pro instances when I'm not using any instruments that they contain.

Currently I'm using about 13 GB of my 16 GB of ram with this template.

The main reason I have SO many proverb instances is because I'm using one instance for early reflection, and the second instance is the hall. Also, since I'm using libraries from many different companies, I have to EQ them differently, as well as position them within the reverb to get a more homogeneous sound.

So does anyone have any ideas of how I could make this more efficient (especially on the side of CPU)? I know I could raise the buffer, but like I said, the computer is able to handle everything now, but who knows what would happen when I start expanding the template, adding more tracks, adding different effects. And I also want to avoid pushing the computer too much that I cause a crash.

Even though I'm not looking for any hardware upgrades at this moment, if anyone has any ideas in that department I would be greatly appreciative!

Thanks! :D

-ea

ps - random additional question - the Mac Pro I have does not have a Firewire 400 port on it. Only Firewire 800 ports. I use a 400-800 cable that came with an external hard drive as my cable. Do you think I should buy a new cable that is more designed for audio/etc? If so, what cable should I get?
Hi:

You may want to xcheck out Plogue Bidule:
http://www.plogue.com/?page_id=56

it's a nifty routing app that works via reWire. Best I can explain is it takes the strain of the VI's off of DP into a separate enviornment. The WS has a much better explanation!

Also, there are some great YouTube tutorial videos by Mike Patti (c0mp0ser), a DP user.

Here's a link to one of them:http://www.youtube.com/watch?v=KBB4Txnpl78

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 6:41 am
by FMiguelez
.

Wow!

I'm jealous now. I wish I could run a fraction of all that in ONLY my Master computer 8)

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 7:37 am
by willheim
sounds like a terrific pallet made with grat care.

my pallet is a mess.

one thing i do that seems to work well for me is to run ewqlso pro lib in a standalone. i have some 40 instruments in it, it runs outside of dp and is very stable. i run ivory the same way. it is easy to do. turn on iac in the MIDI set up. add dp outputs .

assign ewqlso to match. always star dp first then start the stand alones. it is rock solid.

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 11:44 am
by Elektroakoustika
AnthonyS wrote: Hi:

You may want to xcheck out Plogue Bidule:
http://www.plogue.com/?page_id=56

it's a nifty routing app that works via reWire. Best I can explain is it takes the strain of the VI's off of DP into a separate enviornment. The WS has a much better explanation!

Also, there are some great YouTube tutorial videos by Mike Patti (c0mp0ser), a DP user.

Here's a link to one of them:http://www.youtube.com/watch?v=KBB4Txnpl78
VE Pro is doing the same thing as Bidule, just in 64-bit. All of my instruments with the exception of Ivory are running outside of DP within VE Pro "Servers."
willheim wrote:one thing i do that seems to work well for me is to run ewqlso pro lib in a standalone. i have some 40 instruments in it, it runs outside of dp and is very stable. i run ivory the same way. it is easy to do. turn on iac in the MIDI set up. add dp outputs .

assign ewqlso to match. always star dp first then start the stand alones. it is rock solid.
I've tried this before. I actually found it to be a hassle, especially since I have 3 separate instances of PLAY. But I do see how it could reduce some CPU usage. Interestingly PLAY has become super stable for me inside VE Pro since the the new update. PLAY was never stable inside DP. And I've never had any problems with Ivory.

Thanks guys! Keep the ideas coming!
-ea

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 12:07 pm
by AnthonyS
OK. Since I'm not that familiar with VEPro, I didn't know you could do that. Very cool. May be simple, but what if you increased your buffers to 512. Too much latency?

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 12:10 pm
by Elektroakoustika
AnthonyS wrote:OK. Since I'm not that familiar with VEPro, I didn't know you could do that. Very cool. May be simple, but what if you increased your buffers to 512. Too much latency?
Tried that too. Its not a terrible latency at 512, I just prefer to work in 256. And I didn't really see any performance differences between the two on the current template.

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 1:45 pm
by jroadrage
What are your buffers in VE PRO? I never tweaked mine because I haven't hit a bottleneck yet buy I think upping the buffer multiplier could ease the CPU hit some. The latency would be compensated for on playback, you'd only have to deal with it when inputting MIDI. It's also the sort of thing that you can do on the fly or as needed.

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 2:09 pm
by doodles
I'm sorry, but I just can't get my head around bidule. That's a tutorial for how to use it? What the heck? :mrgreen:

Elektroakoustika, I use all my VE Pro's on some slave macpro's, but quick question for you - why are you running Kontakt 3.5? You should at least upgrade to the 4 player, or 4 sampler. You seem to have it in the 64-bit mode, so why use 3.5 in the 32-bit?

I'm running a couple of 32-bit instances with Mach 5, and one version of Kontakt (cos it's from an old template for this score), but 64-bit for all the other kontakts, VE's, spectrasonics, etc. Kontakt keeps crashing the 32-bit mframe for me - just to warn you!

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 19, 2010 6:33 pm
by Elektroakoustika
jroadrage wrote:What are your buffers in VE PRO? I never tweaked mine because I haven't hit a bottleneck yet buy I think upping the buffer multiplier could ease the CPU hit some. The latency would be compensated for on playback, you'd only have to deal with it when inputting MIDI. It's also the sort of thing that you can do on the fly or as needed.
Umm... VE Pro buffers? I'm running in server mode, which means DP is controlling buffer size right? I don't see any options in VE Pro Server for buffer size. Just multi-core threading. Or maybe I'm missing something here.
doodles wrote:I'm sorry, but I just can't get my head around bidule. That's a tutorial for how to use it? What the heck? :mrgreen:

Elektroakoustika, I use all my VE Pro's on some slave macpro's, but quick question for you - why are you running Kontakt 3.5? You should at least upgrade to the 4 player, or 4 sampler. You seem to have it in the 64-bit mode, so why use 3.5 in the 32-bit?

I'm running a couple of 32-bit instances with Mach 5, and one version of Kontakt (cos it's from an old template for this score), but 64-bit for all the other kontakts, VE's, spectrasonics, etc. Kontakt keeps crashing the 32-bit mframe for me - just to warn you!
If you take a look at my template, you'll see that all of LASS is running in the 64-bit Kontakt 4.1 Player. The reason I have 2 Kontakt 3.5's is because True Strike (well the version I have) requires the full version of Kontakt to run it (thus I can't use the player). And alas, I have not upgraded my full version of Kontakt 3.5 to 4 yet because it didn't seem to be a worthy investment at the moment. Especially since I don't have any problems with it. My Kontakt 3.5 has never crashed VE Pro. Also 3.5's memory server keeps the memory free from the 32-bit VE Pro Server.

-ea

Re: Optimizing a Large Orchestral Template

Posted: Tue Jul 20, 2010 12:59 am
by doodles
Elektroakoustika wrote:
doodles wrote:I'm sorry, but I just can't get my head around bidule. That's a tutorial for how to use it? What the heck? :mrgreen:

Elektroakoustika, I use all my VE Pro's on some slave macpro's, but quick question for you - why are you running Kontakt 3.5? You should at least upgrade to the 4 player, or 4 sampler. You seem to have it in the 64-bit mode, so why use 3.5 in the 32-bit?

I'm running a couple of 32-bit instances with Mach 5, and one version of Kontakt (cos it's from an old template for this score), but 64-bit for all the other kontakts, VE's, spectrasonics, etc. Kontakt keeps crashing the 32-bit mframe for me - just to warn you!
If you take a look at my template, you'll see that all of LASS is running in the 64-bit Kontakt 4.1 Player. The reason I have 2 Kontakt 3.5's is because True Strike (well the version I have) requires the full version of Kontakt to run it (thus I can't use the player). And alas, I have not upgraded my full version of Kontakt 3.5 to 4 yet because it didn't seem to be a worthy investment at the moment. Especially since I don't have any problems with it. My Kontakt 3.5 has never crashed VE Pro. Also 3.5's memory server keeps the memory free from the 32-bit VE Pro Server.

-ea
fair enough. In which case, I would stick with it like that. In fact, you've just given me a thought - I might try 3.5 in the 32-bit VE to see if that works. Kontakt 4.1 is a thing of great beauty, But it's just causing me a PITA in 32-bit! Good luck with it, dude. The True Strike upgrade earlier this year (to version 2), is nice -the keymapping's great, but the old ones are pretty cool, too. So no urgency to upgrade. Good Luck! :)

Re: Optimizing a Large Orchestral Template

Posted: Tue Jul 20, 2010 11:13 am
by willheim

I've tried this before. I actually found it to be a hassle, especially since I have 3 separate instances of PLAY. But I do see how it could reduce some CPU usage. Interestingly PLAY has become super stable for me inside VE Pro since the the new update. PLAY was never stable inside DP. And I've never had any problems with Ivory.

Thanks guys! Keep the ideas coming!


Huh. That's great news - VE pro is the vienna thing right? are you running that over LAN?

Yeah, Play blows as a insert - I didn't know you could run it with VE. Makes me want to buy the damn thing. Standalones rock though, for me - great way to spread the resources of ann 8 core around and use more memory and make DP more responsive and, well snappier.

Re: Optimizing a Large Orchestral Template

Posted: Tue Jul 20, 2010 11:14 am
by willheim
btw i use plogue on my old dual gig g5, and it is nice once you learn to drive it.

Re: Optimizing a Large Orchestral Template

Posted: Wed Jul 21, 2010 6:34 pm
by jroadrage
Elektroakoustika wrote:
jroadrage wrote:What are your buffers in VE PRO? I never tweaked mine because I haven't hit a bottleneck yet buy I think upping the buffer multiplier could ease the CPU hit some. The latency would be compensated for on playback, you'd only have to deal with it when inputting MIDI. It's also the sort of thing that you can do on the fly or as needed.
Umm... VE Pro buffers? I'm running in server mode, which means DP is controlling buffer size right? I don't see any options in VE Pro Server for buffer size. Just multi-core threading. Or maybe I'm missing something here.
If you look at the VE connector plug in DP there's a drop down window that indicates the buffer multiplier, *I think* increasing the buffer eases the CPU hit and only adds latency on input, not playback.

Re: Optimizing a Large Orchestral Template

Posted: Mon Jul 26, 2010 10:51 pm
by Elektroakoustika
Just thought I would update with a discovery I found this last week.

Since I have 10 or instances of VE Pro running in my template, I had thought that if the multi-threading setting in VE Pro was set to 16 threads (the highest possible) that I would get the best performance. Boy was I wrong. After reading the VE Pro manual again I found out that the threading is for each individual instance. So I kept setting it lower and then checked my CPU. I ended up at 1 thread per instance and it lowered my CPU usage across all of my cores by 20%. Now 20% may not seem like a lot, but let me tell you, its pretty substantial!

But if you are running multiple instances of VE Pro, I really encourage you to lower the threading! Your CPU will thank you!

-ea