It´s important to separate music making and playback samperates. 44.1kHz/16bit is perfect for playback only. 96kHz/24bit is perfect for making music.
All non-linear processing will benefit a lot of 96k samplerate. Also audiorate modulation is beneficial. If you have the CPU power then go for it.

there are a few links on the web discussing why the double blind test that have been conducted may not have been productive methods of determining whether humans can hear or identify the differences in adopted sampling frequencies for audio. not saying they have merit but may be worth reading.

@Craig Anderton: You did not mention the [x2] button (Upsampling) that has been introduced in Sonar.
Do I understand correctly, that if in doubt, using this feature will solve the VSTi Synth/Sampler artefact problem, allowing users to continue recording at 44k1 (or 48k)?

yeah, what jpetersen asked? I'm intrigued. I have a project that is near completion that was done at 44.1. At what point do I upsample and where is that option? BTW I'm on lexington, just haven't used that feature yet.Is it in the export dialog? or on the front end in preferences? I didn't use any soft synths, but I have lots of plugins on it.Only synth was hardware yamaha ex5 that I used guitar controller and breath and foot controller with their vl engine to achieve a very believeable tenor sax sound (don't tell any sax players that, they ask "who played" I tell them me THEN they say "yeah, I could tell" when what they were really were asking was " why didn't you use me instead of that guy?")That,of course, was recorded into sonar at 44.1 also.

I can tell you where it is: It is (on my Lexington) in the cluster of buttons to the right of the light blue time display.
Hover over the button marked [2x], bottom row, third from the left.
The tooltip states "Enable/Bypass Plug-in Upsampling"
 
But yeah, it would be good to get some guidelines on it's use from those who know.
Should I just turn it on the moment I use any VSTi?
Are there any drawbacks to taking this naive approach?

mikannohako
As far as I know,
There are pros and cons.

 
Good job on the research, which also underlines what I said about the validity of upsampling during recording.
 

Some plugins internally process at doubled of multipled sampling rate (oversampling) to avoid aliasing or that.

 
Yes, and if set to oversample, these will generally not benefit from upsampling. However, oversampling does take more CPU, so for example in an amp sim, the amp itself may be oversampled but not the effects. Also upsampling sometimes has other benefits. The reverb imaging in TH2 "wanders" a bit when recording at 44.1 kHz, but not at 96 kHz.

a lot of information that we can't hear, lile above 20khz, these unnecessary information produces more IMD. So this is a downside. To avoid this, you can use a filter to cut these frequency before you put non-linear process.
So with higher sampling rate with unnecessary information, you might get even worse result because of IMD. I read this thing on gearslutz forum.

 
This is true, but very rare.

For SRC, I think you eventually release your songs at 44khz, so if you produce your music at other sampling rate, you need to convert the sampling rate with SRC. You can use SRC on your DAW but the quality of SRC is different on each program. And whenever you use SRC, the sound get worse and peak changes. (only just slightly and no one might notice though) anyway, usually SRC in your DAW is not ideal, so if you don't have decent SRC, the track converted to 44khz may sound worse than one with 96khz. IFAIK SRC on ableton live is great among DAW. SRC by Saracon or izotope are considered to be best I think. You can check the quality of SRC on SRC comparison website.

 
If you look at the comparisons at Infinite Wave, SONAR fares extremely well compared to other programs. For example, compare it to Cubase, Pro Tools, Reaper, or Vegas.

Reverb or other process may act differently on different sampling rate but I just don't know it.

 
Yes, it can.

I currently master at 96khz with ultrasonic filter to lessen IMD but for other thing like mixing or producing. I don't know which is better because higher sampling rate seems to require more CPU.

 
This is the best part about SONAR's upsampling, you can get the quality of recording at 96 kHz with 44.1 kHz projects.


Thanks again for the concise summary.

jpetersen
But yeah, it would be good to get some guidelines on it's use from those who know.
Should I just turn it on the moment I use any VSTi?
Are there any drawbacks to taking this naive approach?

The main drawback is that upsampling adds time to the rendering process.
 
There's no need to turn it on the moment you use any VSTi, compressor, limiter, etc. If the VSTi already does oversampling (most modern ones do, or give the option to do so), or doesn't generate energy above the Nyquist frequency, upsampling is not needed.
 
After upsampling was introduced, shortly thereafter SONAR introduced upsampling on playback. This makes it easy to determine whether something sounds better or not when upsampled. For example, I've found the SI-Strings sound better when upsampled. Also, some amp sims sound better when upsampled, even if they do internal oversampling...haven't figured out why, though, or whether the sound is perhaps "different" instead of "better."
 
Bottom line: Use your ears

Wow, what a great thread. 

the way i tend to view things and ask is how well the anti aliasing filtering has been implemented, in the software and hardware domain, the daw its self, the plugins, sound interface, so many variables

naturally all anti aliasing filters aren't gonna be created equal