Yes, more transparency would be nice. Because even though I don't think they're lying, I agree that they're probably using an edge case to spin the number positively. Below is my reasoning, but perhaps I don't have a grasp of the fundamentals; feel free to correct me.
My understanding is that typical max bandwidth numbers (like PS4's 176 GB/s) are theoretical not because the silicon inherently can't reach that value, but because of the behavior of memory access code run on it. There's a combination of reads and writes going on at any one time. Sometimes, one in-flight process will need to read data that another process hasn't written yet, so it has to wait. If you have tons and tons of processes going on at once--which you evidently do in graphical work--you might not really lose that bandwidth, though. While the first process waits you can switch to another one to take up the slack. You can't guarantee there'll always be an option, but there usually should be so the lost bandwidth overall is low.
Just Add Water recently said they were getting ~170 GB/s on PS4, which would be ~97% efficiency.
It seems to me that on One the stall issue would be far worse, though. With the eSRAM only 32 MB, fewer processes can be using this memory pool at one time. So when one needs to wait, there aren't as many candidates to switch to. Plus even if you find one, sometimes the needed data will be over in DDR3 instead and you have to wait anyway. Together these things would make stalls more common, and that seems to be supported by Eurogamer's initial report. Back then, when the theoretical max of eSRAM was 192 GB/s,
Microsoft's real-world example only achieved 133 GB/s. That's a mere ~69% efficiency.
So correct me if I'm wrong, but even if the 204 GB/s number is technically correct for the One eSRAM, it seems it isn't comparable like-for-like with the PS4's figure. They're both "theoretical max", but One's is a lot less likely to be reached than the other. If I'm right, comparing them directly would be pretty disingenuous.
