Stepping Up to 4 GB: 2 x 2 GB Buffalo PC2-8500 Firestix Reviewed

Our previous article explained the test bed as well as the old and new memory being used. For the sake of comparison though, we won’t do it all in one step.

 First we will be taking out the two 1 GB sticks of OCZ 800MHz memory and installing a single 2 GB stick of the faster Buffalo memory. This in some sense tests 2GB at 800 MHz against 2GB at 1066MHz, but you may have already noticed a problem. The OCZ pair runs in Dual Channel mode, while the single Buffalo stick obviously doesn’t. What we are really testing is how much better it is to run a pair of sticks in a Dual Channel configuration than a single module – even if the latter is rated faster.

Then we add the second Firesti… what is the singular of Firestix? Anyway, if you need a refresher on how to install DDR2 memory, check this article. With all 4 gigs in, we run tests with the memory at 800MHz and autodetected timings, i.e.: the way they will run if you just drop them in there; and similarly to how standard 800MHz modules would run.

Unless both your motherboard and memory are SLI or EPP rated, they will run as standard PC2 6400 sticks (at 800 MHz, or even more slowly if your motherboard is older) until you tell your computer the extra performance is there. You do this in the BIOS. The process isn’t that complicated and should be detailed in your motherboard’s manual.

All it involves is rebooting, hitting a certain key during the manufacturer’s splash screen and finding the right place to enter the desired (manufacturer recommended) speed, timings, and voltage. If you are put off by this, look into SLI or EPP memory, which can tell some motherboards (again check the manual or website to see if yours is one) how it should be configured automatically.

We set the Firestix up the way the package says (1066, 5-5-5-15, 2.1V) and tested. Finally, we overclocked the memory and tested again.

Many memory benchmarks traditionally revolve around bandwidth. The problem is that bandwidth is heavily dependent on how your computer (specifically the CPU and bus) use the memory. Many other tests rely on moving many very small pieces of data around. The problem with these tests is that moving little pieces really isn’t where memory will slow down your computer.

What we need is to move big honking chunks; whole Photoshop or Illustrator windows or a sound file in a game. Stream type benchmarks do a far better job of showing how a computer will actually use the memory.

Using RightMark Memory Analyzer v3.80, we tested bandwidth, largely to ensure the new modules were working correctly, and the results were pretty consistent. We also ran its stream tests and averaged out the results. A few different overall benchmarks were taken (using different products) and averaged as well, to see how the memory options affected the system as a whole. Results are in the next article.

Note that the Vista experience memory score will often drop after an upgrade. Unless you can back up the drop with another benchmark or the system actually feels slower, don't loose sleep over it.