Just get an FM3 Turbo from G66, and give it a spin - you've got 30 days to test it out and if it doesn't meet your expectations, just send it back for a refund.
When it comes to hardware revisions, I seriously doubt there is an AxeFX IV on the horizon.
AFAIK, the main reason the AxeFXIII was introduced was because the TigerSHARC DSPs in the AxeFXII were discontinued and so they switched to the TI Keystone which has a completely different architecture.
It also seems to me that there is so much DSP power available in the AxeFXIII that it would be largely pointless to produce a new unit that isn't simply another revision of the AxeFXIII (MK2 Turbo XL Supercharger Ultra).
The Axe-Fx 2 came to be because the Axe-Fx Std/Ultra ran out of firmware space. The Axe-Fx 3 was made because the Axe-Fx 2 DSPs got discontinued like you said.
Looking at TI's DSP range, the Axe-Fx 3
TMS320C6657 DSP chip is $47.85 / 1000 units. This has 2x C66x cores.
The next step up is a model with 4x C66x cores, so double the performance of the Axe-Fx 3, if good parallel processing utilization can be achieved. Except this chip costs $130.5 / 1000 units. So almost triple the price of the current ones.
Ok, but what about using two of the original chips? That's still $95.7 per device and incurs more latency penalties.
So for Fractal to make a faster than Axe-Fx 3 device, they would have to put down a lot of money for the DSP alone before they even have a full product to sell and recoup those costs. At the same time, I would wager 99% of users are not maxing out even the original Axe-Fx 3.
While throwing more processing power at it could potentially mean further accuracy for amp/cab modeling and fx, realistically the needle doesn't move a whole lot when all of those are exceptional already.
But the problem is that Line6 is going to release something with a more modern user interface and more processing power, probably using a similar DSP to what the Quad Cortex uses unless Line6 decides to go with ARM instead.
Let's imagine an alternate world where the Quad Cortex didn't have all its problems. What if the last 3 years the QC life looked like this:
- There were never any power/noise/wifi/security issues.
- Computer editor has already been out for a good while.
- Latency issues have been fixed.
- A lot more amp and fx models have been added and their overall performance improved.
- Capture quality has been improved and compares favorably to Tonex and NAM. I mean it kinda does anyway but let's say even for sticklers about null tests etc.
- MIDI capabilities have been improved so even if you don't like the spacing of the footswitches, just add your favorite controller.
- Content management is a lot better, with offline backups via editor etc.
- There's an online marketplace for buying/selling captures, presets and IRs.
How many people would still flock to Fractal, if we assume they did nothing different in the same timeframe? The company whose Axe-Fx 3 product page still reads "A large 800×480 color display" as one of its highlights. I feel NeuralDSP's incompetence has been a big boon for both Fractal and Line6. If NDSP did better, they could have captured a big chunk of the modeling market. Including yours truly, most likely.
For the next gen, I feel like Fractal doesn't need to do anything but modernize their user interface. But at the same time people are going to be expecting them to overhaul a bunch of stuff or to offer even better modeling capabilities.