For a DAC that uses an oversampling filter in the DAC chip itself, you will often find there are a number of oversample rates you can choose, and different digital filter algorithms. Whether you can choose which of these to use is another issue.
So, the question would be whether the oversampling and filtering options in the DAC are better than those options provided by the equivalent functions in your media player software.
Oversampling is means to relax the requirements of the anti-imaging filter required on the DAC output to reduce the sinc function imaging inherent in a zero order hold sample data system.
Experiment with all the oversampling and filtering options available to you, and pick the one you like best.
This is what I finally did- I upsampled the 44.1k 16 bit Red book source to 176.4kHz 32 bit which I believe (please correct me if I'm wrong) provides a 0-22.05kHz signal, a 154.4kHz to 176.4kHz image and a 176.4kHz to 198.4kHz image (following the sinc function) etc. I did not, eventually, end up converting to DSD- I could not tell any audible difference from the upsampled PCM. I then used the Foobar2000 SOX/SSRC plug in to apply an apodizating anti-aliasing filter that remained flat at 20kHz, but rolled off gradually to be -146dB at c. 40kHz. This eliminated the supersonic noise prior to the image frequencies that seems to be generated by the upsampling.
The DAC reported a 176.4kHz PCM input rate and the built in filters which I believe are applied to the input stream and are frequency scaled appropriately, were applied relative to that, so presumably they are now operating c. 80kHz.
We could not hear any effect due to the built in DAC filters, but we were convinced that the upsampled output sounded better with the SOX/SSRC filter in place.
I'm still trying to determine if there's any rational way to perform a measured difference analysis of the original and upsampled bit streams before converting to analog.
I also tried upsampling to 176.4kHz 24 bit and 352.8kHz 24 bit and could not hear or measure any significant difference in the analog output with the DSP set as above, so I just kept it as it was at 176.4kHz 32 bit as the CPU didn't seem to be having any issues (15% utilization) and there were no glitches that could be heard.
By the way, the DSP plug in that does the upsampling/digital filtering can be switched in and out with a single mouse click- reverting the input to 44.1k 16 bits on the fly. The DAC reacts accordingly switching the reported input to 44.1k 16 bit PCM.
The difference in the supersonic behavior can be seen on a spectrum of the analog output, but the audio signal amplitude remains the same. There is a small click produced by the transition so it's not completely non-detectable.
This made doing an A/B test simple.