Buffer Woes

[PTravel]

My mixing system:
computer: 2.6 GHz quad-core, 8 GB RAM, 3 TB RAID 5 (Data), 560 GB SSD (Program)
interface: Fast Track Ultra set to maximum buffer size - 4k (I'm just mixing -latency isn't a concern)
software: Adobe Audition 3.0

Project:
Nearly 8 minutes in length, close to 100 tracks (many instruments, many vocals)

The problem:
I've got a lot of CPU-intensive effects running. I've got tracks ganged to buses where appropriate and, for the most part, effects are applied only to the buses. However, there are around 10 buses. FX include Audition's Studio Reverb, Chorus, and Dynamic Limiting.

As this particular project has grown in complexity, I've been stepping up the buffer size as needed, having started at 256k. However, now, even with the ASIO buffer set to the max of 4k, I'm getting so much crackling and distortion as to make mixing really, really painful. I've been approximating settings as best as I can and then rendering down to a 2-track mix to hear the result. Each render takes 6 minutes or so -- you can imagine what a slow process this has become.

This is from a musical that I'm writing. It's what, in theater, is called "wall-to-wall" which means the music is more or less continuance from curtain up to curtain down, so it's not really possible to chop the piece into smaller "chunks" and then combine them. There are so many tracks because there are multiple lead characters, a chorus that, in some sections, is singing 8 parts, and the orchestration consists of four discrete sections, with three of the four using completely different sets of instruments (it's a science-fiction story -- lots and lots of synths).

Fortunately, this is both the longest and most complicated piece in the show. I've got some shorter numbers with around 80 tracks that will give me the occasional snap and crackle, but nothing like this. I can't afford to buy any additional gear at this point. I do have another Fast Track Ultra, and also a Fast Track Ultra 8R, but that's also limited to a maximum buffer size of 4k.

Any suggestions as to how to get around the buffer problem?

 

[sixer2007]

Any suggestions? Yeah, a quite simple one...
Commit to the sound of your VI and print to actual audio tracks. No more CPU hungry instruments and all your resources can be used to mix it.

The pros often have VI instrument sounds but they don't get the actual VI on a track... That's how they handle 100+ tracks.

 

[PTravel]

Any suggestions? Yeah, a quite simple one...
Commit to the sound of your VI and print to actual audio tracks. No more CPU hungry instruments and all your resources can be used to mix it.

The pros often have VI instrument sounds but they don't get the actual VI on a track... That's how they handle 100+ tracks.

Unfortunately, that's exactly what I do.

I compose and record instrumentation in Sonar X2 and export to .wav files. I import those into Audition, do a rough mix of just the orchestration, then export to a rough stereo master. I use the rough master for recording vocals in a new session in Audition (and to send my actors "music minus one" rehearsal CDs so they can prepare for the recording session), decide which vocal takes I like (which often involves cutting together multiple takes), and then import them into the original orchestration session. Then I do my final mix.

One of the reasons I have so many tracks (I looked last night -- there are 89) is because roughly half of the music was done a number of years ago on Fostex A8LR 8-track deck. Over the years the tape stretched slightly and, once I'd digitized it, I had to do a lot of work pulling it up to pitch and adjusting the tempo. Worse, still, because I was limited to 8 tracks when I did the original recording, I had to do a lot of bouncing to get all the instruments in. For example, I put the bass and drums on a single track, thinking (at the time) that it was okay if both were center channel given that bass and kick drum are, essentially, non-directional. In order to fix this, I actually have three copies of the bass/drum track and use EQ to create stereo field separation, and also to adjust the levels of the higher-voiced percussion as against the bass.

So, thanks, I appreciate the suggestion, but I'm already doing that.

 

[MrWrenchey]

How on earth is your quad-core choking on playing back just audio?
I'll have bloody 20 instrument tracks set up for a session where I need to write a ton of synth, and maybe 30-50 audio tracks depending on the client, and my i7 doesn't even hit 35%.
What are you doing plug-in wise? I mean, if there's 100 tracks, and you're processing EVERY track with it's own EQ or something similar, I suppose you could buss all the background vocals to a buss and apply EQ to just the buss. That'd help with processing power.
Honestly, I'm not sure how you've managed to max a session out with just audio on a quad-core system. Haha, you'll have to give us a bit more information, I think.

 

[PTravel]

How on earth is your quad-core choking on playing back just audio?
I'll have bloody 20 instrument tracks set up for a session where I need to write a ton of synth, and maybe 30-50 audio tracks depending on the client, and my i7 doesn't even hit 35%.

This isn't an i7. It's a 2.6 GHz Q9400. Note that all tracks 89 tracks are audio tracks.

What are you doing plug-in wise? I mean, if there's 100 tracks, and you're processing EVERY track with it's own EQ or something similar, I suppose you could buss all the background vocals to a buss and apply EQ to just the buss. That'd help with processing power.

Most effects are applied at the buss level. There are a few (but not that many) individual tracks that have effects, primarily EQ or dynamic limiting.

Honestly, I'm not sure how you've managed to max a session out with just audio on a quad-core system. Haha, you'll have to give us a bit more information, I think.

Please tell me what do you need to know? One thing: I am working with 96k/32 bits which, of course, is going to be far more of a CPU strain the 48 or 44.1 k / 24 or 16 bit.

 

[DM60]

This isn't an i7. It's a 2.6 GHz Q9400. Note that all tracks 89 tracks are audio tracks.

Most effects are applied at the buss level. There are a few (but not that many) individual tracks that have effects, primarily EQ or dynamic limiting.

Please tell me what do you need to know? One thing: I am working with 96k/32 bits which, of course, is going to be far more of a CPU strain the 48 or 44.1 k / 24 or 16 bit.

PT, the problem is you have a VST or a VSTi that is just CPU intensive or, if your waves (you might have said this) is having to be "real time" corrected that would be a load. Just my thoughts being typed, but if you exported all of your tracks so that it is all aligned, pitched and in the correct time (no eq nothing, just the basics), brought them all back in as a new sessions, then you can begin to treat (using grouping, returns or what ever your DAW uses so you can run multi tracks for the same set of treatments.

Even though it is an i5, (I think an i7 chip would choke as there is some kind of processing going on) it should handle the job. I am thinking your old analog corrected files are the culprit and even though there is no VST, if it is being corrected real time, then it is being processed.

Let me know if I was way off base or didn't really understand the issue.

 

[Steenamaroo]

Hey Ptravel.
The correction thing was the first bit to catch my attention.
If your time/pitch correction is 'live', commit it once your done.

I'm not sure how it works in your DAW, but in ProTools I can have elastic audio turned on for a track and can make time/pitch adjustments.
While EA is active there's a big strain on the PC, but I can choose to commit my changes and the load drops right down.

I had a session of this magnitude once and was forced to split it up and just deal with it as separate sessions.
It wasn't ideal but I ended up mixing so many tracks in one session and so many in another, then brought them all together at the end.
I'm not sure if this is feasible for you.

 

[MrWrenchey]

Hey Ptravel.
The correction thing was the first bit to catch my attention.
If your time/pitch correction is 'live', commit it once your done.

This will affect it. A lot of things like Melodyne are pretty CPU intensive, I always print the tuned audio and just hide the original track and make it inactive. That way, if I hear something is still a little off pitch later, I can just reactivate the track, fix the one note, and print that phrase back out.

Yowza! It's probably a lot of the 32-bit and 96kHz too. I always just run 24-bit 44.1kHz. Personally, I never really noticed enough of a difference between 44.1 and the higher resolutions to want to use the higher ones and waste space, especially considering you dither to 44.1 anyway... But that's a personal preference, and I'm sure someone will come on here and chastise me, But I suppose that's that. xD

Hmm. To me, it sounds like it's primarily those two things. High CPU usage from plug-ins, and it takes even more CPU power to run the higher bit-depth/sample rate audio through those plug-ins.

The only suggestion I can really make there is to maybe use smaller bit depths and sample rates if you know you'll use that many tracks. Either that, or upgrade the CPU maybe. =]

 

[DM60]

This will affect it. A lot of things like Melodyne are pretty CPU intensive, I always print the tuned audio and just hide the original track and make it inactive. That way, if I hear something is still a little off pitch later, I can just reactivate the track, fix the one note, and print that phrase back out.

Yowza! It's probably a lot of the 32-bit and 96kHz too. I always just run 24-bit 44.1kHz. Personally, I never really noticed enough of a difference between 44.1 and the higher resolutions to want to use the higher ones and waste space, especially considering you dither to 44.1 anyway... But that's a personal preference, and I'm sure someone will come on here and chastise me, But I suppose that's that. xD

Hmm. To me, it sounds like it's primarily those two things. High CPU usage from plug-ins, and it takes even more CPU power to run the higher bit-depth/sample rate audio through those plug-ins.

The only suggestion I can really make there is to maybe use smaller bit depths and sample rates if you know you'll use that many tracks. Either that, or upgrade the CPU maybe. =]

I think if you are choking an i5, an i7 will not give you that kind of performance boast. You would have to move to a multi CPU unit and a server grade OS (Windows 2008 or 2010) so you can control the CPUs better. The difference between i5 and i7 are really marginal if you to really look into it.

 

[Steenamaroo]

I think if you are choking an i5, an i7 will not give you that kind of performance boast. You would have to move to a multi CPU unit and a server grade OS (Windows 2008 or 2010) so you can control the CPUs better. The difference between i5 and i7 are really marginal if you to really look into it.

It's marginal with certain chips, but there are i5 chips out there benching along side pentiums and celerons of yester-year.

I know there are so many different kinds of tests, but broadly speaking his chip is about 1/3 as capable as an i7-3970X.

 

[jimmys69]

Keep in mind also, that not all i5's are 4 core processors. The cheaper ones are only two core -4 thread. The i7's are all quad core-8 thread.

 

[PTravel]

Well, this is interesting (at least to me ).

Just for the hell of it, I opened this project on my laptop, which has an i7 and 16 gig of RAM. It played perfectly with not a single snap or crackle. I think, after all this, I'll just edit this one project on the laptop and leave it at that.

 

[DM60]

Keep in mind also, that not all i5's are 4 core processors. The cheaper ones are only two core -4 thread. The i7's are all quad core-8 thread.

I think in that case, then there would be a big performance gain from an i5 to an i7. Plus, there are some i7 chips with 6 cores.

I wasn't aware until Jimmy made the statement that there are 2 core i5 chips. In researching it, I wasn't aware of a 6 core i7 chip. I am not as up to date on these things as I use to be. Plus, I usually stay with AMD chips.

Good information.

 

[Thurgood]

I do not mean to be rude but why 96k? I do understand the use of high sample rates but.... 100 tracks 32 bit+ plugs. WOW. That would kill my older pc and give my current one a bad hangover.

 

[PTravel]

I think in that case, then there would be a big performance gain from an i5 to an i7. Plus, there are some i7 chips with 6 cores.

My mixing computer has a 4-core 2.6 GHz i5. My laptop has 4-core, hyper-threaded, 2.3 GHz i7.

I do not mean to be rude but why 96k? I do understand the use of high sample rates but.... 100 tracks 32 bit+ plugs. WOW. That would kill my older pc and give my current one a bad hangover.

Well, I don't want open that debate here. Short version: My personal belief (which I acknowledge may be dead wrong) is that higher bit rates produce smaller rounding errors when corrections and effects are applied. I have the storage space and, until this particular project, I had the computer power to manage 32-bit at 96k with no problem. As this particular project only gives my mixing computer headaches, and not my laptop, I'll just finish it on the laptop.

 

[DM60]

Well, I don't want open that debate here. Short version: My personal belief (which I acknowledge may be dead wrong) is that higher bit rates produce smaller rounding errors when corrections and effects are applied. I have the storage space and, until this particular project, I had the computer power to manage 32-bit at 96k with no problem. As this particular project only gives my mixing computer headaches, and not my laptop, I'll just finish it on the laptop.

And I think we can all agree, if this is what works for you, then that is the correct.

 

[jimmys69]

And I think we can all agree, if this is what works for you, then that is the correct.

Amen brother...