Motherboards, processors and memory: memory
Now we move on to memory. The CPU is a processing engine, and while it has a small amount of memory (cache) for repeating tasks faster, that is very small in regards to how much memory is in computers.
Memory is simple chips, that as long as they have power will keep the data stored in them. That means when the computer is off, memory is no longer storing anything – hence a reboot will wipe main memory and start anew.
Memory is the working area for the CPU – as applications are loaded, the CPU will call to specific areas of memory to load the application and run it, data in memory as it runs, and then clear it when the application is complete (if the application doesn’t do this properly, and pieces of it run when it’s called to stop, it’s called a memory leak).
Memory is addressed sequentially in a system and some ranges are reserved for operating systems, the system BIOS, and even addresses for usb devices. here’s a simple chart:
So, now we know what memory is used for, let’s go over the types.
First, ECC – standing for Error correction code, is usually reserved for server and important business workloads – it has built in error correction to all but eliminate data degradation. Degradation is found near high-power sources, bad connections, and even bad electronics. This technology comes at a cost – it can 2-4x more expensive than standard RAM.
Second, DDR generation – or Double Data Rate – has to do mostly speed, and are in locked step with generations of CPUs and motherboards. Taking a look at current memory specs, only the newest boards are using DDR5 DIMMs (Dual inline memory modules, or memory “stick”), which is an upgrade on DDR4, which replaced DDR3 and so on. Wikipedia has a good chart on the generations of DDR: https://en.wikipedia.org/wiki/DDR_SDRAM. As you see, each generation has gotten faster and faster – allow for Windows, applications and system to be faster and faster.
Third, CAS latency. Each DIMM or SDRAM stick has multiple memory chips on it. The speed between wanting to use a memory column (a way of getting data from the memory) and then that data being ready for read and write, is the latency. Most of the CPUs and motherboards will support a range of memory latencies, and the lower the latency, the more expensive the RAM. Also, as you see memory clock cycle be faster, the CAS latency number go down. No these are measured in nanoseconds, and the difference in a few nanoseconds is not generally noticeable, so most RAM starts at a CAS latency of 16.
Now the DDR4 memory in my system is listed as CL16-18-18-38. My CAS Latency is 16, but a CAS14 chip would technically be faster, even if you can’t notice it. Here’s a good write up of the numbers: https://appuals.com/ram-timings-cas-ras-trcd-trp-tras-explained/.
So, all that you say, most decisions on what to use is about cost vs. gain. I’ll say that my 5 year old CAS16 system does just about everything good, even if it’s been outclassed by newer systems. Just verify the CAS on the memory is compatible with the CPU and motherboard.