X99 Motherboard Roundup

About The Roundup         Written by: Michael Blystone         Edited by: Vedran Ikonic

For some time, we at AVADirect have wanted to put the hardware we receive for testing of new configurations to a more interesting use, and what better use than to answer a question we get asked all the time:
 
“What is the best new motherboard to buy for my gaming PC?”
 
The answer to this question is often slippery, changes frequently and comes down to the individual features that are important to any given customer. However, the question gets a lot more interesting to answer when someone already has a top of the line build in mind. This is why we have put together the newest 5960X Extreme edition INTEL processor, a massively overdone custom water cooling setup, and two brand new GeForce GTX 980 cards in SLI, and put them to the test in three of the hottest X99 motherboards around!
 
This roundup is focused on the benefits that these top of the line motherboards and a custom water cooling system, paired with extreme overclocking, can bring to the most taxing gaming and productivity scenarios. We are often asked if these types of halo configurations are worthwhile, if overclocking is still worthwhile, so we will be looking at the delta in performance between a stock clocked system, and one that leverages the advantages of these extremely strong X99 platforms. We will be looking at the features of each board, and the performance differences they show between one another, so when it comes time for our customers to pick between the different brands, the advantages of each board is clear.
 
So, without further delay, let’s take a look at the three motherboards that we will be testing: The X99 Deluxe from ASUS, the X99 UD4 from GIGABYTE, and the X99S XPOWER-AC from MSI. The things that are consistent between these motherboards are the things that most likely any X99 board in this class will share, like the support for DDR4 memory, and support (with an appropriate processor) for up to 40 PCI-Express lanes. They also each have an M.2 slot for super high speed SSD and wireless devices.
 


 
The deeper specifications, like number of USB ports and such, can be found at each manufacturer’s individual website here:
 
ASUS X99-DELUXE              GIGABYTE GA-X99-UD4                MSI X99S XPOWER AC                  

The Hardware

MSI X99S XPOWER AC

 
With regard to visual styling we have another winner here!  This motherboard, billed as an overclocking powerhouse, certainly looks the part. Yellow highlights on a jet black PCB, with subtle hints of gold, this board will look right at home in a highly overclocked workstation or gaming rig. 
 

 
Right away I noticed that the XPowerAC came equipped with an extra 4-pin connection for CPU power, which the ASUS X99 did not, advertised as an overclocking monster, I honestly couldn’t tell if this 4-pin connection added anything worthwhile to that equation. With the 5960X we are thermally limited by the lowered thermal ceiling on the processor, despite an incredibly powerful custom water cooling loop.  Suffice to say, if there is something limiting overclocking on this chip, I believe that it is heat, not lack of power. With that said, the effort is appreciated, and I’m sure in a more ordinary quad or even hex-core use case, the extra power will lead to additional stability in overclocking. This board did appear to have the most consistent power delivery of all the boards tested.
 

 
On top of the well thought out connections, useful overclocking tools, and super steady power delivery, MSI has finally done something I wish a lot more manufacturers would start doing. They have tossed the drivers for the board onto an included 8GB flash drive. Flash drives are incredibly cheap, and I rarely build PCs for myself with an optical drive any longer. The flash drive is a welcome, modern, and sensible addition to the well stacked list of accessories that come with this board. 
 
Speaking of accessories, MSI doesn’t really mess around here, and they provide a slew of extra SATA cables and several SLI bridges, beyond any doubt there is enough extra equipment to make up for something getting misplaced or broken. Nothing is worse than not having that last SATA cable, or finding the one and only SLI bridge in the motherboard box doesn’t work, MSI has tried to mitigate this possibility and I definitely endorse that.
 

 
In terms of baseline performance, this board runs our Intel 5960X by default at 3.5 GHz on two cores. We noticed it dropping to 3.2 GHz under full load, with the CPU at 40c. This was not temperature related, and with a 1500W PSU, most likely not power related either. It may be a misreading of the appropriate turbo ratio by the bios.
 

 
The UEFI Bios for this unit is good, but not quite up to the level of quality of either the ASUS, or GIGABYTE boards, the mouse cursor flickers and judders, the options are a bit obtuse, which comes in stark contrast to the actually useful textual descriptions on the ASUS board, and the generally positive feel of the Gigabyte UEFI bios. This is a slightly disappointing aspect of the MSI unit, and if you are a novice thinking of breaking into overclocking or tweaking, the GIGABYTE and ASUS boards would certainly be a more convenient choice with regard to ease of use.
 

The Test

Our Test Rig

During bench testing of the 5960X, it became pretty clear that this was one of the hottest CPUs that had ever come through our shop. Because of this, we got together and collected an assortment water cooling hardware, installing it in a gutted COOLER MASTER COSMOS II case. Our setup featured a quad radiator out front, a dual radiator on top, and both a reservoir and pump from SWIFTECH. All this cooling hardware leads to an XSPC Raystorm CPU waterblock. This allowed us to move from a Linpack stable 3.9 GHz overclock, to a stable 4.3 GHz. Considering the stock speed of 3.0 GHz on these chips, 1.3 GHz over isn’t bad at all!

One note on the 5960X, is that INTEL has lowered the thermal ceiling by around 11 degrees from prior generations of hardware, this means the chip throttles earlier. If we were willing to tolerate a bit of thermal throttling, we could get up to 4.6 GHz. With a water cooling loop that most people would consider terribly excessive for just a CPU, this is a bit of a surprise. The fact is, however, there has never been an INTEL 8-core consumer part to compare against. 1.3v on 8 cores represents a lot more heat to dissipate than 1.3v on 4 cores.
 

 
At this point we had a processor, overclocked to an extreme level, but what video card to use? We started with an GeForce Titan Black, which performed very admirably on a single 1080p 120hz screen. However a day after our test machine went up, we were given TWO GeForce GTX 980 cards fresh from the fab!
 

 
What to do with them but pair them with this newest gem of a processor from INTEL? This led us to a nearly locked 120hz frame rate in every game we cared to test. That doesn’t make for the most entertaining batch of benchmark results, so we had little choice but to push things even further. We did that by equipping our test rig with three 120hz BENQ monitors in an NVIDIA surround setup, and a custom built racing seat by our own RMA manager!
 

 
With all the pieces in place, we were ready to begin our testing in earnest! We decided to stack up a variety of games and applications for testing, notably Assetto Corsa, the Metro series (2033 and last light) and Crysis Warhead. We wanted to focus on the very demanding programs and investigate the type of benefit they get from a solid CPU overclock when paired with best in class GPUs. We wanted to see if overclocking a top end CPU can still buy notable enhancements in performance, and I think our test results show that is still the case.

While absolute frame rates don’t always tell the whole story, at extreme surround resolutions each board, when overclocked, showed substantial improvements in frametime, and a significant reduction in the microstutter that often plagues multi-gpu rigs.

During testing we took frametime graphs and recorded minimum, maximum and average frame rates. Each application was tested three times at each respective overclock setting to ensure consistency, we use the last run of each test for the final score. I don’t believe in averaging these scores, or otherwise playing with the numbers, as an average isn’t going to represent an actual test that was run. Our numbers come from an actual test result, and I believe that any reliable test score should.

The Test

3DMark Fire Strike Benchmark

Screenshot credit goes to www.futuremark.com


For our next test, we will be taking a look at the most recent edition of 3DMark. We are running the Fire Strike Demo and Fire Strike tests to obtain Min / Max and Average frames per second, along with a 3DMark score. The framerates are taken from the visually spectacular Demo run.
 
 



 
Here we see a similar pattern to the prior set of tests, as the GA-X99-UD4 takes the performance crown and runs with it, beating the X99 Deluxe by nearly 2000 3dmarks. It also handily trumps the MSI X99S in performance.  We also see overclocking here deliver a pretty substantial improvement, giving each of the motherboards on display an almost 400 mark bump.  
 
 

The Test

Metro 2033

Screenshot credit goes to metrovideogame.wikia.com
 
Now we will be taking a look at Metro 2033, an intense test when released, Metro is still every bit as demanding as it was, especially with PhysX enabled. For these tests, we run Metro on the Very High setting, with Adaptive anti-aliasing and Advanced PhysX at a resolution of 5760x1080. Needless to say, the game looks stunning when configured in this way. So how does overclocking impact our scores?  How do the different motherboards stack up? Let’s find out.
 


I think the most clear story in this test, is that enabling Nvidia PhysX, even on the strongest GPU and CPU combo available, is incredibly taxing. On average, overclocking buys us an extra frame per second, and the difference between the generally fastest board in our tests (the UD4) and the slowest (X99 Deluxe) is very minimal. Just as a note, when disabling advanced PhysX, the machines kept up a minimum of roughly 40 FPS, and each averaged above 60 FPS. So if you are a gamer who demands 60 or more frames per second, consider disabling PhysX to get a more consistent frame rate.
 

The Test

Bioshock Infinite

 
Screenshot credit goes to www.amazon.com
 
Bioshock Infinite is our next test, released last year and based on the Unreal 3 engine, it tends to run very well on powerful GPU and CPU setups. Let’s take a look at how the different motherboards stacked up using the built-in benchmark. For this test we maxed out settings and ran with Ultra level depth of field effects, at our triple screen resolution of 5760x1080.
 


There are some interesting numbers and behaviors on display in these tests! We again see the UD4 take the performance crown, with little to no stutter and a sustained average frame rate over 90 fps. Both the X99S in stock form, and the X99 Deluxe had some very strange, but repeatable performance anomalies here, with minimum frame rates dropping to 0 or near 0 midway through the monument island pass of the benchmark on both, repeatedly on the X99 Deluxe. For the X99S this stopped happening when an overclock was applied. On the ASUS board the dip in performance simply lessened, as it improved to a minimum of 3 FPS rather than 0. Still, each board put up great numbers otherwise, with an average frame rate well over 60. Each of the motherboards on display would be very much at home in a top of the line gaming or productivity PC. Just incredible frame rates here.
 
Looking at the overclocking results shows us that when not GPU limited there can be a pretty big performance gain obtained by simply changing around some bios settings and cooling a processor well. Each board gains around 5 average frames a second from the overclock, and in the case of the X99S, overclocking cleared up the stutter that was present entirely.

The Test

PCMark CPU & RAM Benchmarks

We're starting the synthetic tests with the PCMark memory and CPU performance tests. A pretty widely used benchmark, this program places quite a bit of stress on both subsystems. We should get a good idea of how overclocking benefits computing speed in an unfiltered way. Let’s take a look at how these X99 motherboards stack up against one another, in both stock and overclocked form.
 




Here we see all of our motherboards trading blows, with no real clear winner. The scores are too close between the different X99 platforms themselves to say that any one of them is demonstrably better than the others. The real shining star here is overclocking, which adds nearly 25% performance for each board we tested. This represents a tangible improvement that definitely translates to real world computing.

The Conclusion

So at the end of the day which motherboard was the fastest? Did overclocking improve things? Did we see a better gaming experience or remove any limitations on performance?
 
I think we can safely say that overclocking gives us a mild improvement, even in situations that are traditionally GPU limited. In situations where the GPUs are not struggling to keep up, there is a definite improvement. If we are struggling to render at 4k or surround resolutions, a highly overclocked CPU will smooth things out, improve minimum framerates and frame consistency, even if the end results fall within a margin of error.
 
In terms of smoothing out stutters, lowering frame latencies and helping bring more frames back into the magical 16ms range which represents a smooth 60 fps experience, Overclocking helps with this quite a bit. What surprised me, and is something that I previously hadn’t considered very frequently, was that the model and brand of motherboard can deliver a pretty large performance delta, sometimes beyond what overclocking itself can accomplish.
 
Now to answer our very first question, if I were building a gaming or productivity focused machine tomorrow, which motherboard would I buy? Well, the GIGABYTE X99 UD4 is no doubt the fastest.  It isn’t within a margin of error the fastest, and it isn’t arguably faster, the UD4 is demonstrably and by a good margin faster.  Since I love benchmarks, and I’m a sucker for cool designs, I’d go for the UD4 above our other two boards today.  But that isn’t to say any of this hardware is sub-par.  The aesthetics of the ASUS are second to none, and their UEFI bios is again, one of the most impressive I’ve ever seen.  The MSI X99S has a set of accessories that put the other two boards to shame, with the included driver package flash drive just really lending the whole package an air of quality and practicality. Cliché as it is, all the boards we’ve looked at over the last week are winners, and I certainly wouldn’t pass any of them up!

Thank you for the input teca!