On modern systems, the memory parity bit is used by the motherboard
ECC circuits. I don't think you'll find any systems that do traditional parity checking anymore.
When a memory ECC error is detected, a
machine check exception is generated. A correctable ECC error will only result in a log entry, while a non-correctable error will typically cause a kernel panic.
ECC errors are detected during memory read access, or by the ECC
memory scrubber.