ASUS P9X79 PRO Motherboard Review

From a performance perspective, we feel that motherboard benchmarking is useless. It’s the motherboard’s job to allow all of the installed hardware to operate at its full potential, so in theory, a $100 option shouldn’t be much (or any) slower than a $300 one. The differences in price instead come down to the quality of the on-board components and other features.

So why do it at all, then? We still perform benchmarks because it’s important to make sure that the board we’re dealing with doesn’t lack in one particular area versus the rest. If board A performs 2% slower than board B in PCMark 7, for example, that’s of no concern to us – random variance is a fact of life with benchmarking. However, if one board consistently performs weaker than the rest, that’s worthy of note – it could suggest that weaker components have been used which do not allow the hardware to operate at its full potential.

Please bare this all in mind when perusing our results. Just because a board under-performs in a single test, it doesn’t mean anything with regards to its quality as a whole. Our ultimate goal here is to make sure that each board we test performs as we’d expect across the gamut of scenarios we pit them against.

With Windows 8 having hit the market this past fall, I had considered moving our test-benches on over. However, because the differences between the OSes are minimal, and also because Windows 8 kind of adds more complexity to our testing with its Start screen, we’re preferring to stick with Windows 7 for the time-being. Depending on how Microsoft’s roadmap looks at the time, we may decide to hop on over to Windows Blue when it arrives (aka: Windows 8.1).

One trick that some vendors have liked to play in the past is to increase their Base Clock ever-so-slightly so as to give their board an edge over the others. With a multiplier of 35, even a single 1MHz BCLK increase can result in seconds being shaved off of most benchmarks. It’s for this reason that we perform this simple check before evaluating the performance.

All checks out – let’s move on.

General System Performance

To take a look at the “overall” performance of our PC configuration, we rely on a simple Windows boot test (cold boot > usable desktop) and dual Futuremark suites; PCMark 7 and 3DMark 11. Please note that we’re using PCMark 7 1.0.4, not the recently-released 1.4.0 (which has resulted in score changes).

ASUS clearly had an advantage in the boot test here, shaving 4.5 seconds off the total time. In 3DMark, both boards resulted much the same, while MSI slid just an inch behind in our PCMark 7 testing.

I/O Performance

To properly give both the internal SATA 6Gbit/s and USB 3.0 controller a good workout, we turn to CrystalDiskMark and Iometer – the latter of which is the first benchmark in our entire test suite that we run.

As pointed out on the previous page, ASUS’ inclusion of support for the UASP USB 3.0 protocol helps to increase overall throughput, and CDM sure proves that well here. MSI’s board does not support UASP, leading to much weaker sequential results. Fortunately for MSI, 4K results come out even.

Rendering & Image Manipulation

Writing files to disk or reading a website doesn’t do much to exercise our CPU, so for that, we turn to a couple of common scenarios; image editing, video rendering and 3D rendering, using a combination of Adobe and Autodesk products.

Our 3ds Max run is the only one that was able to differentiate these boards in any real way, with ASUS cutting close to 3% off of the total rendering time.

Sub-system Performance

For memory and CPU throughput testing, we utilize SiSoftware’s Sandra 2012 (SP4), and for Ethernet testing, we use iperf (or more appropriately, the Java-based jperf which utilizes it). The stock memory configuration uses 4x4GB Kingston HyperX modules clocked at DDR3-1333 and using timings of 9-9-9-24. While 64GB testing isn’t going to be a common theme in our motherboard reviews, we decided to give Kingston’s 8x8GB HyperX Predator kit a test for fun. Those modules are clocked at DDR3-2133 with timings of 11-12-11-32.

It’s become obvious over the past year or two that motherboard vendors believe Intel’s integrated NICs are the way to go for the initial LAN port. Could it be that too many Marvell’s died and caused RMAs to occur? That’d be hard to find out, but it wouldn’t surprise me, based on my experience with those controllers in the past. Given both boards here use identical controllers, it’s no surprise to see them both come out evenly. The same outcome applies to the Sandra tests as well.

With testing out of the way, let’s get to wrapping this review up.