Use intel-mce-tools or amd-mca-tools. Run stress --cpu 64 --io 32 --vm 4 and monitor /dev/mcelog. Also check numactl --hardware to see memory nodes.
# View recent machine checks
mcelog --client
Key fields you may see:
Use vendor decoding tools or platform logs to map status bits to meaning (Intel/AMD provide MCE decoding docs). x64 exception type 0x12 machinecheck exception link
| Tool / Environment | "Link" Meaning |
|-------------------|----------------|
| Linux / mcelog | Interconnect link ID (e.g., "UPI Link 0" or "Coherent Link 2") |
| WinDbg (Windows crash dump) | !errrec shows "Link" pointing to a WHEA_ERROR_RECORD linkage |
| EDAC (Error Detection and Correction) | "Link" refers to memory controller channel |
| Intel PCM / AMD uProf | Link power management or retry count exceeded | Use intel-mce-tools or amd-mca-tools
Thus, when investigating an x64 exception type 0x12 machinecheck exception link, you must extract this field to pinpoint the failing hardware component. Use vendor decoding tools or platform logs to