Thanks for the additional detail. The difference in CPU power does boil down to a cost argument, I think (in contrast to an architectural advantage of the Zynq). The other scope manufacturers could have used faster dual-core ARMs, but decided against it for budget reasons. Not sure whether there is an actual price difference, by the way: When comparing the GW Instek with the Rigol DS1000Z series, one should consider that these scopes are in different price ranges, so GW Instek probably did spend more on the Zynq than what Rigol spent on their CPU plus FPGA.
The other aspect you mention (shared bus between CPU and prorammable logic) is what I also referrred to, as the area where I would expect a qualitative performance difference due to the integrated architecture. Hence I was surprised that the Xilinx sheet does not go into more detail here or give numbers. Maybe this advantage is difficult to quantify, since it depends on the application. I would expect that it helps in a DSO, enabling fast handover of the acquired data to the CPU, and back to the screen. Is it conceptually comparable to classical DMA interfaces (direct memory access), or is there a fundamental difference?
That Xilinx sheet seems more a "brochure" aimed to high level managers than HW/SW developers, it's purposely not intended to go deep in tech details, anyway on their website you can find tons of docs.
Same thing for Altera's CPU/FPGA SOC like Altera's Cyclon 5.
We do not know (at least i do not) how Zynq or "generic" CPU/FPGAs really cost for medium/high quantity (>>1000), Digikey or similar listing may be quite off comparing its prices against what a brand like Rigol, GW, Siglent could obtain under a trade agreement with Xilinx or Altera, like GW seems to have done.
The GDS-2000E series shows off 120.000 wfms/s waveform update rate against the 50.000 wfms/s from the GDS-1000B series (the one analyzed by Dave), anyway these two scope lines are built in a complete different way (the 2000E in a much better way) so i'm not sure if the Zynq 7000 model inside the "E" series is the same for "B" series (7010CLG400) or if they improved a lot the system efficiency.
If we take Digikey prices in account for a rough cost comparison, the Zynq 7010CLG400 comes for about 85USD (there are several variants, anyway) against the average 50-70USD amount for traditional cheap DSO's CPU+FPGA coupling, where you have yet to deal with additional RAM memory chips, PCB design effort and bigger board size.
Not a huge cost difference, maybe a DSO could be even cheaper if built around Zynq, the reason why right now you can find only GW ones on the market (in this segment, of course) should lie in the fact that it's a relatively recent SOC family, supported by a complete different code toolchain with a steep learning curve, not a joke to deal with.
I could imagine that GW placed it's bet on this SOC benefiting from a "special price", and last but not the least a rilevant development support from Xilinx's specialists and/or external consultants.
Anyway, if you look to FFT performance and overall responsivity of GDS2000E, as i did this night, you soon discover an amount of computing power not common for similar or higher priced DSOs.
I can bet that next Rigol DSOs generation will go for the same path.