© Estwald ISI 2015 - 2025
For the Home, Office, Small and Medium Business                                                                                 Fredericksburg, VA

Estwald’s

Information System Infrastructures

Memory Motherboard manufacturers state which type, speeds, and size of memory modules can be used in conjunction with a given motherboard. Memory type will be stated as DDR, DDR2, DDR3, or DDR4 and voltages from 1.35v to 1.8v will be stated and they must be followed. Speed and size are not as rigid. The speed of addressing dynamic random access memory (DRAM) is based on two things the clock speed in MHz and the number of clock cycles it takes for data to be available once an address is provided. The first is denoted by numbers like 1066, 1333, 1600, 1866, 2133, et al; and higher is better. The second is labeled as Complex Adaptive System (CAS) and takes the form 9, 12, or 20; where a lower number is better. To calculate access speeds multiply the reciprocal of the frequency by the CAS. For instance: (1/1,600,000,0000) * 9 = 11.25 ns (1/1,866,000,000) * 10 = 10.72 ns As can be seen, despite the lower CAS of the 1600 it doesn’t make up for the frequency advantage of the 1866 class module and the subsequent reduction in access time. And although speed is important, an increase in speed does not necessarily provide an equal increase in system performance. Memory has two purposes. First, it services storage and other slow devices. Second, it provides the L3 cache, part of the CPU, data it anticipates it may require before the L2 requires it from the L3 cache. So long as memory can upload the data to L3 before L2 asks L3 for it all is well. Tests have shown 90% of the time a 1600 chip will be fine, with an 1866 chip having a hit rate of nearly 99%. The 2133 chip only adds another .004%. It simply isn’t worth the higher cost of the chips except in extreme cases. The amount of memory required by a system depends on the number and type of applications being run both individually and simultaneously. Modern computers set up virtual machines for each running application and the operating system controls their activities. Each application requests a memory space to work in. If it is available the OS gives it to the application. If there is insufficient memory for all running applications then the OS starts offloading applications from memory to a ‘Pagefile’ created on the storage device. Moving applications between memory and the Pagefile is a slow process and it must be performed continuously on a system without sufficient memory. In this instance, to little memory will cripple a system but too much, although wasteful, will do no harm and the OS generally finds ways to use it. Low end workstations tend to use 4GB – 6GB of memory where the more expensive employ 8GB – 64GB. Specialty servers can have 128 GB or more.