Another important perspective of ECC memory is whether your machine can properly respond to the interrupt. ECC without reporting and handling is not a working ECC. I have been using Ryzen machines with AsRock motherboards and the Linux EDAC driver reports errors correctly (verified by overclocking the memory sticks to the edge). On the other hand, I owned a Dell Precision mobile workstation with Xeon W-11955M processor and ECC SODIMM sticks. However, since its EDAC driver never really lands in the mainline kernel, I won't get any ECC reports, so the corresponding process that has a bit flipped will not get killed.