Yeah agree, ideally we would add some smaller MLCCs caps with the larger Tantalums, together in combination. Maybe even an array like:- 10uF, 1uF, 0.1uF, 0.01uF with every 4 x 470uF Tantalums, for each NEC replacement, maybe would be more ideal. Otherwise, the current arriving at the processors is gonna be too noisey without good filtering (bypass capacitors / decoupling capacitors for both low & different high frequencies). The CELL & RSX seem really senstive to wanting a nice clean power supply / line. I want to experiment with adding some MLCCs, soon all being well, and see what results we get.
I only buy backwards compatible models, so I'm also wondering if the higher nm process is more robust in terms of noise. I've run them missing half the caps with like 400mv peak to peak of noise and they ran GT6 and TLOU just fine. So a tantalum replacement is going to work better with them with no additional modifications to reduce noise. I think the console I was checking all the different tantalum specs on was a H or L, and it didn't like some of them.
The failed TOKINs aren't bad because they're allowing too much noise, though, it's something weird with their reaction to current flow. I never got to invite my EE buddy over that works in power electronics. Maybe next year when covid is over I can set him loose in my lab with the schematics and some beer.
Anyway, yeah, I'm on board with the idea that for whatever reason, later non-BC models might need some additional modification.