Some perspective taking for this thread, because its really easy to let a thread like this start to become a mindf$#% exercise that has folks (like me) irrationally worrying away their weekend rather than rationally thinking through a problem and testing to see to what extent it is a real-world problem for them personally. Not at all meant to apologize or diminish the importance of fixing this issue AT ALL, but this thread kept me holding off on pulling the trigger on a III for several months and it really shouldn't have. If I'd thought it through rationally, taken the numbers that are available in this thread and done some testing with the audio interface I was using (all of which is super easy to do), I would have quickly realized that to the extent there is a problem, its easily fixable, and to the extent there is variation around the solution to that problem...that variation is at most the same as the variations that I've been a-okay with for my entire life as a home recording guitarist.
Executive Summary of a weekend of testing/comparing: (1) I don't think i would be able to identify even 12ms of offset in separately recorded tracks (i.e., two different takes); though I can hear that kind of offset easily for re-recording of a track. Listening to double tracked guitars where neither has a 12ms offset; both have a 12ms offset, and one has no offset and the other has a 12ms offset, I'm pretty certain I'd never be able to pick out a difference amongst any of them. (2) Based on this it's worth setting my DAW to correct by ~10ms, which is the average offset I get, just for peace of mind and (3) There is absolutely no way I or, I would wager, anyone else, is going to be able to hear an uncorrected offset of 1-2ms, so there is no reason to worry about whether or not there are any slight variations of uncorrected offset from use-case to use-case, or from project-to-project, or from preset-to-preset. (4) I tested the Behringer UMC1820 interface I've been using worry free for years and it is offsetting tracks 40 samples EARLY -- i.e., it's printing what I'm inputting to it 0.8ms earlier than when it actually received the signal input relative to its playback of the monitor mix.
Here are some audio examples of recordings that were tracked with various degrees of offset in the range of 10ms. The double tracks include various combinations of tracks recorded with either the same or different offsets in this 10ms range.
The uncorrected offset I measured when sending an audio track of a rim shot (this started life as a MIDI track using an EZDrummer rim shot; just to be consistent, I then converted that to an audio track within Logic) out of my DAW to my III, out of its analog Out 1, back into Instrument input and back to my DAW over USB to re-record it was 460 samples. Sending this track to my III and back to my DAW without ever reverting to the analog world (i.e., same scenario as "reamping" a DI with the Axe Fx III) was somewhere in the low 400 samples range, I believe - when I saw that this variation was roughly the same as the error that the Behringer audio interface I'd used worry free for years has in it, I just stopped worrying about it. I am using a 2012 Mac Mini on Catalina OS and its core audio drivers (i.e., two generations old OS). That is less than 10ms (460/48000 = 9.6 ms) for base offset of recording an analog input. So I've put a -460 sample recording delay in Logic and don't plan on revisiting the issue unless there is an update to the Axe III that requires me to.
While nobody wants to give
@Ugly Bunny another reason to start talking about latency again (cheers, UB!), his video on latency gives some context -- he measured Variax acoustic models to introduce 8ms of latency and found that undetectable when playing; when introducing tuning to electric guitar models on the Variax, the latency went up to 19ms and "you could feel it a LITTLE bit, but its not gonna screw you up or anything". Plenty of folks have said there is a "feel" issue with using a Variax to play downtuned electric guitar; but I don't think I've ever read anyone saying "I just can't use it for recording because that level of latency -- I can hear it in the final recorded product. The timing doesn't feel/sound right of recordings I make." Indeed, I've never even seen anyone ask the question in terms of final musical output. I've never heard anyone complain about the rhythmic tightness of the Variax-using band Twelve Foot Ninjas, or say that Rabea Massad is less rhythmically tight when playing his Variax with tuning effects engaged compared to when he plays a non-digital guitar.
Ugly Bunny measured a Line6 G10S wireless system to introduce 5ms of latency on its own. While plenty of folks complain about using wireless for tonal issues, it's the rare bird that says they can feel a latency problem with modern devices. And again, anyone that does talk latency talks about it within the context of feel, not concerns over what the audience/DAW is hearing. Feeding that into an Axe Fx III gives 7ms total latency. How many metal bands that are supposedly the pinnacle of needing super tight rhythms run guitar into wireless to Axe Fx without complaint about timing?
Have you ever read/heard anyone worry about the rhythmic timing issue of recording guitars with a mic 6 feet away from a cab vs. close mic'ing it (which introduces ~6-7ms of "latency")?
Back to the issue raised by unFILTERed, latency based on signal chain in the Axe Fx is just a fact of life in using digital processors -- modelers, pedals, wireless systems, whatever. Many many MANY folks that run the analog output from their modeler into the analog input of their audio interface and never sweat that the track is being laid down a couple ms later than what they actually played because they are using an Axe Fx, much less worry about whether that latency is changing based on turning blocks off/on in the patch. In other words, the only reason this worry ever even manifested is simply because once there is one inaccuracy, folks (me definitely included) start irrationally worrying about every possible inaccuracy, whether they are real-world relevant or not.
Again, not meant to be an apology, or to imply that this pretty basic level of audio interface functionality is an optional thing. Especially years after the product's release. Just giving some context in case there is another person out there similarly situated to me that has been holding off on buying a Fractal product over this issue -- take a basic measurement, put in the appropriate offset in your DAW, stop thinking about it.