The beginners guide to memory overclocking

Getting to know your kit
Tweaking changes on a kit to kit basis, obviously firstly dependent on the ICs used but then also dependent on the bin level from which the chips were selected, through to the tolerance for a given kit to respond to additional voltages.

So now you know what all those numbers and letters mean, it’s time to truly get to know your memory. To do so a number of applications will come in handy.

This list of software will allow for different levels of testing and diagnostics focused towards system memory. For some individuals, these pieces of software have become second nature while tweaking a system. For others, well, they’ll feel bewildered and lost. For the latter group, we’re going to break down what each does and how it is involved in the tweaking process.

Memtest86+: This DOS-based and launched software is one of the most renowned and well known memory testing suites. With little to no OS overhead due to booting from DOS bootable discs (3.5in floppy or CD), this software stresses the memory, running through passes whereby the memory is addressed, has a piece of data written which is then read and checked against a checksum. This is one of many testing techniques used by the software to test for bad memory blocks be it a hardware failure occurring or instability from tweaking the relationship between frequency, latency and voltage. To be considered ‘Memtest86+ stable’ the system should run through at least 30 passes of the testing suite. It is important to note that the time in which this testing will elapse will increase as the memory modules increase in capacity.

MemSet: An application launched from within the OS, this software allows access to the primary and secondary latencies for tweaking on the fly. The level of access to the latencies can vary depending on the timings available within the motherboard itself and on the level of support for the chipset and motherboard manufacturer by the software. For the majority of enthusiast chipsets and motherboards, full access to all primary and secondary timings are available – however the CAS latency cannot be altered within the OS; even though you’re given the option to do so, the value does not actually set. This software is commonly used by individuals seeking to tweak the memory latencies on motherboards that do not otherwise supply the options within the BIOS or for enthusiast benchmarkers looking to tweak that last tiny amount of system efficiency.

CPU-Z: Purely a diagnostic and system display application, this software is run from within the OS and details a vast amount of easy to read and comprehend information with regards to the CPU and motherboard. The memory tab allows for the information displayed to be read from any single module of memory installed in the system. This allows for the system to have multiple modules in multiple slots and for information such as the frequency, primary latencies, a small amount of secondary timings (often one to three readings) and the command rate at which the memory operates. Additionally the application can read information stored on the module, such as manufacturer, manufacturing date, manufacturing batch and the amount of memory available on the DIMM. To round it out, the memory tab of CPU-Z can also display the SPD frequencies and timings as well as any other automated profile as such EPP and XMP.

Everest: A full diagnostic suite is built into this application. It runs within the OS and lets the user read information from many areas of the system as well as allowing for testing between different sub-systems such as the memory read, write and latency. We use this a lot in Atomic testing. As such, this application serves a dual purpose in that it can read information about the memory, both in regards to system set values or SPD readable values, and can also test areas of the memory sub-system such as read and write bandwidth.

Stress Prime 2004 Orthos: A stress testing program commonly used to test overall system stability and a favourite among enthusiasts. This application is also run in the OS and is generally known for its ability to load a full system and its primary sub-systems. These loadings can also be shifted to load more heavily on a particular area, generally either CPU related or memory related. This application is often the secondary testing mechanism after having fully tested the memory outside of the OS using Memtest86+. Depending on the level at which you wish to test, the general load testing period for this application is to let it to run for 12 to 24 hours; however some individuals believe eight hours to be sufficient.


Prices go south
As the industry makes the inevitable change from DDR2 to DDR3 the memory economy is doing what would be expected – dumping DDR2 stock piles. As much as DDR3 has come down in price considerably since its conception and initial release, it still fails to come even close to the now excessively cheap prices of DDR2.

DDR3 is faster, offers higher bandwidth, has greater future-proofed potential and is a standard that can expand to much larger capacities to feed the ever growing need for additional memory demanded by today’s computing industry – particularly gaming and other enthusiast activities. Yet, none of these positive attributes of DDR3 can out weigh the sheer price difference and price for performance ratio maintained by DDR2 memory in the current scene.

There will eventually be a mainstream switch from DDR2 to DDR3, as there was from DDR to DDR2 – and thankfully looking to occur much quicker – but for the time being, the DDR2 platform still has plenty of strength in it. With memory such as TEAM Xtreem Dark PC2-6400 4-4-4-12 kits based on ProMOS ICs going for as little as $100 for a 2x1GB kit and with quality Micron D9GMH based kits from the likes of Crucial’s Ballistix series going for as little as $160 to $200, the definitive difference in value is too strong to ignore. At this pace it is only the most hardcore of gamers and benchmark enthusiasts who are jumping on the DDR3 bandwagon.