Interfaces and Sampling Frequency

I am shopping for a new interface to go with Windows 8 PC and SONAR Platinum.  I've always recorded at 44.1, but I'm seeing 96 and higher.  What is the rate most people are using, and what are the considerations?

^ Colin has nicely summed up hundreds of pages of debate in the above post.
 
If you took a poll, you'd find that the vast majority use 44.1 KHz, a large minority use 48 KHz, and a smaller minority are at 96 KHz. A few clueless noobs use 192 KHz on the theory that larger numbers must be better.
 
With regard to interfaces, most of the devices you and I can afford are specifically optimized for 44.1, even if most are capable of higher sample rates. One point to bear in mind when choosing an interface is that the highest sample rate it supports says nothing at all about the product's overall quality. 

Bob, the only "interface" that did internal sampling conversion I know of was early Creative Audio Audigy.  They did the conversion in real time, and many of my old songs I dubbed from vinyl still have the click click click of that internal conversion of 48 to 44.1 for CD.  Otherwise, I haven't heard of that problem.
 
And optimized or native rate are relative.  Interfaces might work better at one rate and most pro-consumer interfaces we use strain to use 192, but a 96 rate ought to work all right.  That doesn't mean the audio recorded will sound better.
 
So many home recordists  stick to 44.1  Conversion between sample rates used to be a black art, but it is pretty well-written code these days unlike back when, when turning your 48 kHz recording into 44.1 for CD could produce artifacts.
 
48 kHz was and is a native rate for digital video.
 
The optimum for recording PCM, according to Lavry and white papers back when is around 60 kHz.  Not sure if that still holds since technology moves on and he had some specific arguments about filtering, I believe.  20 kHz should capture everything you can hear, but the slopes etc. of the digital filtering can cause artifacts.  So the answer is raise them higher.  But too high and you get info you can't hear, waste space and I believe another reason I couldn't follow even in my simple way.  32 kHz was a 4-track audio rate for digital camera - if that had been doubled (like 48 to 96 kHz) we could hear if there was any difference.
 
Many professionals use 96 kHz - because they can.  It is becoming more of a norm and does since that is the new standard for video, etc.  They can also afford arrays of solid state drives to archive a single song on and transfer around.  They also have SSLs.
 
If you have the disk space, there is no reason not to record at 96 kHz even if the audio doesn't sound better than 44.1.  Good for archiving.  But there isn't much of a reason to either, if you can't hear the difference.  Most software these days do the internal oversampling which was one reason to work at higher rates.  Synths and effects (w/ a lot of math involved) often upsampled to 32 or 64 bits and or higher rates so the mathematical rounding errors were very small, which helped the sound.  SONAR does internal 64 bits math and now can use oversampling with software that doesn't.  So a lot of the reason to shoot for 192 kHz is gone.
 
When I interviewed Rupert Neve, who knows a few things about sound, said he thought the problem w/ digital was the low sampling rate.  Of course, RND doesn't really do digital while Lavry does, so the experts here are 1 to 1.  Just use whatever is the lowest rate you can't hear a difference in.  There are good arguments for either side.
 
@

I record at 48kHz 99% of the time.  Even for high end album releases.  I choose it because of the occasional accompanied video, it moves the LPF slightly above any possible audible range, and my computers don't have to chug so hard to work with it.  Since I have never been able to hear a difference between 1, 2, or 4 times sample rates, since all that does in record frequencies I cannot hear, record, or play back, I don't see any trade-off.  Every now and then I will go to 88.2 if I feel there is a potential to release the material on a more modern format like Blu-Ray Audio disks or HD downloads.  A marketing concern, not a practical audio concern.  
 
Pretty much all modern converter chips I know of sample at ultra-high DSD frequencies and convert that to PCM at a later stage.  The Tascam rep who posted here recently about the DSD capabilities in Sonar did mention that the best reason to record in DSD was to let the expensive software handle the conversion process to PCM rather than a $5 chip.
 
If you are stuck with PCM, however, bit depth and analog stages are infinitely more important to the sound.   

I think it's safe to assume that prosumer-class interfaces are optimized for 44.1 KHz, and that high-end interfaces such as Lavry's are optimized for 96 KHz. However, the difference in performance is going to be negligible for most material. I wouldn't base a purchasing decision on it.