On this page
Memory speed shapes how fast an HPE server moves data between RAM and CPU — but on a server the speed you actually get is set by the processor and how you populate the DIMMs, not the module alone. Here's what memory speed means, and how to get it right.
Last updated: 2026
Memory speed is one of the headline specifications when you buy memory for an HPE ProLiant server, and for good reason: it affects how quickly the processor can feed on data, which matters for virtualization, databases, analytics, and any bandwidth-hungry workload. But server memory speed works differently than the simple "faster is better" story — the CPU, the DIMM population, and the memory generation all have a say. This guide explains what memory speed is, how it's changed from DDR4 to DDR5, what actually determines the speed your server runs at, and how to weigh speed against capacity.
What is memory speed?
Memory speed is how fast data moves between the memory modules and the processor. It's expressed as a data rate in mega-transfers per second (MT/s) — for example, DDR5-4800 runs at 4800 MT/s. You'll sometimes see it written as "MHz," but that's imprecise for DDR memory: because DDR ("double data rate") transfers data on both edges of the clock, the MT/s figure is twice the actual clock frequency. A higher data rate means more memory bandwidth, and bandwidth is what keeps a many-core server's processors from waiting on data.
Why memory speed matters
A server processes data by continuously pulling it into the CPU from memory. The faster memory can deliver that data, the less time cores spend idle waiting, and the more work the server completes in a given window. That's why memory speed matters most for:
- Virtualization and VDI — many VMs share the memory subsystem, so bandwidth directly affects VM density.
- Databases and in-memory analytics — large working sets that live in RAM benefit from every bit of bandwidth.
- High-performance and technical computing — memory-bound workloads scale with memory throughput.
For lighter, less memory-intensive workloads, the difference between speed grades is smaller — which is why speed should be matched to the workload rather than simply maximized.
DDR4 to DDR5: how HPE memory speeds have changed
The biggest shift since this topic was first written is the move from DDR4 to DDR5. DDR4 served HPE ProLiant Gen9 and Gen10 servers and topped out at 3200 MT/s; DDR5 arrived with Gen11 and brought a large jump in bandwidth. Here's the progression across the generations:
| Generation | Memory | Typical max speed |
|---|---|---|
| Gen9 | DDR4 | 2133–2400 MT/s |
| Gen10 | DDR4 | 2666–2933 MT/s |
| Gen10 Plus | DDR4 | Up to 3200 MT/s |
| Gen11 | DDR5 | 4800 MT/s (4th Gen Xeon) / 5600 MT/s (5th Gen) |
| Gen12 | DDR5 | Up to 6400 MT/s (Intel Xeon 6) |
DDR4 and DDR5 are not interchangeable — a DDR5 server can't use DDR4 modules or vice versa — so the first question is always which generation your server is. (AMD EPYC Gen11 and Gen12 servers also use DDR5, at their own rated speeds.)
What actually sets your server's memory speed
Here's the server-specific reality the simple version misses: the speed printed on a DIMM is its maximum, not a guarantee. The speed your server actually runs is the lowest of several limits:
- The processor. Each CPU supports a maximum memory speed, and it's usually the gating factor. A 5600-rated DIMM in a server whose CPU tops out at 4800 will run at 4800.
- DIMM population (DPC). Filling both slots in a channel (2 DIMMs per channel) increases electrical load, so the platform steps the speed down — for example, a Gen11 server running 5600 MT/s at 1 DIMM per channel drops to 4800 MT/s when all slots are filled.
- The slowest module. The system runs all installed memory at the speed of the slowest DIMM, so one mismatched module drags the whole set down.
The practical takeaway: buy memory rated at or above your CPU's supported speed, keep every module identical, and remember that a maximum-capacity build (2 DPC) will run a speed grade below a lighter one. This capacity-vs-speed trade-off is real — see how it plays out at the top of the range in our guide to the memory capabilities of the DL380 Gen11. Two related specs also interact with speed: memory rank (higher-rank modules load the channel more) and CAS latency (why DDR5's higher CL numbers don't mean slower memory).
Memory speed vs memory capacity
Speed and capacity are the two axes of a memory build, and the right balance depends on the workload — they aren't mutually exclusive, but at maximum capacity they interact:
- Favor speed for real-time, bandwidth-bound workloads — heavy virtualization, in-memory databases, HPC — where feeding the cores quickly is the priority.
- Favor capacity for workloads that need large working sets in RAM — big databases, large VM footprints, data caching — where fitting the dataset in memory matters more than peak bandwidth.
The catch is that reaching a server's maximum capacity usually means 2 DIMMs per channel, which steps the speed down a grade — so a maximum-capacity build and a maximum-speed build are slightly different configurations. For most servers, HPE's dual-rank RDIMMs at 1 DIMM per channel hit a strong balance; push to full population only when you genuinely need the capacity. For the ceilings by generation, see our guides to maximum memory on Gen11 and Gen12 servers.
Getting the configured speed you paid for
Buying fast memory is only half the job; it has to be installed correctly to run at rated speed. Follow HPE's population rules for your platform — fill the white-labeled slot in each channel first, populate channels evenly, and on dual-socket servers balance memory across both processors — and keep every module identical in capacity, speed, and rank. After installing, you can confirm the negotiated speed in the BIOS/UEFI or through HPE iLO. Our server memory upgrade checklist walks through the whole process step by step.
Where to buy HPE server memory
Match speed, capacity, type, and rank to your exact server:
- HPE Gen12 server memory and HPE Gen11 server memory (DDR5 RDIMM)
- All HPE ProLiant server memory across generations, including DDR4 for Gen9/Gen10
- Find memory validated for your model on the HPE parts by server model page
Not sure which speed your server supports? Contact us with your model or serial number and we'll confirm the right memory — speed, capacity, type, and rank — before you order.
Frequently asked questions
What is memory speed and how is it measured?
Memory speed is how fast data transfers between the memory modules and the CPU, measured in mega-transfers per second (MT/s) — for example, DDR5-4800 runs at 4800 MT/s. It's sometimes written as MHz, but because DDR memory transfers data twice per clock cycle, the MT/s figure is twice the actual clock frequency. Higher speed means more memory bandwidth.
What memory speeds do HPE servers support?
It depends on the generation. DDR4-based Gen9 and Gen10 servers run up to about 2933 MT/s, and Gen10 Plus up to 3200 MT/s. DDR5-based Gen11 servers run up to 4800 MT/s (4th Gen Intel Xeon) or 5600 MT/s (5th Gen), and Gen12 servers run up to 6400 MT/s. DDR4 and DDR5 are not interchangeable, so the server generation determines the memory.
Why doesn't my memory run at its rated speed?
Because the DIMM's rating is a maximum, and the server runs the lowest of several limits: the CPU's supported speed (usually the gating factor), the DIMM population (filling both slots in a channel steps the speed down), and the slowest installed module. Buy memory rated at or above your CPU's speed and keep all modules identical to get the best result.
Does memory speed or capacity matter more?
It depends on the workload. Speed (bandwidth) matters most for real-time, bandwidth-bound workloads like heavy virtualization, in-memory databases, and HPC. Capacity matters most when you need large working sets in RAM. They aren't mutually exclusive, but reaching maximum capacity often means two DIMMs per channel, which lowers the speed a grade — so a max-capacity build and a max-speed build differ slightly.
Does filling all the DIMM slots slow the memory down?
It can. Populating two DIMMs per channel increases electrical load on the channel, so the platform steps the speed down a grade to stay stable — for example, a Gen11 server at 5600 MT/s with one DIMM per channel runs at 4800 MT/s when fully populated. If maximum sustained speed matters more than maximum capacity, a one-DIMM-per-channel configuration is faster.
Can I mix different memory speeds in an HPE server?
You can physically, but the system runs all installed memory at the speed of the slowest module, so a mismatched DIMM drags the whole set down — and mixing is not recommended. For best and predictable performance, use uniform, HPE-qualified modules of the same speed, capacity, and rank across all channels.
How do I check what speed my memory is running at?
After installing, check the configured memory speed in the server's BIOS/UEFI setup or through HPE iLO, which reports each module's details. If the speed is lower than expected, verify your CPU's supported speed, your DIMM population (one versus two per channel), and that all modules share the same rating.
The bottom line
Memory speed drives the bandwidth that virtualization, databases, and analytics depend on — but on an HPE server the speed you actually get is set by the processor, the DIMM population, and the slowest module, not the number on the label. Match memory to your server's generation (DDR4 for Gen9/Gen10, DDR5 for Gen11/Gen12), buy at or above your CPU's supported speed, keep modules uniform, and decide deliberately between maximum speed and maximum capacity. Browse HPE ProLiant server memory or contact our team to match the right module to your server.
