Upsampling is a gimic to me - I have heard both upsampling DACs and non and prefer the non-upsampling DAC. Also, if you run bit perfect - what you are haring is what is in the file or disc.
Unless you are using a pure binary, R2R (or equivalent), 44.1kSa/s, 16-bit DAC, you are probably hearing some upsampling.
Upsampling isn't a gimmick. It's a technique that allows a designer to use a more gentle reconstruction filter, rather than a brick-wall filter. This has benefits for amplitude and phase integrity, and signal-to-noise ratio of the reconstructed signal; that sounds like a good idea, right...?
https://en.wikipedia.org/wiki/Reconstruction_filter
An upsampling, digital interpolating filter is used to 'join the dots' between the samples coming off the digital source. This moves the alias image up in frequency, and spreads the frequency of the output quantisation noise, reducing the noise power spectral density (the same noise power is spread over a wider bandwidth, because the output steps occur more frequently (e.g. 2x, 4x, 8x Fs etc.). This means your reconstruction filter bandwidth can be pushed out from the input Nyquist frequency, and, since the input samples contain no information above the input Nyquist frequency, the reconstruction filter can be designed to roll off gently from the input Nyquist frequency, as the alias image is now much further from the wanted signal. Result: more linear amplitude and phase performance of the reconstructed signal, and reduced quantisation noise.
Fairly good coverage of the noise sources in digital signal conversion here (page 44 onwards)
https://www.intersil.com/content/da...ulators/noise-in-the-signal-chain-webinar.pdf
This gives a good explanation of the benefits of upsampling filtering:
https://www.embedded.com/design/rea...cks-Reducing-A-D-Converter-Quantization-Noise
Or one from Analog Devices:
https://pdfs.semanticscholar.org/3716/a58310339cb34efd19234c132990f8f8ef78.pdf