Should this box be checked? It is not checked by default so I'm not sure if I'm supposed to enable it.

You don't have to, but some people think that the extra precision can make your mix sound better.  If I remember right, the late Terry Howard and Ray Charles thought it was an improvement.
 
John

Here's the facts:
 
http://www.cakewalk.com/S...kb/reader.aspx/2005278

Processing at 64 bits means smaller rounding errors when SONAR is doing math.  You may or may not hear a difference, but on today's computers it is no problem.  So why not?
 
@

Actually the rounding error thing is probably a red herring. The standard numerical representation is sufficient to produce sound output that is exactly as accurate as the larger numbers when they are rounded back down to anything you can put out through a speaker.
There is a potential for digital clipping during processing (math operations) if you are totally ignoring your levels and just combining (adding up) a gazillion tracks. Summing lots of tracks means potentially coming up with a large enough number to overflow your numerical representation. I doubt most users would ever run into that.
On the other hand, if your computer is powerful enough, you will not notice any degradation in performance by using larger numbers in the calculations, and your friends with the golden ears will not be able to blame the audio engine for your failures.

slartabartfast
On the other hand, if your computer is powerful enough, you will not notice any degradation in performance by using larger numbers in the calculations, and your friends with the golden ears will not be able to blame the audio engine for your failures.

So just how powerful does your computer have to be for the 64 bit double engine to be beneficial?

As has been previously noted, 64bit double precision does indeed technically reduce errors and allow for more headroom.
 
However, 32bit single precision already allows for hundreds of dB's of headroom above 0dBFS before it will clip (!!!), making it utterly impossible to clip (at least without intentionally trying very, very, very, very, very hard to make it clip just to prove you can). So it is ridiculously silly to claim "more headroom and reduced chances of clipping" as some kind of a benefit, but they do.
 
As for the errors - in 32bit single precision they are already so low in level that they will not be audible. You should note that the level of the errors (and how slowly they accumulate) is never ever put into context by people making claims about "more accurate audio".
 
So it really doesn't matter if you turn it on or not.

Mosvalve

slartabartfast
On the other hand, if your computer is powerful enough, you will not notice any degradation in performance by using larger numbers in the calculations, and your friends with the golden ears will not be able to blame the audio engine for your failures.

So just how powerful does your computer have to be for the 64 bit double engine to be beneficial?

 Your computer specs should be fine. Click the box and see what you think.

Mosvalve, it isn't a question of having a computer with a particular spec.
 
Basically, it just handles data to a much higher precision. For most of us, it is the same precision of our processors, so there is little to no extra cost to the increased precision. Almost an unmeasurable increase when using a computer with a 64 bit processor and a 64 bit OS.
 
So, unless you are frustrated by slow bounce times, or you are getting dropouts at sample buffers you prefer, you can use it without really compromising anything.
 
On the flip side, some very smart people argue that the additional precision is beyond the hearing accuracy of humans.
 
I'll tell you, I check it and leave it checked.

I tend to leave it unchecked for playback, but do use the option during Export just in case there's some combination of tracks, synths, FX, routing, etc, that may benefit... even if it's barely noticeable.   Why not?