Glad to see there's enough space for a M.2 SSD and a generous heatsink. That should increase the number of high-speed SSDs we can use for expansion.
So we're looking at confirmation of the sandwich heatsink patent, right?
I'm not sure the bypass capacitor grid below the SOC is entirely out of the ordinary. A lot of high-performance chips have components like that below. Nvidia just had a brief controversy over similarly positioned components with the RTX 3080/3090.
It's for show, so perhaps things were cleaned, but there didn't seem to be good thermal contact with them.
This is the heatsink sitting in the opposite side of the SoC, item 21 in the patent picture:
That heatsink is too far off to the side, and the heatpipe in the shield looks to be routed over that generous ridge of white thermal compound over the underside VRMs off to the side of the GDDR6 and SOC footprint.
The underside of the SOC has at least two intermediate layers between the capacitors and the shield, and I am reluctant to assume good thermal contact in the absence of a sign of thermal pads or compound. The GDDR6 chips show where pads of some kind allowed for contact with the shield.
The capacitor grid is asymmetric, but that might be because there's a bigger field for the GPU, and then a different rectangular region dedicated to the CPU section.
Now, IIRC the patent also describes item 5d as being stacked memory.
If there is stacked memory on the bottom of the SoC, there could be some meaning to putting the GDDR6 chips in the bottom of the PCB.
There's silkscreened spots on the top side that look to align with the GDDR6 below. Could be space for clamshell mode for dev kits with larger RAM capacity.
They had a shot of the SOC and package, and at least externally it looked like a standard package.
So the patent clearly points at a stacked setup (I only see that this is applicable to a plurality of ICs), but it doesn't mention memory on 5d specifically. It only says that 5d and 5c are two distinct chips.
The patent covered a broad range of scenarios, most of which don't make sense here.
That diagram in particular would also hint at the chips being embedded in plastic or resin, which we see isn't the case.
If that is a "heatsink" and not just a custom looking backplate then what does tiny little plate like that can actually do? It is under the metal casing without airflow.
We don't see the underside too well, so it's not clear if there is a thermal pad. Another possible use is keeping the pressure from the heatsink mount's crosspiece from impacting the capacitors. The plastic crosspiece and metal cap could allow for higher mounting pressure being distributed around the periphery. It's possible the plastic in particular adds enough height to keep the cap from touching not just the PCB but the components under it.
Of course, they wouldn’t be as effective as a copper slug on those same vias, but this is probably a best of both worlds. And you’d already be limited in how much that portion of the board could be Swiss-cheesed with via holes for mechanical/rigidity concerns, especially under the static mechanical pressure of the heatsink assembly.
I think the capacitors aren't sized for uniform contact with a heatsink, although the focus or quality of the shot I saw isn't fine enough to be certain. There are components at the margins that might be taller than the grid in general, and there may not be an interface material.
Wow impressive, much of heatsink over a large area, looks very clean. Such a large heatsink can get rid of alot of heat.
Edit: liquid metal? Had that been used before, pc space? Exotic things we get atleast, like the ssd solution and the tempest audio. Like it
Liquid metal compounds aren't standard in the PC space, but there are liquid metal compounds for after-market application. Sony did make note that they needed years to engineer a solution for the console, likely because of the unique constraints in using such a compound in a mass-produced product.
The complexity in applying and isolating it was one of the downsides which I saw noted for a use case like this.
I'm curious about the comment about getting vapor chamber performance out of heat pipes, since it's not like a vapor chamber isn't successfully used for this class of chip.
What metrics prompted the avoidance of a vapor chamber and the liquid metal application? Is it the specific dimensions of the heatsink footprint, noise level, or hot spot concerns with the GPU clock?
PS5 power supply is rated for 350W. Can we deduce something about the real power draw by comparing to ps4 power supply? Probably no huge difference in power consumption between ps5/xbox series x as series x power supply is rated for 315W.
There's usually a generous margin for Sony's consoles. The PS4 Pro has a 310W supply, for example. Some of the highest numbers I've seen at least unofficially for measured draw were in the range of 170-180W.