Here's a story about "specifications" (and "change of specifications") that I believe it's worth reading. If you purchased an Intel NUC in the last couple of years you may find it pretty interesting. And, whether you are aware or not, it may also apply to you. Of course, some kind of comment from someone @Intel would be very appropriate.
I got my first NUC in the spring of 2014. It was a DN2820FYKH which I used exclusively to stream HiRes audio to my Onkyo receiver via Foobar 2000 (directly through the HDMI output via WASAPI). Never had a glitch, which, after all, is pretty obvious: all the NUC had to do was to decompress FLAC audio into a PCM bitstream and send it to the amplifier, which took care of everything else. Easy. So, since 2014, my library of HiDef music increased steadily, including plenty of 192kHz/24bit audio.
Given the satisfaction with the above, in the spring 2016 I decided it was about time to consider a similar approach for movies and videos as well. So, given the CPU limitations of the DN2820FYKH, I decided to purchase a NUC6i3SYH which, ON PAPER, had all the capabilities of my earlier NUC, plus enough processing power to cope with most video sources (after all, if DTS-HD, Dolby TrueHD, and whatever else are simply "passed-through" to a capable receiver/amp, you just need a CPU powerful enough to take care of video decoding... right?).
While such assumption concerning video was right (and, after setting the appropriate audio pass-through, video works indeed smoothly), I soon realized that my brand new "Skylake" NUC wasn't capable of playing any of my 192kHz/24bit audio files. What?!?
Useless to say, as such audio files played perfectly with my old DN2820FYKH, my disappointment was huge... but as the NUC6i3SYH had just been released, I optimistically assumed that such an obvious issue would have been fixed pretty soon by some Intel update. And, please note, I perfectly remember that at that time I double checked the original "Technical Product Specification" (which I still have): and that TPS clearly stated that "LPCM, 192 kHz/24 bit, 8 channel" was supported (and both through HDMI and Mini DisplayPort). So I felt sure that, one way or the other, sooner or later, I would have been able to play such files using the new NUC.
Hence, from time to time, I simply checked the Intel site to see if a new "HD Graphic Driver" (which includes the so called "Intel Display Audio") had been released, installed it, crossed my finger, and tried again.
Until last Sunday, after installing yet another driver version released on June 15 2017 (with no benefit whatsoever: just another waste of time), I decided that it was enough, and it was about time to post something on this forum. As I had browsed this forum in the past (and I was familiar with the typical follow-ups of "try changing the HDMI cable" or "maybe your receiver doesn't support that format"), I started gathering documentation and screenshots to show that everything was fine with my receiver and HDMI cables, and that 192kHz/24bit audio could be streamed perfectly not only from my old DN2820FYKH NUC but even through any other kind of PC I have (even an old i7-2600 with a Radeon HD 6450 can perfectly stream 192kHz/24bit to my amp, and both via the Intel® HD Graphics 2000 HDMI or via the Radeon HDMI!).
Then, when I was just about ready to create a new thread on this forum with all such evidence, I thought that it would have been a good idea to start such a post with a "cut-out" from the "Technical Product Specification", showing what Intel had announced/promised in terms of PCM audio support via HDMI and NEVER DELIVERED... This time though, rather then refer to the original PDF that I downloaded over a year ago (when I purchased the NUC), I decided to check the latest available document on the Intel site... And, guess what?
At the very beginning, on page "iv", under "Specification Changes or Clarifications", among a bunch of other things that are mostly meaningless and have no impacts whatsoever, it states:
"August 2016 Spec Change • Page 23 Table 7 from 192 kHz / 24 bit to 192 kHz 16 bit"
They fixed the 192kHz/24bit problem by simply and silently downgrading the specs...!
How cool is that?
Now, does this apply only to the NUC6i3xxx series?
Or, in other words, would I have been in a better position if I had bought a NUC6i5xxx or better?
No: I would have simply spent more money for the same kind of rip-off!
Because such an "August 2016 Spec Change" also applies to the whole NUC6i5xxx series and even the NUC6i7KYK (which is not exactly cheap, right?).
And what about the latest NUCs?
Oh well, you won't find a "change of specifications" history in their TPS (yet), and they all promise "192kHz/24bit" exactly as last year Skylake NUCs did... but I wonder why I should ever trust Intel again on such promises: maybe by August 2017 those specs will also be "downgraded" to "192kHz/16bit", huh? Or something else that worked fine on previous NUCs won't work any longer...
So, here's my 2 cents about this story:
1) Intel knows very well that lots of NUCs are used as "media players" in entertainment systems, or they wouldn't be providing CIR support in every one of them.
2) Yet, in their designs/implementations (and testing!) they seem to frequently overlook and disregard basic standards and assumptions (192kHz/24bit audio via HDMI was already supported by platforms that are now 6 or 7 years old, not to speak of all NUCs before Skylake).
3) I find it quite amazing to think that, while still at the design stage, people at Intel may have screwed up a simple computation such as:
192000 (hz) x 24 (bits) x 8 (channels)
and not realize that their design didn't provide enough bandwidth to PASSTHROUGH such a bitrate (which, by the way, is far from huge by today standards: just 36,864 kbps, huh?).
4) And it's even more disconcerting to think that nobody at Intel ever took care of thoroughly testing their PCM implementation through HDMI before the launch of all their Skylake NUCs (otherwise, they would have easily spotted the problem with 192kHz/24bit PCM audio and done something about it).
5) It's pretty clear that we (the users) are not treated by Intel's NUC division as customers, but rather as "guinea pigs". If you have the patience to do some searching/browsing on this forum, you will find plenty of people (mostly using the NUC6i7KYK) complaining that 192kHz/24bit audio through HDMI (or whatever else) was not available, and no Intel representative ever took the pain to confirm/address the problem. They just suggested to update the bios, drivers, try a better HDMI cable, or even told them that it was probably the fault of their amp/receiver. And of course all such suggestions turned out to be a waste of time (because the problem was Intel's own design/implementation).
6) It took about ONE FULL YEAR for Intel to recognize the problem with their HDMI/PCM implementation on their Skylake NUCs (6i5xxx were launched in Q3/2015, the spec revision occurred in August 2016), and, after one year, how did Intel address the problem? They just simply (and silently) "downgraded" the specs.
7) I was not born yesterday, and I am well aware of all the fine prints and "specifications subject to change" stuff, but that's meant to give manufacturers freedom to update specs and features for future revisions of their product, not to downgrade the specs of something that has already been sold on false promises. If you sell me a car saying it has 4 wheels, and deliver a product that has only 3, changing the specs one year later (and downgrading the description to "3 wheels") doesn't do the trick: I'm entitled to a refund, even more so if, for nearly a year, you mislead me suggesting a should get a new pair of glasses (or in our case get a different HDMI cable or even a new amp/receiver) to see and enjoy the "fourth wheel"... And/or implicitly lead me to believe, given the obvious nature of the issue, that some kind of fix was surely going to come (and wasting my time in multiple useless updates of BIOS, drivers, and whatever).
8) All in all, it seems that the whole "customer care" concept appears to be unknown to the Intel NUC division. Because once you recognize your error (as they eventually did), at least you should do whatever possible to alleviate the problem for your customers. Or not?
In this case, once you realize you have not enough bandwidth to passthrough 8 PCM channels @ 192kHz/24bit via HDMI (as they originally stated in their TPS), at least provide support for 2 channels (stereo!) or 5.1 (six channels). But they didn't, and simply removed 24bit audio entirely from their specs.
9) Please note that the kind of fix-up I just suggested above (supporting at least stereo and maybe 5.1) doesn't imply write tons of code. They already support PCM passthrough streaming (via HDMI and whatever) at different frequencies and bit depths, so all that's needed is to allow the PASSTHROUGH of any 192kHz/24bit bitstream that, because of the limited number of channels, still fits the available bandwidth, so that applications can at least stream 192kHz/24bit stereo. But it seems that for Intel NUC division their customers don't deserve not even such a partial solution to the problem. Too much work, huh? Let's simply remove any hint to 24bit audio all together from the specs: who cares... And not even an apology.
10) As I already mentioned, though Intel knows very well that lots of NUCs are used in entertainment systems, they clearly don't seem to have a clue of what people assume and expect to be "standard feature" nowadays. 16bit audio is 30 years old stuff (compact disc). And all subsequent advancements in digital audio implied higher bit depths (DVDs, DVD-Audio, BluRay, etc. all assume bit depths > 16) for the very simple reason that higher bit depths provide better dynamic range and accuracy. So seeing a "change of specs" in which Intel in August 2016 tells us that its HDMI implementation supports 16bit PCM is simply laughable. And that's exactly what they have done. Period.
11) Here's how the PCM support/implementation for HDMI and DigiPort is now described after the change of spec:
"LPCM, 192 kHz/16 bit, 8 channel"
Honestly, who cares about 16bit PCM nowadays?
Dear Intel/NUC guys, 24bit audio has been around for years, and, guess what, we expect your specs to tell us something about that: what do you support in terms of 24bit audio through HDMI, DigiPort or whatever? Up to what frequency, and how many channels?
Instead, after the screw-up, it seems that Intel/NUC guys prefer to stay clear of any further reference or commitment on the 24bit audio front, and thus they just don't write/specify anything at all...
We just have to find out by ourselves, huh? Wow!
12) So, guess what: before writing all this I spent some time experimenting with different audio files and programs and came to the conclusion that, at least on my NUC6i3SYH (and latest driver), 24bit audio is supported up to 176kHz (multichannel). So, if such conclusion is correct, there's plenty of bandwidth to support at least 192kHz/24bit stereo and even 5.1 passthrough. If only Intel cared, of course. But they don't: the fact that in their revised specs there's not even a single hint to 24bit audio through HDMI tells it all.
Now, I'm sure someone (maybe even from Intel) will tell me that the difference between 192kHz/24bit and 96kHz/24bit is not that discernible, so I should just feel lucky and be happy that the latter works and just stick with it. Yes, indeed. Even the difference between FULL HD and 4k and is not that discernible unless you are sitting pretty near to the screen... so what?
I have an amp that accepts 192kHz/24bit and that worked perfectly with my old NUC. So I purchesed a lot of audio files in that format and I don't see why I should have any trouble playing them with a platform that "promised" to support that (as just about every single PCs does!). That's all.
Am I wrong?
Indeed, I can play those files at a lower frequencies on the NUC6i3SYH. For instance, while Foobar simply gives an error on any such files (saying that 192kHz/24bit is not supported), if I try with Kodi such files gets played anyway... But guess what: if I check what my amp receives it's a 176kHz/24bit bitstream. In other words, Kodi resamples the audio at the highest supported frequency for 24bit. This resampling, of course, doesn't certainly do any good, and performing such resampling each and every time I play the same file it's just a waste of CPU. Right?
Sure, I could spend days (maybe even weeks!) performing such resampling offline, once and for all, on all my library files. Yet I would need additional storage, because I don't want to throw away the original Hi-Res files, so each and every audio file in my library would end up being duplicated in different formats (and we are speaking of Hi-Res audio: files much larger than CD). And, of course, I would also need to reorganize my entire audio library accordingly. Given I was sold a NUC that promised 192kHz/24bit support out of the box (exactly as earlier products did), and after more than a year of useless driver and bios upgrades, I'm not that happy having to waste more time like this (not speaking of adding more storage, reorganizing the library, etc.).
Of course, I could put the old NUC DN2820FYKH back into use (though I now use it for different purposes in a different room). In other words, as ridiculous as it sounds, I could use the old DN2820FYKH just for audio, and the NUC6i3SYH just for video. But, guess what, all 7 HDMI inputs on my amp are already in use. And anyway it would probably be a mess having those two NUCs sitting side by side and operated by the same remote control. Right?
Lastly, I could buy a new NUC... But why in the world, after this experience, should I trust anything written on an Intel NUC TPS anymore... huh?
You tell me.
A very disappointed "Guinea Pig"