I believe it's a cut down chip. Just like Tonga's 384 bit bus and full shaders which were cut. We later learnt that from detailed die shots.
The reasons why I think it's 40 CU, is because of two reasons:
1. Polaris 11 was leaked as a 20 CU 1280 SP chip (manifest and SiSoft database). In GPU design, it is cost effective and time effective when scaling up or down to double up for a chip that's more than twice as big. 20 -> 40 CU. This is especially true for the same generation of architecture.
Then the leak was obviously wrong. Polaris 11 isn't 20 CU's. It's 16, which neither matches up to being half of 36 or 40 and completely invalidates your theory. Furthermore, 28nm GCN didn't follow your guide to being cost effective by doubling or halving chip functionality. Tahiti was 2048 CUs, Pitcairn was 1280 CU's. Hawaii was 2,816 CU's, while Tonga was 2048, Fiji was 4096, and Bonaire was 896 CU's. The only thing doubled there was Tonga to Hawaii. No matter how sound your reasoning might be, your argument is entirely wrong.
I believe it's a cut down chip. Just like Tonga's 384 bit 2. PS4 Neo confirmed 36 CU Polaris 10. Nobody sane would build a mass volume APU with only 36 CU and try to get most chips usable with all 36. It would have to be a redundant harvested part that makes it into the final volume PS4 Neo. This means >36 CU.
It's only 230mm2. The PS4 and XB1 are mass volume APU's that do not have built in redundancy. Why does it suddenly matter now to build it in? I'll give you 2 additional reasons why it's not a cut down chip, on top of countering both of yours.
1. It's ~230mm2. There is absolutely no reason to have 3 SKU's at launch - a full die, a cut die, and a further cut die - for a smaller sized chip when AMD already facing a massive performance deficit. Pitcairn didn't have 2 cut down parts, Barts didn't have 2 cut down parts, GF116 didn't, and neither did GM206. GK106 only intro'd a 3rd SKU (GTX 650 TI BOOST) more than 6 months after it came to market to compete with Bonaire when they left a huge gap between the GTX 650 TI and GTX 660. There is a very real and clear pattern of ~215-230mm2 chips only getting 2 SKU's, with a third emergency SKU coming half a year later to fill large gaps being attacked by the competition.
2. AMD's own slides say Polaris 10 has 36 CU's. It doesn't say RX 480 has 36 CU's.
There really isn't more to say. Your graph you linked has an obvious error between the 390, 390x, and RX 480 which brings it to question it's validity. Everything is shaping up to RX 480 and P10 being a complete disappointment on the performance front. ~65% slower than GP104, an absolutely massive gap when compared to first gen 28nm cards at their release.