Sample Rate and Softsynths

I am sorry if this is covered somewhere, but I couldn't find exactly what I am looking for.
 
I understand the benefits of recording and mixing in 88.2 or 96 Khz as you will get more accurate recording and editing, even if publishing finally at 44.1k/16 bit, however, I do have some confusion when it comes to recording softsynths.
 
I create music mostly with softsynths but sometimes external synths and other instruments as well.  Should I be starting out in a 96k project even when programming up the softsynths?
 
I also would like to know how Cakewalk's internal upsampling feature comes into play here.   I don't understand its use-case.
 
I am just trying to get a proper workflow going here.  I don't mind recording my synths in one project and then bouncing down and importing the stems into another "mixing" project if you will.  I just want to know what is best here.
 
I have a very powerful PC which can handle anything I throw at it, if that matters.
 
Thank You!
Adam

higher sr should give a more "accurate" sound, indeed some synths recommend higer sr (NI's monark defaults to 48 iirc)
try it - bounce a synth at lower and higher sr and compare - for some it won;t make any difference, but for others it will
can't comment on sonar/cbb upsampling feature

As far as CW's internal 64 bit use, you don't need to worry about that. CW takes care of it's end.
 
The main thing to remember when starting projects at 96 (I use that bit depth) is when you do go to downsample. you need to dither. And when you downsample depends on how you work and what your intended work is meant for.
 
I start with 96 and keep it there as long as I can. Most of the time that is right before I get a wave to burn to CD. Then I downsample making sure to dither at this time.
 
This is all general info and there is much more info to be had on the internet. And others that know more than me may pop in here and give you the scoop.

It doesn't really matter, tho it is best to keep everything at the same sample rate etc.  Higher rates used to strain computers, but nolonger.  48 to 44.1 conversions used to cause problems, but no longer.  Many pro studios use 96, tho my favorite studio in Dallas sticks to 48 and his stuff doesn't suffer.  It basically matters a lot less than one would think.  SONAR has its own built-in "upsample" process to insure there are few overs when processing effects or mixing channels.  Turn on the 64 bit engine.