I decided to revisit an old rumor, and it looks as though it may have been genuine.
- monolithic die ~22.4mm by ~14.1mm
- 16 Samsung K4ZAF325BM-HC18 in clamshell configuration
- memory vrm seems like overkill with multiple Fairchild/ON Semiconductor FDMF3170 power stages controlled by an MP2888 from MPS
- 3 Samsung K4AAG085WB-MCRC, 2 of those close to the NAND acting as DRAM cache (unusual 2GB DRAM per 1 TB NAND)
- 4 NAND packages soldered to the PCB which are TH58LJT2T24BAEG from Toshiba
- PS5016-E16 from Phison
Die size: 454x256 pixels is 1.77 aspect ratio. 22.4 x 14.1 is 1.59 aspect ratio. The die size of the latter (316) is in the right ballpark, but a miss on the aspect ratio this large is suspect when they're providing sub mm measurements. The image may be thrown off due to uncorrected angle, but that would worsen the aspect ratio, not lengthen it.
Memory: The higher speed modules may have been necessary to run in a true clamshell configuration. Perhaps they weren't 100% decided at that point, so it's plausible.
VRM: We can't tell part numbers from the shots, but this is plausible because there are similar packages on the board. See this
image of a 2080 ti board for comparison of both the power stages and the controller.
Memory: Plausible. Same manufacturer and part line (DDR4). Plenty of DDR4 RAM to match up to 2GB. Though just 512MB would probably be fine.
NAND: They did turn out to be Toshiba NAND from the same series, though the quantity is off. I'll comment here that is very tough to find references to this series online, even scouring the latest device teardowns. This adds a lot of legitimacy, for me.
Controller: Now that we know that the controller is separate from the I/O complex, perhaps it's plausible that it is a custom Phison design rather than Sony (unfortunately we can't see the part number in the video). The E16 was the faster controller available from Phison at the time, maxing out at 5GB/s, which could explain fewer NAND chips.
Summary: If it weren't for the chip aspect ratio, I'd say this indeed looks genuine. It may still be, given the difficulty of nailing down sizes from images, and the fact that chips can undergo revisions (though this one would be significant).