They aren't official die measurements, just speculative ones based on information around the time they were made & assumption of additional units on Oberon that might've been disabled.
That's the graphic I was referencing. You can see basing the measurements off of the following:
These were done by Liable Brave over on Era; just put aside the fuckery of Era as a toxic dump of political nonsense for a moment and take these posts for what they are: very reasonable measurement estimates and predictions. And when you take other things like the ITMedia article into consideration, 48CUs would give you the Radeon VII equivalent they quote (13.8TF) within the known Navi sweetspot of 1.7GHz - 1.8GHz (10.4TF - 11TF)
However, while I personally don't see the 36/40CU Oberon in the leak as the full picture of what Oberon actually is in its retail form, I do suppose it's possible on the low-end. And with the system using a very TDP-friendly CPU that does leave room to clock the GPU above the sweetspot range (assuming the sweetspot range would scale with the efficiency improvements Disco_ brought up i.e 30-35% efficiency in architecture and 10% additional breathing room on the MHz sweetspot clock (upper limit of 1.8GHz would then be 1.98GHz).
So in that scenario yeah, all 40 active at 1.8GHz would probably be better for the chip than 36 active at 2GHz. But like I said earlier, I think it's more probable Oberon is a 48CU chip, not a 36CU one (talking active units), but I'll only stand by that if there is a GPU benchmark that comes around to confirm it, or by some chance actual specs are mentioned for PS5 @ GDC that fall within the 10.4 - 11 TF range (or even 11.4 - 12.15 TF if, again, the supposed improved efficiency of RDNA2 that was suggested to be possible earlier puts the improvements closer to 30-35% instead of 20-25%, assuming 48CUs still).
Conversely though, if those improvements in efficiency being around 30-35% hold true (just for the efficiency of architecture, not a bump in the sweetspot range) ,then that would apply to XSX as well, so it could see TF performance of around 14.19TF (assuming it's a 56CU chip). But that honestly sound kind of ridiculous to me (but I assume it would bode very well for Big Navi going up against Nvidia's upcoming GPU line of cards).
Honestly tho my bets are still pretty firmly hedged on
10.4-11TF PS5 and ~ 12.1TF XSX being the most likely performance ranges (taking all things into account from the GPU benchmarks, insider claims, Github leak, and articles such as the one from ITMedia plus vague claims from people like Phil Spencer himself), minus RT. In other words, 48CU chip PS5, 56CU chip XSX, PS5 GPU clocking @ between 1.7GHz-1.8GHz, XSX GPU clocking in @ closer to 1.7GHz, both using at least some RDNA2 features (PS5 might be closer to RDNA1 & 2 hybrid design, XSX closer to fuller RNDA2 if not complete RDNA2), and about an 8% performance differential. And both easily at the very least matching up to Radeon VII's performance of 13.8TF (arguably moreso when considering the improvements in microcode for Navi and the benefits of RT).