64bit Engine

[Ezekiel Tsman]

Hi Cakewalk Gurus,

please what's the effect of 64 bit engine in Sonar, what quality does it add to audio?

[Jim Fogle]

It is worth noting that the 64 bit audio engine you're asking about performs the internal math  used to process audio data.

At its most basic the audio engine is a calculator.  As the audio engine processes audio it is adding, subtracting, multiplying and dividing.  Bigger numbers means bigger data sets.  The bigger number an audio engine is composed of means more memory the engine can access, the more audio tracks it can handle at one time and the more accurate the resulting processed audio data will be.

Ever see a calculator give an error message because the answer has more numbers than the calculator can compute or display?  The largest raw number a 32 bit engine can manage is 4,294,967,295 which is a 10 digit, base ten number.  The largest raw number a 64 bit engine can manage is 9,223,372,036,854,775,807 which is a 19 digit, base ten number.

Edited June 2, 2019 by fogle622

[slartabartfast]

My understanding of the 64 bit audio processing is that it just uses larger numbers to represent data. I fail to see how it can make more memory available, if anything it is going to require more memory to do the same operations as using  smaller numbers would, and that might result in slower processing with big numbers as well. The only argument for using larger digital representations is that the cumulative errors due to truncation or rounding at the least significant digit with each calculation/operation will be less likely to show up in the final result.  As fogle622 points out, the numbers that can be represented are enormous, but remember you are only concerned with errors in either the 32nd bit or the 64th bit depending on the base, but you are likely to be saving your audio at 16 bit--so errors below that level are not going to make it through to the final result. It is conceivable that your audio calculation could encounter a very large number of calculations each using the previous result as input, and that every error is systematic i. e. each rounding error is in the same direction. In practice, errors tend to be more random and cancel each other out, so it is pretty unlikely that you are going to use up all the room you have to drop/hide them by processing with 32 bits. But if your computer has the resources to crunch the bigger numbers, and you want to cover for very unlikely computational events,  it is there.  

There may be some confusion between using Windows 64-bit as the operating system and using Cakewalk's setting to use its internal 64-bit calculations/audio engine, which are not at all the same. Windows 64 does indeed give access to a larger amount of memory than Windows 32-bit, and depending on how the applications are coded may result in faster processing.

Edited June 3, 2019 by slartabartfast

[Jim Fogle]

I apologize my first response rambled.  I was in a hurry.   I intended to remove the memory reference but failed to.  That was said in error.  slartabartfast made the point I was trying to make; the larger the data set the audio engine can handle the more accurate the resulting audio will be.