AMD Plumbing Linux Support For Reading The CPU's Protected Processor Identification Number (PPIN)

Going back to Ivy Bridge processors, Intel has supported "PPIN" as the Protected Processor Identification Number as a globally unique identification number set in the factory. It turns out recent AMD CPUs are also supporting PPIN and that reading their value is about to be supported on Linux.

The Protected Processor Identification Number (PPIN) is effectively a unique serial number for each processor. One of the intended use-cases for PPIN is in large data centers and multi-socket servers to be able to more easily identify a particular CPU, especially in case of problems. At least in Intel's case, Intel can also translate a customer's PPIN number back into the fab and production run of that particular CPU along with any other internal data in isolating any issues. Intel has supported reading the PPIN under Linux for years and plumbed it into the MCE (Machine Check Exception) code for allowing server administrators to potentially more easily identify a particular CPU in the event of problems as well as tracking CPU inventory.

AMD developers have been working on a patch for PPIN support within the AMD MCE code. The patch basically follows Intel's implementation given their implementations of PPIN are nearly identical. The patch only mentions "newer" AMD CPUs support this feature and not specifying if that means Zen 2 or forthcoming Zen 3 or potentially only implemented by EPYC CPUs (the actual code is simply checking a particular feature bit, waiting to hear any clarification from AMD).

For now the AMD PPIN number is just being reported as part of the MCE logs and not exposed by other means, for those concerned about this globally unique CPU identifier being used for tracking purposes.