Since the performance of systems equipped with memory working at the same frequency differs insignificantly, we used fewer benchmarks this time. We selected those that are most sensitive to the memory subsystem parameters. First of all, there are synthetic benchmarks, such as the well-known Everest Ultimate Edition utility, for instance.
Another synthetic benchmark that we used was the one that calculates the Pi number according to Gauss-Legendre algorithm.
To estimate the average performance in general-purpose applications we used CPC Benchmark 2007 test that measures the performance during work in GIMP graphics editor application, during video transcoding with Handbrake utility and during multi-threaded load involving HD video playback and simultaneous data compression by 7-zip archiving tool.
We have also paid special attention to archiving performance. This type of tasks works aggressively with the memory subsystem and depends a lot on its speed. In this case we resorted to the benchmark integrated into the 7-zip archiving utility:
And in conclusion we performed a couple of gaming tests in Crysis WARHEAD and Far Cry 2:
The main conclusion that we can make judging by the obtained results is that it doesn’t really matter for the regular users what DDR3-1600 SDRAM kit they have in their system. All kits tested today demonstrated almost identical performance at their default frequency of 1600 MHz. However, if performance difference of a fraction of a percent matters for you a lot, then you should go for the memory modules with lower timings, which is absolutely logical. In fact, this is exactly why the performance leaders in our tests are the kits from Geil, OCZ and Mushkin – the kits that work wit CAS Latency = 7 instead of 8.