It's only that I read that people do oversampling while from what I understand it's useless.
The only reason I can think of is reducing jitter on a synchronous physical interface like SPDIF, i.e. sending a 44.1/16bit stereo signal only needs 1.4mbits/sec, while 352.8/32bits stereo needs 16 times that (22.6mbits/sec).
Higher bandwidth allows for a more accurate clock recovery, and I think it is
necessary to increase the bit depth when you oversample to avoid quantization noise. Simple example: 2*upsample a straight line through -1,0,1 and you would need -1/2 and +1/2 as values, i.e add another bit.
Interestingly, even very old DAC chips like the BurrBrown
PCM58P internally use a similar mechanism: they sample an input stream of 44100Hz/16 at 176.4kHz and operate on 18 bits internally.
All of this was my motivation to buy a USB DAC, this completely removes the transport jitter problem and all upsampling is done inside the (very capable) DAC chip - simply feed it the original input data unchanged -- and then we're back at DSD: when you have DSD source material you want to feed it directly to the DAC and be confident that the DAC processes it in the proper domain.