Our testing protocols for these kits of RAM involved installing the RAM using the XMP profile for that specific kit of ram to ensure that the RAM performed at its intended levels. We then ran the RAM in a series of tests including AIDA64, SiSoft Sandra 2012 and SuperPi Mod 1.5. The first two we chose as simple measurements of speed and theoretical performance and SuperPi we chose because it is actual final performance of the kit factoring in both latency as well as clock speed.AIDA64
In AIDA64 we ran the Read, Write and Copy tests for each kit of RAM and compared them against each other.
When you take a look at our AIDA64 results, you'll notice that we had nearly identical write performance across the board regardless of the frequency or timings of the kit. The write speeds were also significantly lower than the read and copy speeds which ranged between 15,000 MB/s and 18,000 MB/s. When it came to read speeds, the performance was almost directly attributable to the frequency of the kits as the faster the kit, the faster the performance. Interestingly enough, the lower latency Patriot 1600MHz kit did worse than the higher latency 1600MHz kit from Kingston. We believe this is because X79 simply does not like nor benefit from timings lower than CAS9. When it came to the copy test, the results were less drastic but still effectively the same placement as the read test with the faster kits coming in higher. In this test once again the Kingston 1600MHz kit beats out the Patriot 1600MHz kit.SiSoft Sandra 2012
In SiSoft Sandra 2012, we took all of the memory kits and compared the measured memory bandwidth figures of each kit based upon the frequency and timings that they were set to.
With the SiSoft Sandra 2012 memory bandwidth benchmark, there is a combination of two different tests which dictate the systems overall memory performance. Looking at these speeds, we see that once again the trends from the AIDA64 tests are repeating themselves with the fastest 2133MHz kit breaking 45GB/s of memory bandwidth. The difference between the float and integer memory bandwidth tests were negligible in all the tests, so there's no cause for any concern there.SuperPi XS Mod 1.5 32M
In SuperPi we simply took the same concept as in the previous tests and compared all of the different kits against each other. Except, in this test we also included the overclocked times against the stock ones so that you can see the performance improvements as a result of the overclocks as well as compare overclocked kits against stock kits and overclocked kits against other overclocked kits of the same frequency. Note that what SuperPi does is calculate multiple loops of n digits of Pi. The 32M test effectively forces the computer to calculate 32 million digits of pi more than 2 dozen times. Note that in this test lower is better.
With SuperPi we were able to get a real feel for memory performance and we could see that on stock speeds that the Patriot Viper kit was the slowest followed by the Kingston 1600MHz genesis kit. Once those two were taken out of the way, the SuperPi performance improved significantly jumping to 11 seconds over 9minutes instead of 17 and 19. In reality, the Kingston 2133MHz kit did beat out the 1866MHz Corsair Vengeance kit, but the truth is that the performance difference there is in the single digit percentages.
The real surprises came when we started overclocking these kits of RAM. We successfully managed to get both of our 1600MHz kits from Kingston and Patriot to 2133MHz and surprisingly enough , the Patriots lower latency helped it edge out the Kingston 1600MHz kit overclocked to the same speed. We also were able to get the Corsair Vengeance RAM to 2133MHz as well, and as you can see the kit was clearly running at lower latencies as it beat out both overclocked kits running at the same 2133MHz. And the performance crown once again goes to the Kingston HyperX Genesis 2133MHz kit as we were able to get it up to a lofty 2400MHz which further pushed our SuperPi time all the way down to 9m and 8.764s.
© 2009 - 2013 Bright Side Of News*, All rights reserved.