Advice needed - How important is POWR dither (vs Triangular) for audio exports?

Dithering should only be applied at the last stage of mastering.
If you recorded at 24 bit, no need to dithering for exported 24 bit tracks.
You would only user dithering when downsampling, eg. . . from 24 bit to 16 bit.
When you need to dither, Powr dither would be preferred over Triangular.

A lot depends on the type of music you're doing. If you're doing classical sessions with acoustic instruments that exhibit a lot of dynamics in a complex acoustic space, you might notice a difference. For pop music, few can tell the difference between dithered and undithered material, let alone the difference between different dithering options.
 
Furthermore, with 24-bit audio, the issue becomes pretty much moot as any rapid switching in the lowest bits will almost certainly be happening below the noise floor of any associated audio electronics. I'd be happy to stand corrected if anyone knows more details about this.

I'll always use dithering when going to 16 bits as noynekker suggests at the mastering stage -- I've read somewhere (maybe the R. Izhaki "Mixing Audio" book?) that there's enough bit reduction going down to 16 bits that dither is typically worthwhile in that scenario.  
 
I think 24 bits is a bit more up for debate (as Anderton suggests, at 24 bits maybe any of the bit action is happening at such low levels that it becomes moot).
 
The question for me really is this: if I'm going from 32 bits (VSTi source file) to 24 bits, is 1) better to use Triangular dither OR just leave dither *off* and 2) maybe more to the point, will anyone even notice when I'm doing a pretty full production (i.e. not classical or something similarly dynamic).
 
 

To answer the original question: it's not important.
 
Use triangular dither when exporting to 24 bits, it'll be fine. Or don't dither at all. That'll be fine, too. But please don't upgrade to Producer for PowR-N dither alone.

Me thinks that a 32 bit VSTi has nothing to do with audio depth, rather the code running as 32 bit (originally optimized for 32 bit OS).  The important question is, the output resolution of the VSTi is sample rates.  For example, most KONTAKT libraries have 24 bit output resolution (the instruments were sampled at 24 bit).  Outputs from virtual synths that DO NOT use samples, are also usually 24 bits but not always.
 
So, for your question, the fact that your VSTi is a 32 bit plugin would only be material if its output audio resolution isn't 24 bit.  
 
When your project is complete and then you mix down to 16 bit (CD quality), then some type of dithering would be in order, in which Craig's advice would be appropriate.
 
Caveat:  I think I know what I am talking about but it wouldn't be the first time that I don't know what I think I know.
 
Jim

Anderton
The easiest way to hear what dithering does is to reduce the level of a piece of audio to -85dB, or whatever level gives you truncation. Then you can apply dithering and hear the difference. I did an audio example for an online webinar using an orchestra, and you could really hear what the dithering did. Realistically, few listening environments allow hearing those kinds of low-level signals but when you can, it helps. It's not something you "hear" as much as "sense."

Thanks for confirming my idea. I've heard what it does in situations created artificially to showcase it, but I have never been able tell any difference with real world use for my mixes.

bitflipper

I think I can go forward now with the reasonable assumption that if my mixes suck, it won't be due to dither.

Very wise conclusion, David. It also won't be because you haven't spent enough money on plugins yet.

I know this to be true as I have tested it empirically over the course of many years ;)