In attempting to help another poster, I discovered this on my own system:
 
Under Preferences | Audio | Driver Settings, my Audio Driver Bit Depth is 24.  That drop-down is grayed out.  Under Preferences | File | Audio Data, it says I am at 32 for Record Bit Depth and Render Bit Depth.
 
Is that mismatch a problem?
 
Note: I am using a VS-700 with the audio interface set to 44.1 (just in case that matters).

No audio interface can produce or work with 32 bit audio. Its a A/D D/A converter limitation.

The audio driver bit depth is the communication with the audio interface. I don't think there are any 32 bit audio interfaces available. 24 bit is common.
 
The audio data record bit depth is how the file is written to the hard drive. You can write deeper than you have available from the audio interface but it will pretty much just be wasted space.
 
The render depth is defaulted to 32, and it is for bouncing and freezing.
 
These topics are pretty well covered in the help file.
 
I imagine that the use of 32 bit to store audio is more or less wasting your hard drive space and your throughput.
 
The on line help says to use 24 bit for "soft music" or 16 bit for efficient storage.
 
But I also imagine if you take a bunch of 24 bit recordings and add them together, you might loose data due to truncation. Or perhaps some effects add data beyond the original bit depth. I don't really know.
 
About 10 years ago I tracked a drummer for a bunch of tunes, after I upgraded all my equipment. I really wanted 24 bit, but somehow recorded the whole thing in 16 bit. Must have hit a wrong button.
 
You can check the bit level of a clip by right clicking it and going down to the Associated Audio Files item.
 
Personally I like the 16 bit 44.1 option, for computer performance and reliability. I mix to MP3, so for me high bit rates and depth are a waste.

32 bit is a high-rate format for working and saving a file.  As others have said you can't record 32 bit files - even the best hardware doesn't have 144 dB of resolution, much less more.  Commonly, you work on a file, esp. a stereo master file, at that rate so you have all the headroom for the software to do the calculations with the rounding errors pushed way down (the same reason you should use SONAR's 64-bit mix engine and why the better software effects back when upsampled before doing their calculations).  You typically store it at 32 to do further work or archive it.  It isn't common, as far as I know, to do multiple tracks at 32 bit, tho there is no harm except for hard drive space.  If you make a CD you down sample it to get to 16 bits (or even further for mp3s - note: do your CD version first if you are converting from a downsampled file not the rougher mp3).  It gets dithered (final final process) and what not, basically chopping off the bottom 16 or 8 bits worth of errors and noise at the very bottom of the sound, so what your interface receives is high quality stuff.
 
That is best practice, but many systems don't provide and most people can't hear hardly any differences between the format.  High bit rate recordings also give you slack, however, with a lower noise floor so you don't have to slam the sound into your converters in order to cover up the noise (as in the analog tape days) inherent in the system.  It gives you a wider margin of error to use, which one day on your most important project you'll appreciate when you don't miss the perfect take because of digital overs when the singer/guitarist/hornist finally warms up and is louder than what you set them for.  That is why you use 24 bits- so you can record at a lower volume, too.

Not completely accurate regarding the driver bit depth. There are drivers that support 32 bit float (RANE is an example) and potentially also 64 bit double precision. The ASIO spec supports it and so does SONAR.
The reason drivers may support float is if they have internal mix buses or have onboard FX that can be in the input signal chain. In such cases they may support float to avoid loss of precision before the signal hits the DAW inputs.

Gee, I'm just a low IQ songwriter. Should I change those 32s to 24s?

konradh
Gee, I'm just a low IQ songwriter. Should I change those 32s to 24s?

24 bits is really all you need for recording and rendering, and even then you're bumping up against the practical limits of audio interface record/playback capabilities. All the calculations that happen in SONAR are 64-bit anyway, which is where the extra resolution matters the most.
 

Thank you, Master Anderton.  Someday perhaps I will be able to snatch the pebble from your hand.
 
https://www.youtube.com/watch?v=j1wjvP-raOI
 

Ah grasshopper.