Relative performance (percentages) is the best for comparisons, but the actual difference should also be considered. Along with compute units, there were other aspects of the PS4's GPU that gave it an advantage over the XB1's, and I think similar differences will be seen between Neo and Scorpio, favoring Scorpio. It can be assumed that Neo and Scorpio will be based on the same architecture, but not the same GPU configuration. I think gpu core clocks greater than 1000mhz are rather unlikely, in which case it can be assumed that Scorpios GPU will have more compute units in order to hit 6TF.
It's clear it will be awhile before we really know how Neo and Scorpio compare with each other in real world gaming scenarios. I feel that the end result differences will eventually manifest into a larger delta between the two than exist between PS4 and XB1. PS4 had a better GPU, and arguably a better memory system setup, but a slightly lower clocked CPU and the same amount of RAM. Scorpio has (based on Neo rumors) a significant GPU advantage and a significant memory bandwidth advantage, while probably (based on info from the motherboard render at E3) having 50% or more RAM, and potentially a significantly more powerful CPU (if it's Zen). There's a real chance that on a component level Scorpio will have a significant advantage in every area - something we really haven't seen in two competing consoles since PS2/Xbox (which where on different architectures, making some comparisons more difficult).
Well, lots of speculation on your side, all best case scenarios as well for Scorpio. The truth is, NEO hardware is not finalized, but at least we have some specs. Knowing whether Scorpio will have more CU's or a Zen CPU is not something we can use to estimate a divide between the two systems as yet, farless saying that it will be huge/significant in every area.
All we're doing right now is taking MS's word for it and comparing 4.2TF against 6TF, we have no idea how MS arrived at 6TF, but we do know how the 4.2TF was arrived at.. AMD has already said there's no 40CU polaris and that RX480 is the top end polaris product, maybe they're bluffing and waiting to unveil RX490 with more CU's, who knows, but I'm not sure why the consoles won't be able to clock polaris 10 beyond 1000Mhz. At reference rx480 is 1120Mhz and 1266 on boost, MS has been known to have huge box consoles, so I don't see how their console won't get to reference clocks. Neo according to current spec is already over 900MHz and normally Sony pushes smaller FF designs, so I don't see how 1000MHz is impossible for either console for that matter.
I think the bigger discussion will be how these consoles evolve during the design process till release, there's quite a few hardware innovations and products both for mobile and desktop platforms that will hit in the next couple of months and early to mid next year, so final products for either NEO or Scorpio may not be as picture perfect to what we esteem them to be at the minute.
One thing we must understand is that NEO and Scorpio are not consoles developed with billions of dollars in R&D+, they're iterative consoles where there's more flexibility in what makes it inside the consoles relative to the latest technologies available at the time or within a 3-6 month window of availability.
How is this thread still getting so much discussion over what is a very generic spec list.
Those who argued that 900p is acceptable will now argue that nothing less than 1080p is acceptable.
I imagine that will happen most certainly. I can already see Digital foundry writing articles on how resolution matters and that higher resolutions coupled with 5-10 fps better is a world of difference as it was in the PS3/360 era. For either scenario, NEO vs Scorpio will make for some very interesting reading and takes ......