For ME, even if my interface supported 192k sample rates, I would still opt for 48k, and possibly 96k.  I choose 48k as it doesn't drive my system to the edge - it just works without complaints (dropouts, etc), and I have no issues with the sound quality I get at that rate.
 
I am a keyboard player, and most of the commercial sounds out there are only sampled at a bit-depth of either 16 or 24 bits and a sample rate of either 44.1 or 48k.  So for MY particular needs, I would  be making my system work harder for no significant gain, as capturing audio from samples that started as 44.1 or 48k rates would not get any enhancement by running Sonar at 96k or 192k.
 
Vocals, live drums, or sounds generated by synths that are NON-sample based would have greater resolution at the higher sample rates, but since I cannot detect a difference, I don't bother with anything over 48k.
 
Higher sample rates also generate HUGE and MASSIVE audio files, and it takes WAY more time to process things at those rates.  So, again for ME, I prefer to have things take less time, and I don't have to worry about fussing with things - everything on my system is cheerfully tooling along at 48k.
 
Here is a link to some thoughts on higher sample rates and other white paper topics:
 
http://www.lavryengineering.com/lavry-white-papers/
 
Try some experimenting at different rates and see what makes sense to you, balancing out the potentially better sound against the performance and time costs of running at higher sample rates.  At the end of the day, it boils down to that which makes sense to YOU, for YOUR system.
 
Bob Bone
 

Hi Bob, absolutely. As it happens, I only use 48kHz most of the time for the reasons that you've already explained. I was just really interested in why the latency would drop when (as you rightly say) higher resolution audio is synonymous with a higher use of resources.
 
It has been beautifully explained by Jim on page 1 :) I will check out the white papers though - thank you.
 
 

wizard71
No it probably isn't right, I googled it :-)

:-D
 
++ Great post from Jim.  Is interpolation a factor d'you think?

doncolga
In the big picture, is the lower latency impacted by the machine working harder @ 96K?

Yes, the higher the sample rate, the greater the load on the CPU.
The higher the sample-rate, the less time the CPU has to fill the buffer (ASIO buffer).
If the buffer isn't filled in time, a dropout or glitch will occur.
 
There's no absolute right/wrong answer when it comes to sample-rate.
I believe I can hear a more 3D soundstage and more detail when using higher sample rates (88.2k/96k).
That said, I don't think any record has been (or will be) bought because the sample-rate at which it was recorded/mixed.
There are far too many other important factors.  The first (and main) being the song itself.