can someone explain how my system (Task Manager) shows CPU usage at approx 21% with a C3 song (with 3 VSTi’s loaded) but in C3 it shows it at well above 50% - with occasional jumps to 70%?
Task Manager and Load in C3 are completely different things. There is an interesting article in the blog explaining strong differences between load in C3 and forte - maybe a good starting point for understanding things behind:
Why? They are both only measuring CPU cycles…why the differnce?
From the Glitch Free book:
⦁ CPU Load vs Audio Load
Most audio programs have a load meter – an indicator that shows how much load the system is under with respect to audio processing. Windows Task Manager reports a similar load measurement as a percentage of CPU resource usage.
Although these two metrics are similar and seem related they are in-fact measuring quite different things and can’t be compared.
The load reported by most audio programs is the amount of time taken to process an audio cycle as a ratio of the buffer duration. For example, if it takes 3ms to process one audio cycle and the audio buffer duration is 6ms, the load will be displayed as 50%.
In contrast, the CPU load reported by Windows is the percentage of available CPU cycles that are being used by a particular program. Note that it’s possible for a program to be waiting for something (which takes time) without consuming CPU cycles (because Windows has suspended the program’s execution and possibly using the CPU for something else while waiting).
In other words there’s not necessarily a relationship between time taken and CPU resources used – which is why these two metrics rarely align.
The reason that audio programs display load as a measure of time taken is that time is the critical factor in providing reliable audio – remember that if the program fails to deliver the next audio buffer in time, a glitch will result.
You should use the audio load meter to determine how close you are to exceeding the capabilities of your computer and adjust the buffer duration accordingly.
The best way to test this is to load up the most demanding configuration you have and if you’re using virtual instruments, play a MIDI file that exercises all the loaded instruments and keep an eye on the load meter.
A rough rule of thumb is for the load meter stay below about 50% - however this also depends somewhat on the sample rate and buffer size. You’ll notice that shorter buffers are far less stable than large buffer sizes. This is because a delay in a smaller audio buffer leads to a bigger percentage change than the same delay with a longer buffer.
For example, a 1ms stall with a 6ms audio buffer equates to about 16% difference in the load meter. That same 1ms stall on a 12ms audio buffer adds just 8%.
In other words the smaller the audio buffer – the higher the risk of a dropout and therefore the lower overall load you want to start with.
oky doky, thanks Terry that makes sense. This came about because in a couple of songs the C3 meter was jumping to 70% and I was getting crackling (the first time it has happened) so I guess I should check out the buffer and perhaps give it a bump up.