Many amp sims have an oversampling or "HD" option. Clicking these doubles the internal sampling rate, and makes a major difference in sound quality. This has nothing to do with extended frequency response, but about reducing foldover distortion and having greater accuracy in the calculations that define the distortion algorithm.
 
Now, TH2 doesn't have an oversampling option (at least that I've found), but lately I've been experimenting with running it at 96 kHz and I think it makes a substantial difference. I don't mean like "maybe I hear something, can you switch back and forth one more time" but something really obvious. Also, the latency is much lower because you're running through the sample buffers over twice as fast compared to 44.1 kHz. I can run at 96 samples and get a round-trip latency of 8.4 ms including all sample and hardware buffers both coming in and going out. The sample buffer itself is only one millisecond.
 
I'm preparing for my [shameless plug] workshop on amp sims coming up at Sweetwater's GearFest, and running at 96 kHz because it sounds like it makes such a major difference with the distortion. But I'm wondering what y'all think. If you run the TH2 distortion algorithms at 96 kHz, does it sound like a big difference to you?

Its not very easy for me to switch back n forth from 44.1 to 96, but I do hear a sonic difference from TH2.
Many thanks for that!

I think for amp sim standalone use live, the better sound quality and lower latency justify 96K if the computer can handle it.

So... if one were to bounce TH2 down while at 96KHz and bring that audio into a 44.1 KHz project, the results could be improved?     I may have to try that at some point...
 

stevec
So... if one were to bounce TH2 down while at 96KHz and bring that audio into a 44.1 KHz project, the results could be improved?     I may have to try that at some point...
 

Well, I did an experiment that was pretty interesting. To try and keep everything within a single project and avoid any sample rate conversion that might introduce another variable, I set up an experiment with Guitar Rig. GR has an HD option which does 2X oversampling internally; this has the same practical effect as running Sonar at twice the sample rate.
 
I recorded a track at 44.1 and duplicated it. One used GR as a processor set to standard. The other used GR set for oversampling. As expected, the HD version sounded better. 
 
I rendered both tracks, flipped the phase on one of them, and nudged the non-HD track a sample at a time to line up with the HD track (HD adds a tiny timing difference--not enough to notice, but enough to require nudge if you want to null). This revealed the foldover distortion/aliasing in all its glory. Now, you could say it was from the HD track, not the non-HD. However, it was clear that the HD track sounded better, so I think it's reasonable to assume that the distortion products were more part of the non-HD than the HD track. 
 
So the short form answer is that the improvement of running at a higher sample rate gets "baked into" the track, so when you play it back at 44.1, the improvement remains. Remember this isn't about extended frequency response, but reducing artifacts.

I got questions but don't know what, so that must be a confusion.
 
Is there any relationship oversampling on 64bit compared to 32bit O/S?
 
I have tried hitting the oversampling mode of various plugins but I don't hear any differences?
 
I will give TH2 a try with oversampling. thanks
 

Interesting. Did the same occur for AmpliTube?
 
It seems like this may well extrapolate out to just about most virtual instruments.
 
If that is the case, run 96 on EVERY project.
 
Is my logic on point? 

BJN
 
Is there any relationship oversampling on 64bit compared to 32bit O/S?

 
No, it's strictly a sample rate thing.
 

I have tried hitting the oversampling mode of various plugins but I don't hear any differences?
 

 
It's more obvious on some sounds that others. Distortion algorithms is where it makes the biggest difference.

hockeyjx
Interesting. Did the same occur for AmpliTube?

 
Amplitude, Guitar Rig, and Waves G|T|R all have oversampling buttons. With GR, it's the HD button to the left of the NI logo. AT has multiple options to let you do oversampling on various modules. G|T|R also has an HD button. Doubling the sampling rate with these when oversampling is already in play seems to have no effect. I don't know if it's because they top out at 96k, or whether past a certain point it just doesn't matter. I'd need to test further to find out more.
 

It seems like this may well extrapolate out to just about most virtual instruments.
 
If that is the case, rum 96 on EVERY project.
 
Is my logic on point? 

I don't know yet, but the same thought occurred to me. I think the big question is what kind of sounds exhibit an obvious difference. Guitar distortion seems to be far and away the most noticeable. Perhaps synths with lots of highs would react similarly, but I haven't tested yet. The one thing I know for sure is doubling the sample rate can make a difference with amp sim distortion algorithms.
 
The advantage of doubling project sample rate instead of just enabling internal oversampling within a sim is that the latency is lower at higher samples...of course, you're also stressing out your computer more, and can't stream as many signals to and from an interface. 

well then it goes that trying 192 and beyond may create even more awesomeness