Table of Contents
How Memory Errors Happen
RAM stores data as electrical charges in capacitors. Cosmic rays (high-energy particles from space), electrical noise, and thermal effects cause random bit flips — single-bit errors where a 0 becomes a 1 or vice versa. These occur approximately once per 4 GB of RAM per year under normal conditions. In a server with 512 GB RAM, this means errors every few days.
How ECC Corrects Errors
ECC (Error-Correcting Code) RAM adds additional memory chips containing parity information. The memory controller continuously checks this parity data, detecting and correcting single-bit errors automatically. Double-bit errors are detected and reported, triggering a correctable error (CE) or fatal machine check exception.
Impact on Database Integrity
Without ECC, a bit flip in a database's RAM buffer can silently corrupt data — a transaction committing correctly in RAM but writing wrong data to disk. With ECC, this error is corrected before it reaches the disk. For financial systems, medical records, or any data where accuracy is critical, ECC is mandatory.
FAQ
Does ECC affect performance?
ECC has a 1-3% theoretical performance overhead due to parity calculations. In practice, this is imperceptible in real-world workloads and is a trivial trade-off for data integrity guarantees.
Ready to deploy?
ECC RAM on All Power Down EPYC and Dedicated Servers
AMD EPYC and Intel Xeon platforms with ECC DDR4/DDR5 memory standard.
