Performance results
SuperPi/BootRacer
SuperPi is a single threaded 32bit application that can really show changes in memory performance. It is a highly tweakable program that runs better on windows XP but for testing no tweaks nor XP testing was done, a standard Windows 7 bench system was used.
BootRacer is a an application that will measure the time it takes for windows to load up to the logon screen. It also gives a total time to a useable desktop. Many things will affect this result including startup applications, overclock and memory speed and timings. Multiple runs are performed and the third fourth and fifth runs are averaged together to produce a result. These normally are identical and represent “best case scenario”.”
SuperPi 1Mresults averaged .02 seconds slower when the RAM was overclocked, but this falls within the margin of error. 32M however improved nearly 5 seconds, and this is exactly what was expected, 32M relies heavily on memory performance, faster memory plays more of an increase than timings, but both help.
BootRacer improvements were far beyond what was expected, with a 9 second decrease in total time to desktop from the higher multiplier, despite actually taking 1 second longer to load up to the login page.
PCMark Vantage
PCMark is a software suite from Futuremark that covers a number of different scenarios that the average person uses a computer for. Individual scores are recorded as well as a total PCMark score for simple comparison to other systems.
With the massive bandwidth available from the quad channel memory controller the average desktop applications showed a larger benefit from the tighter timings of the 1600 Multiplier than the looser timings of the 1866 Multiplier. the only test that scored better with a higher memory multiplier was the storage test which resulted in an almost 5% increase in that specific subtest.
AIDA64
AIDA64 is system benchmark suite from finalwire that focuses on synthetic performance. There are two benchmark suites to run, one covering CPU and FPU performance the other focusing solely on Memory. The memory benchmark is used in this review. Hopefully future versions of AIDA64 will be updated to represent the true bandwidth a quad channel controller is capable of.
Surprisingly, latency took a hit when Overclocked, this is normally not the case, but repetitive runs followed the same results. Read speed also took a small hit, but this was well compensated by the large increase in write results and a more modest 1.5 GBps increase in Copy speed. Overall this test shows that the faster memory multiplier will garner higher bandwidth but oddly higher latency as well.
