How We Review Memory
You may not think it, but writing useful reviews of PC memory kits is not an easy task. Unlike a motherboard or graphics card, there really is not much that can set one apart from the others. Many of these kits use similar specs, and they are all priced about the same. We will focus on overclocking a bit, but not too much as overclocking mileage varies dramatically from module to module. We may have a kit that reaches far beyond its specifications, but the one in the box underneath it may not have gone anywhere at all.
Besides, as you’ll see, memory speed and cas latency doesn’t affect performance all that much on Sandy Bridge. As it turns out, 1600 MHz offers plenty of bandwidth for the CPU, even at higher clock speeds.
Here’s a look at how various timings, along with clock speeds, can affect latency performance:
As you can see, even though we have to increase timings up to CAS 12, latency remains relatively stable as clock speeds increase. And looking at the same clock speed (1600 MHz), moving from CAS 9 to CAS 8 hardly has any effect at all. And this isn’t even in real-world performance – this is a best-case scenario test for latency speed differences to show up.
In SiSoft Sandra’s artificial memory bandwidth test, we see pretty large jumps in pure memory bandwith when increasing memory speeds. But what does that mean for actual computing tasks? To find out, we turn to PCMark 7, which was released today.
We’ll look at PCMark 7 more closely in future reviews, but for now we will be using the “Overall Performance” group of tests. If you have seen the dozens of real-world benchmarks we use in our reviews, and wondered to yourself just what all those numbers mean, then PCMark 7 will make things simpler. Instead of telling you that a game runs at 71.2 frames per second compared to 72.8, or that you can run 15.9 pages per second in a browser test instead of 16.3, PCMark lumps everything together to give one total score. This obviously shouldn’t be the only thing to base decisions on, but I think it is a useful tool to use in a review like this, where the products are all so closely related.
The Overall Score is based on two groups of tests – the PCMark test, which performs the tasks based on things a typical desktop user would do on their computer regularly – gaming, image manipulation, video playback and encoding, and web browsing. The second group is the Lightweight test, which performs tasks users would do on their lower end systems, like netbooks and tablets. It includes tasks such as adding music, importing pictures, text editing, and light web browsing.
All tests use actual components built into Windows 7, so can be considered real-world tests. The single PCMark score given will tell us just how much memory specs affect performance. So here we go:
Increasing memory speed at the expense of latency does improve performance overall – however the difference is very slight. What is more interesting is that you can improve performance a bit by dropping the timings just one step – something that we will look to do with every single kit in this review. This tells us that with the i5 2500K, we are near the limit of how much performance 1600 MHz (or even 1333) can provide.
So what we’ll do for this review is take a look at each kit, see how its specs look, see how IT looks, do some overclocking, and move on. At the end of the review, we’ll see how each kit performs at default settings using the brand-spanking-new PCMark 7 benchmark suite. By the end, we will hopefully have gathered enough information to make a conclusion.
All testing was performed with an Intel DP67BG motherboard. Yes, an Intel brand board. Believe me, along with the ability to use Intel’s excellent XTU overclocking software, these boards are fantastic. We would have reviewed this board, but I believe it was discontinued after the P67 recall.
The CPU used is a Core i5 2500K running at stock speed, with a Radeon HD4890. All latest drivers as of late April 2011 were used.
Before testing each kit, their XMP profiles were loaded using Intel XTU. We’ll post a screenshot of those default settings on each page.
Now let’s take a closer look at each kit, and see how they compare in terms of actual performance.