Buying a true separates setup means each device is blind to the capabilities of all others in the chain.
That's definitely where I am. I think compromises are inevitable if you combine multiple functions into one unit, like an AVR or active speaker.
I've worked in complex electronics systems all my career, and experience shows you can do it the hard way, or you can do it the easy way, which is with a layered architecture.
If you have one function per layer, you can optimise the design, improve the performance, reduce the price, and take advantage of specialist capabilities, like companies that ONLY make DACs, and do it very well, or companies that ONLY make HDMI switches, and do it very well. or power amps, .... etc.
You need to have established interfaces for that to happen, but a lot of those are already there, like HDMI for video, AES/EBU for digital audio, line level audio, etc.
When you organise the layers like that, you're quite right, the decoder doesn't need to know what the DAC is doing, it only knows that it needs to output digital audio, etc.
There are some elements of the system that are mature and well-established, like power amps for example. They don't need to change frequently.
Other things like video gaming technology or digital room correction do change rapidly, but you don't want to change every layer when only one of them changes.
Playing Devil's Advocate, let's say we don't mind further going deeper into separates and forgoing the "V" from "AVP", surely there exists a device out there that is a great DSP that is also ART-capable, and has more than 3 HDMI inputs?
Part of the answer might be to use a moderately priced AVR like a recent Denon that supports ART, and use it in preamp mode.
Another answer might be to use an AVP with digital audio outputs, and feed them to a PC running Dirac. I believe that wouldn't be limited to 48kHz sample rate like most AVPs are. That can then drive a high quality multi channel DAC.
Another option might be to use an audio interface with it's own Atmos compatible per-channel parametric EQ, like Apogee or Antelope or Grace Design, but these days there are an increasing number of interfaces that support
SoundID Reference room calibration, like Universal, Merging, DAD, Avid, RME, Avid, Lynx, or Audient.
Why must my DSP still rely on ARC/eARC to maximize the full potential of my system?!!!
Of course that's not essential, and Amir's test of the Tide16 may yet show that such measures aren't necessary, but eARC helps to avoid video compromising audio.
All depends on how much the premium is though, at some point I would not pay more just to chase the better numbers.
