When just over-sampling, it is to reduce the analog filter requirements pretty much exclusively. Though there is certainly a "following" for non-oversampling DACs, most of us want to hear what is on the CD or in the digital source, and not artifacts that were never part of the music.
If your sample rate is 44.1KHz, and the audio bandwidth is 20KHz, there is only a 2KHz band between sample bandwidth (44.1/2 = 22.05KHz), and music 20Khz. To eliminate all the artifacts requires a very high order analog filter (many stages), which is going to itself create large phase artifacts in the audio bands.
When you upsample audio, you don't interpolate between samples (or shouldn't), you insert 0's between samples. 20-30-40 would become 20-0-0-0-30-0-0-40-0-0-0 (for 4x oversampling). This shifts the spectrum of the artifacts to 4x the sample frequency. Then you can use mainly digital filters to remove the artifacts. These digital filters can be something called linear-phase that does not wreck the relative timing of different frequencies. This is then followed by a much simpler analog filter which has much less impact on the phase response.
Almost all modern audio DACs chips have built in over-sampling. Most that build their own DACs, also use over-sampling, sometimes at very high frequencies. There are "claims" made w.r.t. some using FPGAs about the size of their digital filters and an improvement in "accuracy" but for all the hand-waving, good luck getting a number pinned down (to justify).
If your sample rate is 44.1KHz, and the audio bandwidth is 20KHz, there is only a 2KHz band between sample bandwidth (44.1/2 = 22.05KHz), and music 20Khz. To eliminate all the artifacts requires a very high order analog filter (many stages), which is going to itself create large phase artifacts in the audio bands.
When you upsample audio, you don't interpolate between samples (or shouldn't), you insert 0's between samples. 20-30-40 would become 20-0-0-0-30-0-0-40-0-0-0 (for 4x oversampling). This shifts the spectrum of the artifacts to 4x the sample frequency. Then you can use mainly digital filters to remove the artifacts. These digital filters can be something called linear-phase that does not wreck the relative timing of different frequencies. This is then followed by a much simpler analog filter which has much less impact on the phase response.
Almost all modern audio DACs chips have built in over-sampling. Most that build their own DACs, also use over-sampling, sometimes at very high frequencies. There are "claims" made w.r.t. some using FPGAs about the size of their digital filters and an improvement in "accuracy" but for all the hand-waving, good luck getting a number pinned down (to justify).