but are the 2Mpts at Siglent really better now or is that just marketing ?

Of course it's better, because it means much more versatility. The number of FFT points (aka FFT length) determines the frequency resolution for a given FFT bandwidth.

Look here:

https://www.eevblog.com/forum/testgear/rohde-schwarz-rtb2002-vs-siglent-sds2104x-plus/msg3239832/#msg3239832It means that the Siglent can have 16 times finer resolution in the frequency domain.

It also means that you can have 0.1 Hz frequency steps, hence about 0.4 Hz RBW (resolution bandwidth) for 100 kHz FFT bandwidth. The max. FFT length is selectable on the Siglent, so you can go down even to 64 points if you don't need any higher resolution - in short, you have the full versatility.

At the moment I would be interested in FFT in the audio (up to 20Khz) and I know that a soundcard is actually better suited for it ;-)

Both the Siglent and the R&S scopes have excellent low noise frontends, but as with all this modern split path input buffer designs, this is only true for frequencies above some 100 kHz. Below that, noise rises significantly. This has to do with the input protection of the LF/DC path, which has to withstand e.g. 400 volts. So this is certainly not an microphone amplifier design...