EVGA X58 3X SLI Motherboard - Synthetic System, CPU and Memory

Article Index
EVGA X58 3X SLI Motherboard
Close Look at the EVGA X58
X58 Features and Specifications
BIOS and Overclocking
Test Setup, HDD Testing and More
Synthetic System, CPU and Memory
Real World Application Benchmarks
Game Performance and Final Thoughts

Synthetic Performance:

To start things off we'll take a look at PCMark Vantage numbers.  This entire system is identical to the one we built for the Core i7 920 review as previously mentioned and is a whole lot better in some areas that any of the Phenom and Phenom II stuff from AMD.  We are comparing the numbers below with a Phenom 9950 Black Edition processor.  The Core i7 is a couple of generations later and really light-years ahead in terms of technology and architecture so you really can't compare the numbers directly.  The Phenom system uses DDR2 while the Intel system uses DDR3.  These numbers are merely for reference.

 PCMark Vantage

 

To gauge memory performance we used Everest Ultimate and SiSoft Sandra.  These numbers are gleaned from the Crucial Ballistix Tracer 6GB Kit review.  If you want to compare your triple-channel DDR3 performance, you can head on over and grab SiSoft Sandra here and compare away.  Please keep in mind that if you are using a dual-channel kit, your performance numbers will look dismal.


Benchmarking - Everest:

Because there is a lot of system tweaking that goes along with clocking up memory to non-standard speeds, we haven't included any "real-world" benchmarks.  In order to get the memory to run at speeds other than 1066MHz, 1333MHz, 1600MHz or 1866MHz the motherboard bus and CPU must be clocked up in order to achieve these speeds.  As we've shown before the CPU plays a huge roll in benchmark performance and so it's not fair to compare WinRAR compression when the CPU is clocked up 400MHz faster.  Of course the "RAM" will look faster, but in reality the CPU is the one doing the work.  For this reason we use two simply synthetic benchmarks to show memory performance.  How it affects performance in the real-world will depend largely on CPU, motherboard and other hardware bottlenecks so we have to be content with synthetic benchmarks to gain our number metric.  First up is Everest.

 Everest

 While it should be no surprise that CL7 performance at 1626MHz is faster than CL8 performance at 1600MHz, the results aren't quite as impressive as we hoped due to the poorer command rate of 2T.  This is effectively twice as long as the 1T command rate at CL8.  Regardless, when the CPU is overclocked a bit and the memory is running at its fastest stable speed of 1808MHz the performance numbers in Everest are pretty impressive.  Latency measures under 28ns and copy speed is well over 20GB/sec.  Even write speed measures an incredible 15037MB/sec.

 SiSoft Bandwidth
SiSoft Bandwidth
SiSoft Latency
SiSoft Latency

 

SiSoft shows that there is little to no difference in performance at CL7 or CL8 when running right around 1600MHz.  One test shows 0.05GB/sec less performance and the other shows 0.04GB/sec better performance.  I think it's safe to say, it's pretty much a draw.  What does take the lead and the cake is the performance over 1808MHz CL8.  We reach over the magical 30GB/sec performance with this memory according to SiSoft.  In terms of latency, we use SiSoft Sandra to measure linear and random latency and while linear latency remains static at 6ns, we do gain a whole 1ns of latency when moving from DDR3-1600 CL8 to DDR3-1808 CL8.  Tighter timings doing seem to help here as the Command Rate of 2T seems to negate performance when running at CL7.

 

Finally we take a look at PMCore - a program that calculates prime numbers.  This program is multi-threaded and we used it to calculate 10,000 prime numbers.  The results below are in minutes:seconds.tenths.  For reference sake we put up a Phenom II system against the Core i7 just for kicks and giggles.

PMCore

On the next page we'll carry on with some real-world application tests before we go gaming with this motherboard.