My old system was a p4 2.6 with 1 gig RAM. Using FAT mode in Cubase SX2, CPU usage was around 60% for 1 instance of MC. The same system running wavelab only saw 18%.

I just upgraded to a Quad Core 2.6 mac pro with 4 gigs of RAM. Using PTHD, and fat mode, a single instance of MC uses approx 45%.

All of my other plugins have experienced HUGE performance gains. Why doesn't MC reflect this huge increase in processing power? It seems oddly inneficient in certain DAW apps.

-Z-

The performance depenends on many issues including plug-in version, sample rate, audio buffer setup and other factors.
I don't know all those important details of your setup but from our expearience you should notice a similar increase in processing power that you notice with other plug-ins.
Please notice that every host is using CPUs and internal CPU cores in a different way. It may also display performance ina different. For instance it is possible that the application is turned to run on a single core r it shows how much the plug-in is using of a signle core while other cores are still ready to run other instances of the plug-ins.

Regards,

The PT sessions are 24 bit/44.1 and the audio buffer is set to 256 samples. More importantly, these settings are exactly the same as I used with the previous projects/system.


AFAIK, PT is set up to handle multiple cores, but I will check into that and report back...

Thanks,

-Z-

Which version of the plug-in do you use?

I'm using 1.5.4 ilok.

Using MasterComp 1.5.4, in a 24/96 project with I/O buffers maxed at 1024 and Process Buffer maxed to Large, I can't run even one instance of MasterComp with FAT setting engaged at all. Without FAT my CUP meter shows 50%, with FAT it red-lines into 100%. No other plugs at all. Ouch, that hurts. What can I do?

Mateusz,

I'm not trying to be a pain here, but there are clearly performance issues with mastercomp in various hosts. None of my other plugins are nearly as sensitive to the host application for performance as MC.

Have you actually tested this plugin with Protools HD on a Intel Mac? If you have then you must agree that the current situation is unusable.

Thanks,

-Zak

I just tried to use MC in a PTHD session at 96k. This is a Quad 2.6 macpro and in fat mode, I was unable to get even a single instance!!!!!
When I switched it to regular mode it crashed PT and I had to force quit...not happy.
Clearly something is wrong.This is the ONLY plugin that's giving me grief with my system.

Can you please address this problem or at least aknowledge it?

Thanks,

-Zak

MasterComp is CPU hungry plug-in. The real CPU usage depends on host type, buffering settings, host's CPU scheduling, sample rate and FAT of course. We will take a closer look at this problem.

Which version of PTHD do you work on?

Regards,

just a question: i thought fat is for upsampling internally to get better results (less distortion etc.).
so if i have already a high samplerate like 88.2 or 96 khz, does engaging fat make any difference to the sound ???
mathias

Mateusz, thanks for the reply.

I'm using PTHD 7.4 cs2
MacPro quad 2.6 4gigs kingston ram.
OSX 10.4.11
Buffer setting typically set to 512, but raising to 1028 had minimal effect with MC. Sessions are 24 bit 44.1

No problems with ANY of my other plugins...

Have you actually tested MC in PTHD 7.x?

Thanks,

-Zak

This is not just a problem with MacPro. I am running PTLE 7.4cs4 and see unbelievably severe cpu utilization on an AMD Quad Opteron at a 1024 buffer setting. I have trouble running one instance with FAT on at 44.1/48K sample rates. This is a computer that can run about 180 D-Verbs while recording 24 tracks.

Now here is the interesting thing; I do not see this behaviour when I use the VST version in Vegas, Sound Forge, Reaper, etc. I know that both PTHD and PTLE are real time implementations, but there is something amis here.

Greg

Yes indeed, something is amiss...it appears that Mateusz has been unable to properly test MC in Protools. I realize this is a small company and you must be stretched in many directions but Protools shouldn't be ignored this way. It's literally months between postings in this forum. Why bother even having it.

Hello,

I am sorry that you waited so long.
The problem with some hosts looks strange to us. Testing it in various hosts doesn't help to solve this issue. It can be some kind of CPU related problem. For instance on my intel core 2 duo the vst version is using up to 40% of CPU in peaks at 44.1 without FAT.

Some of you wrote that he didn't have such problems on previous system. Would you give us some information on what system was previously running what host and which MC version compared to nowadays system, host and MC version? Maybe this would help us to better understand the problem.

The FAT mode is a double sampling algorithm. It is the best to use it at 44.1kHz. At low sample rates it gives the greatest difference in sound. However some engineers say that they are not using FAT be default in low sample rates. Others say that they hear the difference at 96kHz.

Regards,

Is there anything new on this. MC is using up more of my new BlackBook 2.4 than of my PowerBook 1.5!

- please help...

Hy! i have problem to with Master Comp i like it very much but...i have a 2.8 mac intel with 8g mem.and pt HD2 and i not happy at all with how it's run. zakco have perfect right.and the same problem with neon HR. May be if you try to make a TDM and TEST IT, will work.till then i will used anyway but very desapointed.
all good and no affence.

What is the status of this problem?

We still don't know the reason why there are such performance issues on some machines running Pro Tools. Did you try to readjust audio buffers?

Regards,

Yes... and it's not just PT. Can you publish the performance benchmarks from your testing?

We are working on MC performance. The real problem is that MC is using very exact math which costs a lot of CPU cycles. There is no good optimization for this algorithm without shortcuts.
We can make an alternative "every day use" version of the MasterComp which would have very close sound to the MasterComp, it would still work in double sampled mode but it would not be exactly the same.

Regards,

Or we can add a switch to disengage very CPU expensive code of level detector and exchange it with very good and similar sounding but more traditional solution. This would allow us to have the plug-in almost twice faster. There is very little space to optimize current version of MC without any cuts on the algorithm.

Regards,

- thanks, Mateusz! I'm glad you're working on it. Any or all of those ideas/improvements would be good.

I'm not through setting up my Intel Mac for work, yet - so I'm mainly back on my PPC. I'll try to post more findings during the work transition.

I have noticed one thing that I hadn't before, though - sometimes when I turn off FAT or even bypass MC, I don't seem to get any CPU usage reduction.

Mateusz, 2 questions:

1. Have you ever specifally tested MC running on a Protools HD, TDM system? If so, which OSX versions have you tested?

My system:
MacPro quad 2.6
4 gb ram
OSX 10.4.11
PTHD 7.4.2
Mastercomp ilok version 1.5.4


Disclaimer: I have ZERO experience as a programmer so I am making no claims or assertions here, just a end user's question...

2. I don't understand how this can be a simple case of MC being CPU hungry. Other plugins seem to respond to computer power upgrades in a more or less proportional manner. MC actually performs LESS efficient on some newer, industry standard machines such as Macpros. What is the explanation for this anomoly?

Thanks for your responses on this thread,

-Zak

Hello,

I made various comparisons on my Core 2 Duo macintel and it looks like in every application there is a huge CPU usage for PSP MC PT 7.4 performance looks similar however PT looks more sensitive to CPU usage then some other applications.
This plug-in was designed for mastering purposes so we didn't care CPU usage, it was also developed on Pentium M and it was working pretty good for its complexity.
The comparison between PC and Mac version shows that the MS compiler deals a bit better with MasterComp's code then XCode does.
Sadly switching to newer CPUs may give a worse performance in some cases. In the case of the MasterComp the algorithm is like a series stream of processing instructions, they cannot be reorganized in the way to get easy through a nowadays more parallel architecture.
So we cannot make considerable optimizations without changing the code. This is why I think about giving some options to internally simplify some unusual - hyper-hi-end parts of the code.
So if you want all bits of the MasterComp with FAT on and limiter on, RMS and soft knee you will not get easy on CPU, at least we have no exact solution right now. If you need most of its features and need the MasterComp in the mix I think the option to simplify the algorithm in its most mastering parts is a good idea.

Regards,

it is posible to do a TDM version ?

We don't plan any TDM versions nowadays.

Regards,

Is it just me or does the FAT mode seem to suck out the low mids,I've just been realizing this lately, what exactly happens in FAT mode? I still use this more than any comp I have but I'm tending to hold back on the FAT mode.

There is no difference in the algorithm except that the plug-in is working in doubled sample rate. The conversion itself is very transparent.

Regards,

i hatve to desable Fat mode to.I am too sory becuse i buy the MC and Neon.i use now just for small project.Anyway...vst is work good.but i want MC to work good in pro tols to.

i don't notice performance difference between RTAS and VST but I always noticed that Pro Tools is more fragile to high CPU usage and stops processing earlier then other hosts even with high CPU limit setup.

Regards,

Thanks for clarifying that.
Another question is there any benefit at all in engaging the FAT mode when you already work at high sample rates like 88k or 96k???

Because whenever engage the FAT mode when working at those rates all I get is distortion and the plugin can't cope, it's not my computer I have an 8core MacPro but even when my CPUs aren't full like when I'm just starting to mix when I'm on a high sample rate the plugin just can't cope.

is there any benefit of engaging FAT mode when your already working in 96K?? anyone??

Some users still hear the difference however from technical point of view double sampling at 96kHz would give just a little fraction of a benefit when you double sample at 44.1 or 48kHz.