In AIDA64, we combine three different tests into one graph. We combine the read speed, write speed, and copy speed for each kit of RAM which is running at its XMP 2.0 certified profile settings. Note: All measurements below are in MB/s.
If you take a look at the graph above, you'll notice that the performance winner in the AIDA64 benchmarks was without a doubt the 2400 MHz Kingston kit. This kit beat out all others by around 10% in all three tests and is quite visible. Interestingly, this kit was the only one that really gave us a much higher write speed than all the others which remained relatively flat at around 14,300 MB/s. The 19,639 read speed also gave us the highest result we've ever seen on this platform to date beating the nearest competitor (2133 MHz) by 1.6 GB/s. These two results were once again mirrored by the copy test as well where the board attained 17,270 MB/s compared to the nearest 16,383, admittedly a much smaller margin than the other two tests but still quite beyond our margin of error.SiSoft Sandra 2012 - Memory Bandwidth
In this test, we run a series of two tests to measure both the integer and floating memory bandwidth of the memory while the CPU is running at 3.7 GHz and the memory is running at XMP certified profile settings. Below you can find the results of our tests. This test is measured in GB/s
In this test, the Kingston HyperX 2400 MHz kit once again out shone the competition by a fair margin. The kit we reviewed came in at just above 48 GB/s in both tests (48.23 GB/s and 48.19 GB/s) compared to the nearest 45.67 GB/s and 45.6 GB/s from the 2133 MHz Kingston kit. This gives us a pretty good representation of what kind of performance improvements one can expect by going from 1600 to 1866 to 2133 and finally up to 2400MHz.SuperPi XS Mod 1.5 - 32M Benchmark
In this test we are running Xtreme Systems' SuperPi Mod 1.5 benchmark which allows you to calculate millions of digits of Pi and is a great measurement of aggregate memory performance taking into account both memory clockspeed and latency. For this test, we have the system calculate 32 Million digits of pi over 20 times until it finishes and once it finishes we get a time result. Note: These times are in minutes, we used to measure in seconds over 9 minutes, but the Kingston 2400 MHz ram changed that metric for us. Note: Lower is better.
In this test we were extremely surprised to see such a big difference between the Kingston 2400 MHz kit and the rest of the kits. So much so, that we actually had to change the metric by which we measured the test. Our old metric was seconds over 9 minutes, but we had to change to simple measurements of minutes (converted appropriately to decimals). In our findings, though, the Kingston HyperX 2400 MHz kit pretty much crushed everything in sight, even the 2400 MHz overclocked settings we managed to get out of the 2133 MHz kit. Admittedly, though, we did have to run at kit at 12-13-12-33 instead of CL11 in order to get it to take 2400 MHz. Clearly in this case, latency does impact this test quite a bit considering the performance difference between them.
© 2009 - 2013 Bright Side Of News*, All rights reserved.