Employing a complete system-on-chip as the power distribution network
Employing a complete system-on-chip as the power distribution network
The issue lies with the presence of defective chips within the SoC. This suggests problems in manufacturing or design. Using separate silicon for dedicated components would be more efficient. Modifying the PCH layout to fit its own chipset could save costs and improve performance.
They finally integrated that onto the same PCB as the CPU? If so, it should be considered a SOC feature. Intel has allocated its own PCH, which is more cost-effective to produce than a full CPU. With Intel's sales likely to rise—especially at 14nm, despite not meeting yield expectations—the company may limit using underperforming CPUs for the PCH, saving on their 14nm supply constraints. This approach might become more common in the future when 10nm takes over the workload from 14nm, though Intel hasn't announced any PCIe 4 PCH yet. For now, Intel's PCH will remain dedicated.
I imagine there might be a place where a failing mobile SoC can be reused as a desktop PCH. If certain components like the GPU, CPU cores, or memory controller are completely nonfunctional, this becomes possible. This aligns with what I previously said — the Loongson 7 PCH for Loongson 3 CPU is essentially a Loongson 2 SoC die with dead CPU cores. That’s a system in package.
With 10th generation ULV CPUs using their own chipset on a distinct silicon slice, it seems unlikely this will occur without the PCH taking on additional roles. As seen with Zen 2 and X570, the market leans toward maintaining the PCH as a separate component. Why design broken cores when they can just be omitted entirely?
Additionally consider the PS4 design – it utilized a complete functional Marvell ARM processor integrated into the system-on-chip for its APU.
The PS4 PCH is based on a Marvell Armada SoC. The developer who worked with the Linux version spent considerable time explaining how they accessed it, and that chip posed significant challenges.