I think it's almost a certainty it'd have a better CPU too, considering the original XBO had a better CPU then the PS4 it would be quite strange for the Scorpio to not out perform the Pro in that regards.

To the best of my knowledge the PS4 and X1's CPUs are functionally the same thing, MS just took the risk of clocking theirs a bit higher (just as Sony have done with the Pro). While i do think it's likley the Scorpio will present some kind of CPU advantage, it'll have to involve either an updated version of Jaguar, or an entirely new CPU. If they simply further upclock the X1's CPU, it'll be similar to the Pros. There's a limit to how far that boat can be pushed out.

I guess so I'm not technically knowledgable enough to know just how far the particular CPU can be feasibly upclocked. Yeah basically they're the same CPU but the PS4 is 1.6ghz, XBO 1.75ghz, Pro 2.1ghz.

I whould find it strange if the Scorpio was just the same CPU with a 20% increase in Clock Speed. Though honestly I expect it to feature a different CPU rather then just an upclock. It's going to be a costly machine I think.

It would be a bit strange given how they've been presenting the Scorpio, but it wouldn't be an entirely unreasonably decision either. For the Scorpio's hardware advantage to matter, it also needs to be priced competitively against the Pro, and a new CPU will work against it on that front. They're not going to have the same situational price advantages Sony enjoyed with the base PS4 (Kinect at launch, Sony getting lucky with the timings on GDDR5, some more expensive design choices MS made, and in-general manufacturing scale savings), so every decision they make is doubly important.

I think it's mostly going to come down to what they want the Scorpio to achieve. If 4k (or as close to that as possible) is their core focus, then an upclocked X1 CPU wouldn't be unreasonable. If their goals are broader though, such as 1080p-1440p/60ps being a consistently viable performance target, then a stronger CPU is an absolute must. 

I'm afraid so and suspect whether it'll keep up on worldwide sales.

Scorpio CPU depends on the size of the chip. If they keep the massive 363mm2  surface area (or go larger?) of the original XBO chip, that could explain how they can packed it to the gills with leaner GPU cores and could have room for larger CPU cores and perhaps even a larger ESRAM pool (if not scrapped).

CPU is an area where Phil mentioned 'wasn't quite there' for their design last year, so it should be a step up from what is in Pro.

Actually the ESRAM stuff will be interesting. I don't see how you could offer the speed with main memory (and before the someone comes up with the useless maxbw GDDDR5(X) stuff - you're wrong) and I don't see how they solve that with software alone.

It'll be interesting to see, for sure.

Digital Foundry: The stated figure of 320GB/s can be achieved with 8GB of G5X using a 256-bit bus, or alternatively it could be using a 384-bit interface paired with 12GB of GDDR5.

What's useless/wrong about that?

The latter seems what they are going with based on the 12 memory modules on the Scorpio board.

I suppose the ESRAM is not needed in reality since the system won't be used to run games like an OG XBO.

I'm sure the software layers needed are already baked into the Win/Xbox API anyhow, since games have been shipping over a year already across those architectures sharing the development platform.  

The Win10 transition was practically the groundwork needed for bringing Scorpio into Xbox development, unified.

As we know from PS4, these are just theoretical numbers - real life numbers are *way* lower than that.

The onion bus destroys the bandwidth by a lot when comparing the maximum bandwidth with real numbers. This is what happens when you share data between several busses.

The ESRAM on XBO is dedicated for the GPU and it's way nearer to the numbers they said, they even gave out real application numbers.

For numbers, look here: http://www.eurogamer.net/articles/digitalfoundry-xbox-one-memory-better-in-production-hardware

But that is not even the end of the story - access patterns to ESRAM are also different on this memory. You can use it as a cache or as a scratchpad and even as usual memory, also the addressing of the memory is different to what RAM offers (which is clear given the size of the ESRAM but also its very nature).

The third point is that especially this is low latency memory. It sits directly next to the GPU and you have way less cycles to wait for data to arrive. I know this is sometimes confusing for people to see but memory has always to be seen as a pair of bandwidth and latency. Both combined show you the "power" of memory, not just one big number alone.