RSS Feed

HCW Tech Blog

For the latest info on computer hardware, tech, news, video games, software tips, and Linux, check out our new improved front page: HCW Tech Blog

Reviewed by: Carl Nelson [02.11.07]
Manufactured by: Kingston



Judging performance of memory is a tricky subject for most hardware reviewers (at least it should be). I mean, memory is memory, and for the most part, one pair of modules should run at the exact same speed as another pair, provided the timings are identical and correct. Therefore, you could just say that this memory performs as well as it should at PC2 6400/4-4-3-15. The other aspect you must consider is overclocking, which we did on the previous page.

So is that the end of the review? No way! We need to look at some aspect of performance! So this is what I decided to do: How do all these timings affect performance? I mean, are you better off with a higher memory speed, but loosened timings? How about tight timings, but with no overclocking? Where does CPU speed fit into all this? That's what we're going to find out.

So I tested this memory with my maximum overclocks with tight and loose timings, and at stock speed. Here's a quick run-down of the settings used:

CPU Speed: 2.40 GHz | Memory Speed: DDR2 800 | Timings: 4-4-4-12
CPU Speed: 2.27 GHz | Memory Speed: DDR2 908 | Timings: 4-4-4-12
CPU Speed: 2.72 GHz | Memory Speed: DDR2 908 | Timings: 4-4-4-12
CPU Speed: 2.45 GHz | Memory Speed: DDR2 980 | Timings: 5-5-5-15

So what we have is a standard benchmark to compare the rest against. In the first configuration, we can find out what an increase in memory speed does, while the CPU and timings are relatively the same. With the second config, we'll see how much a higher CPU speed figures in performance, and in the third, we find out how adversely loose timings will affect performance, even though the memory is runnign as fast as it can, and CPU speed is higher than normal.

Our test rig is:

CPU: AMD Athlon X2 4600+ AM2 (2.40 GHz)
Motherboard: Foxconn C51XEM2AA-8EKRS2H-WTF? (NFORCE 590-SLI)
Video Card: GeForce 6800 GT 256 MB

Software: WindowsXP 32 bit with latest updates, latest drivers, etc, etc.

The first test, of course, is SiSoft Sandra 2007's Memory Bandwidth Test.

With the memory controller on the CPU itself, so it makes sense here that the configuration with the highest CPU speed yields the most memory bandwidth. Comparing the 2.72 GHz DDR2 908 result with the stock 2.4 GHz DDR2 800 result is interesting; even though the CPU has a lot more bandwidth available to it, the performance remains the same; anything gained from the higher clock speed is lost with the loose memory timings. Furthermore, adjusting the memory speed itself while maintaining the same CPU clock speed offers no bandwidth advantage over stock settings.

This time, the combination of a lower HTT bus and CPU clock speed is what affects memory latency results the most. Our fastest setup was once again the DDR2 908 setup with the CPU running at 2.72 GHz. The same DDR2 908 speed with a slower CPU is about 3ns slower than that. And although the memory was set to 5-5-5-15, the relatively high clock speed kept things snappy, allowing it to keep up with our tighter settings.

How does this affect the one thing that matters? Let's fire up Quake 4 to find out!

As expected, results in Quake 4 is pretty much directly related to clock speed. The higher the clock speed, the faster the game will run (at this low resolution, anyway). With the CPU running at 2.45 GHz, it is faster than 2.27 GHz or 2.4 GHz, despite the relaxed timings of the memory.

Next Page: (Conclusion)