There is no direct sound quality benefit to upsampling --- like mentioned, you can't create something from nothing.
Rather, it is an engineering benefit. Signal reconstruction requires a process called convolution, which is low-pass filtering the impulse train. This is not an analog low-pass filter; it is a digital one --- that's what people refer to as the "digital filter". Brickwall filters usually contain a lot of pre-/post-ringing. Shallow filters are bad at suppressing aliasing images. The ones that have the best properties of both worlds (i.e. good rejection of aliasing and low ringing) are complicated. Complicated filters require a lot of processing power, e.g. the one used on the Chord Hugo, which is why readily built-in filters are often not preferred by engineers, given enough time, budget, and processing power. If you upsample, you raise the Nyquist limit beyond the hearing range, which makes filter designs a lot easier because they can be constructed shallowly. This is less work for the engineer, and it makes component selection easier as well.
The Sony actually has least penalty to the act of upsampling, processor-wise, because its D/A conversion is dependent upon altering the speed and length of 1s and 0s in its digital timer (usually with an MCU). The operational bottleneck is basically a simple function of the rate at which the chip itself operates. However, the technical issue with these PWM DACs is noise --- capacitors don't switch instantaneously, they have finite filling/dumping times, so these switches need to not only switch in time with the data, but also relative to each other; naturally, high-frequency capacitor switching creates a lot of noise if not well-managed. Thus, the issues it has to overcome are mostly electronic in nature --- if they can overcome noise issues in designing switching capacitors and a very clean constant power supply, then they are essentially able to create highly efficient DACs with minimal parts. Choosing good parts for those few critical components is a romantic approach that many designers hope for, rather than having to optimize many components together.
However, since we're on the topic of the X7, upsampling itself is an issue relevant to all DACs, even to the X7. FiiO engineers will have the ability to load custom digital filters into the X7 (via the SoC and its ROM) if they so choose, or choose from the built-in Sabre filters, which are fast or slow roll-off symmetrical FIR filters. The choice of filters will have an impact on the sound produced by the X7, as each different filter will have different aliasing suppression properties, group delay, and dispersion. The master clock choice and frequency for the X7 will also help determine the amount of oversampling the X7 employs with the ES9018 (the desktop version ES9018 has some restrictions to what features on the chip can be turned on/off, and thus many people use workarounds to force the chip into certain modes), in addition to the jitter introduced by the clocks.
The point is, there are technical challenges for all digital designs, regardless of the choice. Some of the challenges are overlapping, others are not the same. We cannot just say one is better than the other merely by discussing small marketing choices. Sony has gone all in with its high-resolution audio push and they have indeed put their know-how into this new wave, in the hopes that it'll pay off for them in the long run, whilst continuing to work on this technology. Is the ZX2 kind of a rip-off at its price? Almost certainly. But it should still sound very good, and there will be people who buy it. No need to knock the hustle. The PHA-3 serves as their fallback, using traditional, tried and true techniques of D/A conversion with the ES9018K2M, etc.
At the same time, and this is the main point, because this is the X7 thread, FiiO has a host of challenges ahead of it to make the best possible halo product for its brand. Given their exponential improvement in engineering ability and software capability in a short amount of time, FiiO is deserving of recognition for continuing to work at its products while always pricing things at a point that is always reasonable. Four years ago, FiiO products were at best average and mostly mediocre (no need to sugarcoat it). They've improved drastically in the last couple of years. Now, they have a product in the works that can possibly dethrone or at the very least match competitors at many times its own cost.
