You can download the rockbox simulator from google and use the text_codec plugin (or anything that can capture windows audio) to record the output losslessly and test it yourself. You don't need to have an actual player, at least not until someone actually implements 24 bit output on a real device.
This is probably not something you have to worry about. The actual decoding is done at much higher than 16 bit precision and then dithered down to 16 bit for playback. This is much better than most devices (actually virtually everything with a battery) out there. The actual quantization error introduced by the process will be orders of magnitude lower than what was introduced during the encoder process, making it basically irrelevant. Usually when people have complained about decoder accuracy in the past (which is very rare) it was on decoders that operated at 16 bit internally, and therefore had real output much lower (often 10 bits or less).
Actually, years ago I experimented with optimizations that traded off decoder accuracy for more battery life. Testing with MP3, AAC and ATRAC3 I could go shockingly low before there was an audible difference because of how quantization error compounds in a decoder. Most of it ends up masked. I ended up never finishing the work though except for ATRAC3 !
This is interesting to me. What was the actual accuracy of their decoders such that people noticed a problem?
@Marlene references this in the X5 review at Marlene's blog http://marlene-d.blogspot.com/2015/06/review-fiio-x5-loaned-by-fiio.html. If you look at section five -
"Since releasing
firmware 2.0, FiiO likes to brag about the
32 bit floating point decoding of some lossy formats, namely
MP3 &
OGG. They describe their decoding as industry leading; that is the truth and I´m partly responsible for this as
I started a discussion in November 2013 in the X3-thread on Head-fi. I literally bombarded FiiO with emails regarding their then botched decoding of lossy formats, when that didn't prove successful, I wrote the post I linked above. Shortly after, others started to chime in while I continued providing measurement data revealing how bad the X3 was decoding
MP3,
OGG and
WMA. To make a long story short, FiiO started to work on revamping their decoding engine. It took them half a year to pull it off, the results are stellar and well beyond my expectations. The X5 and X3 now decode MP3, OGG and WMA as pristine as
foobar2000. Which is in fact industry leading; I don´t know any other company producing portable players that decode these lossy formats the way the devices from FiiO do. So what I called
'bragging' at the start of this paragraph is completely justified. I have to mention Head-fi member
JoeBloggs, who works for FiiO and was incredibly helpful in getting it done (he will read this article and he will know what I´m talking about)."
If you follow the embedded link to Marlene's post in the Fiio X3 thread, many other users noticed very poor performance of lossy codecs initially on the X3, with the original 16-bit decoder, with OOG decoding being particularly horrible.
Here's a neat summation of the basic problem from http://marlene-d.blogspot.nl/2013/11/mp3-and-other-hires-formats.html "By all means, the 16 bit, quasi-standard decoding of MP3 & Co. isn´t a good thing. Imagine a CD you ripped yourself to MP3; these files were derived from a normal 16 bit source. 16 bit decoding should be enough then, right? Should, but is not. While the decoded data boasts the sources' original bit depth again, something new and eerie has been added... and I don´t mean the inlying compression errors produced by the
encoder. No, this thing from the crypt is additional quantization noise produced by the
decoder. Cause: truncating floating point values to integer values. These additional artifacts are produced
only because the decoder works at half speed and with half of its options."
I recommend reading the full article and you'll understand why a cheap Win 10 tablet, like the less than $200 Onda Air V919 with foobar2000, can become very attractive over hires DAPs for mixtures of lossy and lossless playback (aside from storage capacity, which is less of an issue with lossy hires that foobar2000 would play perfectly), especially with a DragonFly DAC/Amp or the like.
Given all this, I think I'm asking a legitimate question about lossy performance of this device and particularly whether 32-bit floating point decoders are used for lossy files. I don't believe Marlene will get one of these just to measure it, but it would be sufficient for me to know that the xDuoo X3 decodes in 32-bit floating point resolution for lossy files, like Fiio devices or foobar2000, to determine if this device is something I will buy. Maybe RockBox's lossy decoders are 32-bit floating point and so doesn't have this issue, in which case this device in dual-boot becomes viable for me, though is still unattractive for a mainstream audience who will never RockBox anything their whole life.
I misspoke in the part you found interesting and meant to say that I assume this is a common issue for hires players, including TOTL DAPs, since Fiio appears to be the only portable claiming 32-bit float lossy decoding as recently as 2015, when Marlene wrote that X5 review. I admit that I overstated that as a claim based in anecdote rather than an assumption based on logical deduction. My bad, I was rushing a little through that post in the end and I doubt too many people willing to shell out TOTL dough bother to try lossy playback, let alone measure them for quantized noise from unnecessarily low resolution lossy decoders, so the anecdotal evidence may not exist and almost certainly objective measurements do not.