Performance Results

The above three images are screen captures from within the ASUS P9X79 Deluxe's BIOS. The first image shows the Samsung module's offered profiles, including the primary as well as secondary timings that the module uses depending on which profile is enabled. This handy screen can help users manually set timings, but it's also worth mentioning that simply selecting and enabling "XMP" in the ASUS BIOS will automatically set most of these timings, but we did find that a few of the timings were different than what is reported in the screens above with some sticks. I've also included the same screenshot from the G.Skill and Mushkin sticks, so we can see the obvious differences and similarities in timings.

My testing involves using only the automatic settings applied by the BIOS, as this is what most users will do as well, so the results that follow should be able to be replicated by anyone with the same components. I have tested the different modules with the primary JEDEC profile, as well as the XMP profile for sticks that offer it, and the results are below. The numbers below reflect performance results with two modules installed for the Samsung kit, and four sticks for the other two kits.

SuperPi

SuperPi has been a standard in memory clocking in enthusiast circles for many, many years. Being highly sensitive to timing and speed adjustments both on the CPU and on the memory, SuperPi is also good for stability testing for those just benchmarking. The Samsung MV-3V4G3 sticks ended up on the bottom of the pile here, but perhaps there's more to them than this one test shows.

wPrime

wPrime is much more focused on CPU performance, but memory plays a role too, although timings are far more critical. Here, the Samsung memory is just ahead of the other memory kits, but performance still remains fairly close even with the vastly different speeds and timings between all the kits.

WinRAR

WinRAR makes use of both CPU and memory again, and the test results highlight how perhaps quad-channel mode can be a hinderance to performance. The Samsung kit, with its JEDEC profile, is half-way between the other JEDEC scores, and the XMP scores, showing that WinRAR seems to prefer dual-channel mode on the X79 platform.

AIDA64 Read Performance

Memory adjustments on the X79 platform affect Read performance more than Copy and Write performance. Here the JEDEC and XMP profiles put the G.Skill kit ahead of the Mushkin kit by roughly 700 MB/s in the AIDA 64 Read Performance test, even though AIDA 64's tests don't fully utilize all four channels on the Intel X79 Express platform. However, the Dual-Channel Samsung kit beat both of the other JEDEC profiles, which really does seem to bolster the idea that these tests may not reflect performance fully. Thankfully, I ran many more tests.

AIDA64 Latency Performance

The gains in latency are pretty linear too, with the Samsung MV-3V4G3 kit winning out between the JEDEC profiles again. This is more than likely due to the lower tRC value the Samsung sticks use.

SiSoft Sandra Bandwidth Performance

SiSoft Sandra's memory test suite provides a slightly different workload than the AIDA 64 tests do, so we've employed it for testing too. With all four channels used, the performance gaps become much larger overall, and the Samsung kit takes the bottom spot. Although it's on bottom, it's using 50% less sticks, but offering more than 50% of quad-channel performance. Interesting result.

SiSoft Sandra Latency Performance

Latency Performance in SiSoft Sandra closely follows the trend seen in AIDA 64 with quad-channel kits, but again we are left with a different result from using just two sticks, Leaving the Samsung kit over 200% faster than all the other kits in this test. Quite amazing.

SiSoft Sandra Cache Performance

The SiSoft Sandra Cache test works not only the memory, but also the CPU cache, highlighting how memory performance affects not just the memory itself, but also how your CPU operates. The Samsung kit, with two less sticks, and two less channels active, put out pretty significant results, almost catching the four Mushkin sticks at 1333 MHz.

HandBrake Encoding

Handbrake encoding testing followed the results given by wPrime, showing that sometimes timings are more important, but not always. Again the Samsung kit wins out of all the JEDEC profiles.

CineBench Encoding

Cinebench encoding, on the other hand, provides a much different encoding workload, where it seems raw speed wins out overall. Samsung MV-3V4G3 kit is again on top, without a doubt.

PCMark 7

PCMark7 provides a bit of a daily usage comparison, testing different parts of the system in different ways. Even in daily performance testing, the Samsung MV-3V4G3 wins out of all the JEDEC profiles, but only marginally so.

Shogun 2 CPU Bench

We fired up the Shogun 2 DirectX 9 CPU Performance Benchmark for some game workload testing. We've updated all of our test results because previously we used an HIS HD5450 for testing memory, but we found that the results given might be a bit VGA-limited. With an XFX HD6950 2 GB card installed, we get a much better picture of what's going on, and although the Samsung kit is not the best JEDEC result, it is fairly close to even the XMP profiles we tested!

This group of tests has enlightened me to what a lot of other reviewers are saying: Quad Channel memory might be hindering the X79 Express paltform's performance in many apps. At the same time, these results might help some users who are wondering about re-using an existing Dual-Channel kit; I really have to say I cannot tell much difference between dual- and quad-channel during daily usage at all, as long as the Dual-Channel kit has adequate capacity.