I remember people used to recommend this because it was easy to convert to 44.1 for the final, but gave the benefits of higher sample rates.
 
I never see it mentioned any more.  Am I just not looking in the right places?

I tend to record at double rates like 88.2 or 96 most of the time. When I record more than 8 tracks I usually drop back to 48 or 44.1 just because bounce time stops being worth it to me. I get impatient.
 
The math involved in smoothly converting from 96 to 44.1 is definitely in place in Sonar, so the even division from 88.2 to 44.1 no longer really matters. imho.
 
With my interface, I really can't tell the diff between 88.2 and 44.1 from listening, but why not use a double rate even if benefits are only theoretical if there is no noticeable cost or inconvenience to it?
 
My brother pointed out to me that none of my microphones have a high enough quality spec to be able to push the 24 bit 48K sample rate to it's limits. In other words, the noise introduced by the microphones exceeds the noise introduced by the interface so significantly that I couldn't actually assert the specs of the interface with the microphones I have.
 
Craig Anderton has made it a point that synths sound better at higher rates. This is noticeable. For this reason, we now have the feature to upsample for FX.

BobF
I remember people used to recommend this because it was easy to convert to 44.1 for the final, but gave the benefits of higher sample rates.

It hasn't become any harder to do the math over time, but computers are probably now at least 100 times as fast. Since there's no need to try to conserve CPU now, converting between 96kHz and 44.1kHz (or whatever else) can be done without sacrificing audio quality in any way.

This gets batted around quite regularly, but the input (24/96 or so) and output (16/44.1 if going to CD) is locked, as it were. It is what goes on in the middle that may or may not matter depending on effects/processes used, and how they are coded (invisible to us end users). Preferences and/or experiences then come into play, so nothing about this will be "cut and dry" for some time to come.
 
Practically speaking, if your computer has the horsepower and storage to process at higher rates, SONAR's dithering/SRC is more than capable to get you from point A to point B (the middle is really what you prefer).

I've always been in the camp of "if you can't hear it, why worry". I've done lots of testing on this for myself to the point of having other people load up the files to see if I can tell the difference. See, that's the key.....you can't do this yourself. It has to be totally blind. With a good sound interface, you shouldn't be able to tell between 24/48 all the way up to whatever your interface goes to. You *may* be able to hear a difference with 44.1 but even that is a little shady.
 
Those that can hear a difference most likely have cheaper interfaces. The cheaper ones WILL sound better using higher sample rates because....well, you get what you pay for. I have an RME Fireface, Layla 24/96 (there's an oldie) Audio Fire 12, and an Apogee Ensemble. I can't hear any differences that sway me one way or the other. This whole math thing (though valid) means the same to me as comparing some piece of analog gear to a plugin. You know, when you do the comparison and listen 12 times to the point where you *THINK* you can hear a difference that is so minimal, you're either trying to convince yourself or justify you made a $3000 purchase that just isn't valid enough? Listen 3 times...if you can't hear a noticeable difference....there is no difference worth wasting time on. But that's just my way of thinking. I just never got the science end of the recording field....and I'm so glad I didn't. I'd spend more time reading and splitting hairs over stuff that would take me away from actually getting something done. ;)
 
-Danny

only bats can hear the difference.

batsbrew
only bats can hear the difference.

Yeah, batsbrew that is! LOL! :P

My last CD recording (Contrabass and Piano) was done at 88.2kHz.  Just because I felt like it.  Bass itself rarely reaches beyond 10kHz in the overtones,  I could have recorded at 32kHz and had room to spare were it not for the piano.  I almost never go up to 96kHz and I've only recorded at 192kHz once in my career.  Complete waste of effort.  I personally can't hear any difference.   Like what Danny mentioned, back in the day of cheap consumer interfaces with unstable crystal oscillators and poorly designed low pass filters, running the interface at a higher sample rate increased the performance slightly.  I can't think of a single product on the market today where that is still true.   
 
The whole 88.2 easier to downsample to 44.1 is bogus and always has been.  Just because we find it easier to divide the number by two in our heads doesn't mean a computer has trouble with the math, especially with a 32-64bit floating point engine.  
 

rumleymusicThe whole 88.2 easier to downsample to 44.1 is bogus and always has been.  Just because we find it easier to divide the number by two in our heads doesn't mean a computer has trouble with the math, especially with a 32-64bit floating point engine.  
 

Actually that's not true - the math does indeed require much more processing power for dissimilar rates vs. integer multiples. 
 
But again, processing power is just not a limitation today vs. 15 years ago.

Actually that's not true - the math does indeed require much more processing power for dissimilar rates vs. integer multiples.

 
Not with modern asynchronous sample rate conversion it doesn't.  It is a more complex process, but is much higher precision than older synchronous methods and there is no measurable difference between simple or complex ratios.  
 
96 to 44.1 will take more processing power than 88.2 to 44.1 simply because there is more information.