I did some analyzing of this GPSDO and indeed, both versions work identically. They are both a hardware PLL with one of the XOR gates of the HC86 as phase comparator. The GPS module outputs a 1kHz square wave based on the GPS satellite atomic clocks.
The OVCXO (ovenized voltage controlled crystal oscillator) 10MHz output signal is divided by
1000 10'000 (thanks to @Hans_18T for making me aware of my mistake...
) -- by the ATMEGA328 on the older versions of the GPSDO and by the two HC390 double decade counters on the newer. The HC390 route is straight forward (except that by setting the jumpers R25, 26 or 30, different divisors can be selected, maybe to adapt to possible different GPS module choices), while in case of the ATMEGA's internal division, things are a little more tricky. Since the microcontroller's counter/timers always synchronize the clock input to their own clock frequency, using one of these channels to scale the OVCXO frequency would be a bad idea. Instead, some of the OVCXO's (buffered) output signal is fed via C6 to the PB6/OSC1 terminal of the ATMEGA in order to synchronize the attached ceramic resonator to the OVCXO while still permitting the microcontroller to run if the external oscillator signal isn't present.
One of the ATMEGA's PWM/Waveform generator modules is used to exactly divide the clock frequency by
1000 10'000. I'm not sure what may be the advantage of the external dividers that are getting used in the later GPSDOs, but I'm pretty certain that the differences are marginal, since the frequency generation inside the ATMEGA's PWM module is performed completely in hardware and not desturbed by CPU operation.
Other than dividing the OVCXO frequency, the ATMEGA only performs supervisory jobs like monitoring the GPS module's serial data stream for a GPS lock and digitizing/checking the phase comparator's output signal for droop (to identify a PLL lock) and driving the signalling LEDs accordingly.
The phase comparator signal is filtered through an analog low pass (R59/C77) and then buffered / amplified by U11 before it's fed to the voltage control input of the oscillator. And that's all about this GPSDO. Simple as can be. The big advantage of the new version is that the designer did without the buck converter to generate the 5V supply. It's now all linear, granting much less potential for interference.
Maybe this little write-up helps some fellow member to better understand the working principles of this GPSDO and gives some hints to those who are inclined to buying one of these units.
Edit: Corrected the divisor specification