Upsampling doesn't create a better quality, the quality is determined by the original sample rate of the source.

Besides that, upsampling from single to double speed can solve issues if you feel limited by the treble roll-off of certain D/A filters, see the D/A filter frequency curves in the ADI-2 * manuals.
By this, you move the treble roll-off to frequencies much over 20 kHz, which exceeds our hearing capabilities.

I would only do it if I hear a difference. I do not feel limited by the SD-Slow filter, so I do not care much about it.
With NOS, the treble roll-off is easier noticeable.

Upsampling has its place if upsampled signal is further processed by nonlinear effect. For pure playing it could be advantegous if nonoptimal DA converter is used. As Ramses already mentioned. DA converter with poor output lowpass filter would benefit from upscaled signal.

But it is not case of RME D/A. So, possitive effect of upscaling is rather quetionable. I guess, RME has added it there to please HiFi customers, not because they would see any real benefit in it....

But if someone hears the difference and upscaling sounds bettter to him. Why not to use it.

The “Pro”-series ADI’s are primarily made for studio use (which doesn’t exclude private), and here the option of having an asynchronous sample rate input can be a life-saver:

E.g. to digitally connect a player device without wordclock input there’s no other option.

The other point:
Quality nonlinear digital effects typically have their own oversampling built in.
Only in the early times of digital this wasn’t possible or ignored for various reasons.

I got my 1st ADI-2 Pro after the purchase of an Accuphase E-600 Class A integrated amp with Accuphases own DAC module.
This way I was able to make A/B tests with only a little switching lag.
My HiFi speakers work precise so it supported comparisons well.

The ADI-2 Pro with NOS D/A filter sounded very similar to Accuphase's DAC module.
Between NOS and other D/A filter it was easy to distinguish for me.

But I am fine, I am not worried in any way to miss something. At the end all sound great.

But still, very old HW, old plugins, one develops plugin, but does not want to fiddle with oversampling at beginning... One runs HW with limited CPU capabilities.

And yes, it is possible to have powerfull computer, but many do not have it.... There are constant complaints about dropouts on Native Instruments forum.....

But you are probably right that person who has ADI-2 Pro has also good computer. But still, I am able to fully load 8C/16T AMD Ryzen 7 (4+ GHz) with just 16 synth patches. It is not that difficult to run out of CPU....

I am aware, one could have 100+ core workstation with up to 1 TB of RAM. But it is not common, it is just possible.

That’s not entirely true, as the upsampling process needs anti-aliasing-filters the same way like a DA-converter.
That is 1/2 the frequency of the original sampling rate.

If one’s heading for the best qualitiy, off-line (opposed to real time) resampling’s has an advantage:

Filters with much higher filter coefficients can be used, as there is no limitation in processing time.

“Longer” filters improve audio quality by something amount, but can’t completely remove their side-effects, as frequency response, phase, and impulse response is still connected physically.

And no upsampling process can bring back what has been cut away in the first place.

I’d agree with you, upsampling just for listening typically is not worth the effort.
Specially in the light that ADI-2 has a selection of the DA-filters to choose from, where are you can find one that sounds good to you.

I‘m not sure about it‘s design, but Voxengo R8brain offline SRC has an excellent audio quality and comes in a free version:

https://www.voxengo.com/product/r8brain/features/

I use it e.g. for tiny adjustments of sample rates of cameras and audio recorder after on-location video-shots, to precisely sync them.
Perfect for situations where it’s not possible to use SMPTE and/or Workclock sync.

Never heard any difference between original and resampled version in this application.

I found it when reading a comparison - which included compressive measurements - between multiple SRCs, it was found top notch, with no artifacts at all.

For daily use in the studio my DAW (Sequoia) has highest quality offline SRC built in, but for the special purpose above R8brain is easier to use, as it‘s fully parametric.
Like my DAW R8brain includes batch processing, so it can do a lot of individual files in one run.

Thanks for your answer, KaiS.

What about re-coding (I don't know if this is the appropriate term) from PCM to DSDx (x the higher the better :-) ) just for improving the listening experience?

Thanks KaiS!!

... I already have. Actually, I have read the manual cover to cover.