Test System and Max OC

Test System

Testing Rated Speeds

With DDR4 being an entirely new platform, using new CPUs as well as new motherboards, comparisons to by gone technologies proves an exercise in futility. I don't feel that it's quite accurate to compare things without some sort of common denominator present, and compared to past test platforms, the only common item in my test system right now is the MSI GTX 780 GAMING VGA. Adding in numbers from past X79-based systems wouldn't truly provide the right field for comparison either since the new Haswell-E CPUs offer a substantial performance increase compared to past X79 Express IvyBridge-E and SandyBridge-E CPUs. SO the following numbers are from my current test platform only. In the coming weeks, our CPU reviewer is sure to post a comparison that covers exploring the differences in platforms, which will answer any questions readers may have about that specific subject.


There is also another subject to consider, and it must be mentioned, although it does have to do with CPU-related performance. Getting the most out of any DDR4 memory is only truly possible with some overclocking. The following numbers will show that simply running the XMP-rated 3000 MHz profile offers gains pretty much across the board, but getting the best out of these Ripjaws 4 3000 MHz sticks requires an increase to both CPU and CPU-cache speeds. Combined, the performance the platform offers sky-rockets, and not every memory kit is going to be able to keep up with these increased workloads. Figuring out where exactly the performance bottleneck resides was pretty easy since there are other reviews out that cover the performance offered by running 3000 MHz DDR4 memory. I also saw that nearly none bother to increase cache speed to help the CPU feed the memory properly, so I've included numbers here which show the maximum performance I could attain with a reasonable voltage any user might want to run.

Most software is not yet fully compatible with DDR4 and X99-based systems, with system detection issues prevalent in nearly all popular monitoring tools, including CPU-Z, so the maximum OC numbers shown below are provided with 4500 MHz cache speed on my retail 5930K at 4625 MHz. Both CPU and cache were fed 1.3V. For consistent booting, VCCSA voltages also required a boost to 1.275V. With voltages so high already, getting more out of the G.SKill 3000 MHz kit proved quite difficult as it seems that either board BIOSes have not fully matured or my CPU sample is not the best performer, so I don't have any numbers which show the full memory OC possible. Not that such should matter much since it is obviously going to vary from kit to kit.