OK, so I've spent several days now listening to the demo unit and thought I'd pass on my conclusions.
In a nutshell, this is a wonderful DAC at an amazing price, and I'm buying one. It almost equals the best DAC I have here, an Yggdrasil, which is twice the money for a quarter the channels and is frankly a hard act to follow, beating out even a $4000 DAC I had here. In fact, for the first half of my listening, I thought the Okto = would beat the Yggdrasil, and in the end, there were still some things it did better and some cuts as well.
All comparisons were level matched with a balanced A/B switch. I fed the Yggy from my Lynx E22 card, a slightly better source than my AES16e, and of course the demo version of the DAC8 is USB (but I'll be using it with the AES3 outputs of the AES16e, so hope they're as good as the USB one is). I usually triamp, but since my A/B switch has only two channels I went back to the passive crossover. I used my A21 rather than my AHB2 and the source was JRiver, using linked zones (damn, I wish JRiver would fix that feature) to feed the two DAC's.
The comparison was a challenging one, because there are so many filter options on the Okto. I had to familiarize myself with the sound of the filters first. That, though, was an interesting experience. I ended up preferring the linear phase apodizing filter, with the brick wall filter my least favorite -- which is curious, since they measure very similarly (in fact, it could be said that the apodizing filter measures worse since it has some probably inaudible ripple). Otherwise, overall, I found I preferred the linear phase filters, and the fast to the slow. It's hard to be absolutely sure, though, since you can only cycle through the choices, limiting A/B comparisons.
Anyone interested in how the various filters measure can check here:
http://archimago.blogspot.com/2019/02/measurements-look-at-hqplayer-325.html
A friend sent me that link after I'd listened and in retrospect, it seems I preferred the filters that had less high frequency garbage, and of course linear phase.
Once I'd settled on the apodizing filter, I used it for most of my comparisons. (Interestingly, piano was better on the Yggy with all filters but the apodizing filter, but better on the Okto with the apodizing filter.)
Anyway, overall, the main difference between the DAC's was in tonal balance. The Yggy had more midbass, and the Okto had more pronounced highs. I don't think that frequency response measurements account for this, since while the Yggdrasil has as I recall a downslope, it's very slight. (Though as I recall someone wrote a paper some years back claiming that such gradual slopes were responsible for perceived differences in the tonal balance of electronics.)
Unlike most of the other differences I noticed, this one was not subtle. As with every Sabre DAC I've listened to -- e.g., my Exasound e28 (ES9018), and the Exasound e38 Mk II that I had on trial here a few months ago (ES9038), the sound seemed hollowed out in the middle. It is a very light, bright, clean sound that's often exciting. However, at the top, it exhibits a high frequency emphasis, the infamous "Sabre glare." At its worst, violins and sopranos literally screeched (OK, screeched more than they usually do), and when that happened, switching to Yggdrasil was like running cold water over a burn.
That said, the Okto had its own virtues. It was cleaner on quiet passages (as was the Gungnir that I compared to the Yggy some months back), including chorus. It had more solid lateral imaging -- instruments were easier to pinpoint (I didn't notice much if any difference in depth). As I said, it was better on piano -- in general, the Okto had more pronounced attacks than the Yggy, perhaps because of the underdamped filters in ESS DAC's (which may be responsible too for the Sabre glare).
The Yggdrasil, on the other hand, was cleaner on orchestral climaxes. It had more and cleaner midbass, and of course less prominent highs. It also had more detail -- the Okto seemed to homogenize instrumental detail, so you'd hear the valve noise on an instrument through the Yggy but not on the same passage played through the Okto. (Based on experiments with HQPlayer, this may have to do with the closed form filter.) Curiously, it also had more reverberation than the Okto. I believe that I read a recording engineer say that when he tried a Benchmark (don't quote me on that) DAC in his studio, it stripped away the reverb. If so, there's an issue here.
The Okto sounded compressed and hard on massed strings -- not just violins, but violas as well. It had a slight layer of hardness overall.
I don't want to give the impression that these differences are huge. Except for the difference in the highs, which can be so pronounced you'd think someone had messed with tone controls, most were subtle. And I suspect that for many, the choice would be a matter of taste, other components, and the music you prefer.
Something that may not come through is that on some cuts, I prefered the Okto overall, and on others the Yggdrasil. I won't bore you with my nauseatingly exhaustive notes (will I ever even read them myself?), but I think some people would prefer one or the other on the basis of the music they listen to (I didn't listen to any rock but I did listen to some jazz and the Okto sounded
great).
In the end, I preferred the Yggy for its rounder, glowing, more relaxed tone -- both because of subjective frequency balance and the lack of a slight hardness of the Okto. At essence, the music just sounded more real, and I found myself more and more inclined to keep that A/B switch on the Yggy side. So I'll be keeping the Yggy for the mid/tweet panels and using the Okto lower down. But hell, level matched A/B comparisons tend to show up things you'd never notice if you just sat down and listened -- the Okto is a wonderful DAC and if it weren't for the occasionally painful Sabre glare, I'd probably be selling the Yggy and vacationing in France.