He's also well connected to official sources. Maybe he reached for comment and got an off the record info about cooling, but was asked not to say. Hard to tell since he's being so vague.
The old sony patent about cooling through the PCB is still interesting. The machining necessary to pull it off would be expensive, and production would be slow.
Which patent was that? I saw a picture from a Sony Japan patent that had metal pillars through the PCB and into the underside of a chip. That seemed related to some kind of component or space constraint in the area on the upper side of the PCB or above the chip itself, rather than there being a normal heatsink on top.
It doesn't seem like the consoles are that space constrained, and I think the power situation is self-limiting with that method.
The first constraint is whatever is crowding the upper side of the chip, which places certain limits on the amount of heat that can be dissipated, as the through-PCB method is likely to be inferior to a solid metal base, much less a heatpipe or vapor chamber.
The other constraint is that thermally significant pillars of metal burrowing up to the bottom of the die are going to block out equally significant areas of the power/ground pads, removing whatever power they could deliver to the SOC.
It would seem like there would be a narrow power range for a device crowded by chips or components on one side, and missing significant pad area needed for the power delivery for devices we're assuming are in the 100+ W range.
It's possible that vapor chambers or an even greater number of heatpipes can be present if the chips are at a similar power envelope but need to deal with hot spots or overall power density at 7nm.
The following is marketing, but it gives some rules of thumb for heat sinks using heat pipes or vapor chambers:
https://celsiainc.com/heat-sink-blog/vapor-chamber-vs-heat-pipe/
A single heatpipe is likely preferred in the <15W/cm range, but multiple pipes can scale. A vapor chamber becomes more likely at higher heat densities, and also does better with keeping temperature uniform across the base of the cooler.
Something like Scorpio at 16nm was likely getting into the region where heat pipes were giving way to a vapor chamber, and the PS4 Pro might have been approaching that range.
A 7nm chip that was more compact could justify a vapor chamber even if its power consumption wasn't out of line with the Pro. If we're going to theorize about multiple components or nearby memory, the package and chips would do better mechanically if they were at similar temperatures, which a vapor chamber is better at.
A potential limitation is that it's possible that a chamber wouldn't have the physical range of a heat pipe, if we're theorizing the PS5 dev kit photos imply ranks of heat sinks mounted at an offset from whatever is generating the heat.
