Entering conditions:
1. I have a delta-sigma DAC so it plays 1bit data (DSD) better than multibit data (PCM) due to this article:
https://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/2. I encode my hi-res records from PCM 48k24 to DSD128 (5.6M) by JRiver MC25 on the fly seamlessly without any unpleasant clicks between the tracks.
3. Focus Assist is on to eliminate interference from pop-ups, USB DAC is controlled by JRiver only in Exclusive Mode, in order not to have some additional sounds from other apps.
4. CPU is usually loaded by 15%-20% which is bearable, no fan sound is heard.
Resulting facts:
1. The thing you notice at once is that all cymbals become more “sparkling”, “airy”, “analog” and natural. The “curious explorer” inside me immediately demands for a logical explanation of the fact and fails (probably my knowledge of the subject lacks something). Interestingly but conversion to DSD64 doesn’t give the same effect, it seems that the 2.8M noise-shaping and phase of the cutoff filter is too close to the 20-20,000 Hz range. Conversions to PCM with higher sampling frequencies of 88.2, 96, 176.4 and 192 kHz don’t add this perception of the cymbals.
2. The stereo panorama of the track becomes how to say, more spacious, predictable, linear, natural. It is especially heard on orchestrated tracks, classical or progrock. It can be attributed (I suppose) to a better phase handling by a DSD data and 1bit DACs. But the fact is that the track becomes more “ear-friendly”.
3. The bass lines of the tracks where they especially low and fleshy becomes more “tasty” and natural.
4. The singer / singers voices become more human like.
5. The overall perception is much closer to a very good analog tape reel-to-reel recordings than their PCM version.
Conclusion:
1. Having thoroughly studied “Coding High Quality Digital Audio” by Bob Stuart of Meridian Audio I personally agree with Mr. Stuart and think that hi-res audio music information should be digitally encoded reasonably in accordance with average human hearing specs. I personally think that 16bit encoding ladder is insufficient, 24bit encoding ladder is quite enough, and not only due to difference in dynamic ranges 96dB vs. 144 dB (moreover the latter is painful for a human ear). I also personally think that the closest sampling frequency to a Reasonable One is 48 kHz, Fs 44.1 kHz / 2 is too close to the 20 kHz end of human hearing, and the cutoff filters may affect the phase of sound, and higher sampling frequencies music programs may contain some ultrasound noises which can result in unpleasant shade of cymbals due to non-linear behavior of amplifying elements. So I think that a digitally encoded music program with 48kHz / 24bit specifications and by a professional ADC is reasonably sufficient.
2. But, the music program must be also decoded properly. PCM data is better handled by R2R ladder DACs, that’s quite clear. But they cost a lot. Delta-sigma DACs are more affordable but they are 1bit decoders. They certainly decode PCM but the algorithms inside them seem to be simplified or better to say compromised. (I don’t know for sure), but the result is clearly heard. The notebook CPU can encode 1bit data with more precision and proper filtering, so if you have a Delta-Sigma DAC it’s better to feed it with DSD128 at least. DSD64 is insufficient. DSD512 is probably better (or even DSD1024), but my DAC doesn’t support this, so I simply don’t have any listening experience, and CPU will be probably closer to overload. So to me DSD128 is reasonably enough.