What Is Server Memory?
Server memory is the RAM built for servers. It is usually registered ECC memory (RDIMM or LRDIMM) that fixes single-bit errors, supports very large capacities and runs reliably around the clock. Desktop RAM is usually non-ECC UDIMM, which costs less but is not designed for continuous heavy workloads or very large capacities.
Why Servers Use Registered ECC Memory
- ECC: detects and corrects bit errors so one flipped bit does not crash a workload.
- Registered design: a register chip reduces the load on the memory controller, so each channel can carry more modules and more gigabytes.
- Server features: register parity and other reliability features help keep large memory pools stable.
Server DIMM Chipset: What It Means
On a server module, the "chipset" people mention is usually the register clock driver (RCD) on an RDIMM, or the register plus data buffers on an LRDIMM. DDR5 modules also carry a power management chip (PMIC) on the module. The memory controller itself sits inside the CPU, not in the motherboard chipset, so the CPU platform decides which module types are supported.
Server Motherboard Memory Basics
- Channels: each CPU has a set number of memory channels. AMD EPYC Genoa uses 12 channels of DDR5, which gives very high bandwidth per socket.
- Slots per channel: one or two modules per channel is common. Two per channel can lower speed.
- Population rules: follow the motherboard's guide for which slots to fill first, and use matched modules.
- Supported types: most current Xeon and EPYC servers use RDIMM. Many do not accept UDIMM.
Frequently Asked Questions
Can I use desktop RAM in a server?
Only if the server platform supports UDIMM. Many do not.
What is the most common server memory type?
DDR5 ECC RDIMM on current Intel Xeon and AMD EPYC platforms.
How much memory can a server hold?
It depends on the CPU and slot count. RDIMM servers commonly reach 1 to 4TB, and LRDIMM servers can go higher.