X99 Overclocking

With the move to Haswell-E, the X99 platform changes to a more Haswell-like approach. Due to the use of the FIVR (Fully Integrated Voltage Regulator) on the processor, the motherboard only has to deal with the VCIN to the CPU and then tell the CPU what to do with those voltages. This, theoretically, makes the motherboard manufacturer job much easier in terms of design, but could also make it tricky when it comes to differentiating itself from the competition.

Several points need mentioning here relating to motherboard design. If you skipped over our analysis of ASUS’ overclocking socket earlier in the review it is worth a read, as ASUS is using a different socket design to help extend the range of extreme overclocking, although this might not matter much for 24/7 overclocking.

Next is the power consumption, or more appropriately, the current draw. Haswell-E draws a lot of current through the EPS 12V connector (the 8-pin near the socket), with up to 25 amps with a decent overclock and beyond under more extreme cooling. As a result, a sufficient power supply is needed to do this – typically anything with an 80 PLUS rating should be fine, although it makes sense to check. Also there has been some backlash against motherboard manufacturers placing only one EPS 12V 8-pin connector on some motherboards, with a certain section of users suggesting that motherboards with two of these connectors will fare better with high end overclocks. While this may sound preferable, each EPS 12V connector is rated up to 36 amps, so the only danger comes from spiking when the power supply has insufficient over-current protection.

In our Haswell-E CPU review, we explained that Haswell-E acts much like Haswell when it comes to overclocking – if it takes only one bad core out of four to reduce maximum overclocks in an i7-4770K, then having eight cores in one CPU raises the chances that one core might be the limiting factor. Our two CPU results showed 4.6 GHz on the good processor and 4.3 GHz on the poor one, although ASUS’ overclocking guides suggest that 4.6 GHz at 1.3 volts is actually a good processor, whereas anything below 4.4 GHz is below average.

The i7-5960X is cool at stock and peak load due to the low turbo modes not requiring much extra voltage. Power consumption is relatively slow to raise as well, however above the 4.0 GHz mark, voltages, temperatures and power consumption all start to rise quickly. This means that most CPUs should handle 4.0 GHz with ease, giving a nice 33% overclock. Intel recommends a good air cooler or their TS13X liquid cooler as the bare minimum for the 140W TDP.

For our overclocking tests, we measure the automatic overclock features of each motherboard as well as our standard manual overclocking procedure.

Many thanks to...

We must thank the following companies for kindly providing hardware for our test bed:

Thank you to OCZ for providing us with PSUs and SSDs.
Thank you to G.Skill for providing us with memory.
Thank you to Corsair for providing us with an AX1200i PSU and a Corsair H80i CLC.
Thank you to MSI for providing us with the NVIDIA GTX 770 Lightning GPUs.
Thank you to Rosewill for providing us with PSUs and RK-9100 keyboards.
Thank you to ASRock for providing us with some IO testing kit.
Thank you to Cooler Master for providing us with Nepton 140XL CLCs.

Test Setup

Test Setup
Processor Intel Core i7-5960X ES
8 Cores, 16 Threads, 3.0 GHz (3.5 GHz Turbo)
Motherboards ASUS X99-Deluxe
GIGABYTE X99-UD7 WIFI
ASRock X99 WS
MSI X99S SLI PLUS
Cooling Cooler Master Nepton 140XL
Corsair H80i
Thermalright TRUE Copper
Power Supply OCZ 1250W Gold ZX Series
Corsair AX1200i Platinum PSU
Memory Corsair DDR4-2133 C15 4x8 GB 1.2V
G.Skill Ripjaws 4 DDR4-2133 C15 4x8 GB 1.2V
Memory Settings JEDEC @ 2133
Video Cards MSI GTX 770 Lightning 2GB (1150/1202 Boost)
Video Drivers NVIDIA Drivers 332.21
Hard Drive OCZ Vertex 3 256GB
Optical Drive LG GH22NS50
Case Open Test Bed
Operating System Windows 7 64-bit SP1
USB 2/3 Testing OCZ Vertex 3 240GB with SATA->USB Adaptor
WiFi Testing D-Link DIR-865L 802.11ac Dual Band Router

Methodology:

Our standard overclocking methodology is as follows. We select the automatic overclock options and test for stability with PovRay and OCCT to simulate high-end workloads. These stability tests aim to catch any immediate causes for memory or CPU errors.

For manual overclocks, based on the information gathered from previous testing, starts off at a nominal voltage and CPU multiplier, and the multiplier is increased until the stability tests are failed. The CPU voltage is increased gradually until the stability tests are passed, and the process repeated until the motherboard reduces the multiplier automatically (due to safety protocol) or the CPU temperature reaches a stupidly high level (100ºC+). Our test bed is not in a case, which should push overclocks higher with fresher (cooler) air.

ASUS X99-Deluxe

The X99-Deluxe has several modes of overclocking. The onboard TPU switch has two modes, one for multiplier only and the other for BCLK + Multiplier. The included software uses its 5-Way Optimization method to overclock the system, however it was not functional at the time of testing due to an error in our setup and our early sample.

Overall overclock performance of the ASUS X99-Deluxe comes top of the bunch, supporting 4.6 GHz within our absolute temperature window and 4.3 GHz very comfortably.

At 4.4 GHz we are still below 75C for our peak temperature readings and 4.3 GHz gives under 1.200 volts into the processor. It is hard to gauge if the OC socket is partially responsible for the ASUS scoring well compared to the rest.

GIGABYTE X99-UD7 WiFi

GIGABYTE loves its overclocking. Having something like the GIGABYTE OC Lab, manned by their in-house overclockers HiCookie and Sofos, shows that they are willing to dedicate time, money and resources in that direction. So while the UD7 is not the primary target for extreme overclocking, because it sits at the top of the Ultra Durable line it needs to have something there to help it perform. This is why we get things like voltage read points and buttons on board to help with rebooting.

GIGABYTE’s ecosystem around easy overclocks for most users is via the software, giving options from Light to Extreme with Auto Tuning in the mix. With our 5960X, these all gave reasonable results and voltages up to 4.1 GHz. When it came to manual overclocking, our sample gave a good correlation from voltage applied to voltage read, although our 75ºC peak was lower than expected at 4.2 GHz:

When the temperature gets too high, such as at 4.6 GHz and 1.350 volts, while the system is stable if we monitor the frequency it jumps below the 4.6 GHz mark on occasion. So at this point a non-voltage limited CPU in the UD7 ideally needs plenty of cooling.

ASRock X99 WS

While a Workstation class motherboard might not necessarily be geared towards overclocking in terms of OC-specific buttons or software, the nature of the components should be towards a higher stability, offering some amount of headroom for those versed in pushing frequency limits. The ASRock ecosystem offers a number of automatic overclock options in the BIOS and with the A-Tuning software, and the new layout in the BIOS for manual overclock settings, while easier to follow than before, does require a few more keypresses to go in and out of menus. Results from the ASRock are as follows:

The automatic options are fairly aggressive with our average CPU sample, as given by the +0.060 volts at the 4.5 GHz level affording a power delta of +163W over stock. The 99ºC temperature is quite high as well, indicating that we are perhaps temperature limited overall. Note that if we draw a 75ºC line as a suitable peak temperature, none of the automatic settings fall under this, and it gives 4.2 GHz during manual testing.

MSI X99S SLI Plus

The X99S SLI Plus is not being marketed specifically as an overclocking oriented motherboard, although the press materials mention the OC engine to ease transitions from one CPU strap to another. The X99S SLI Plus is a motherboard designed to get up and go with a flick of the OC Genie button. Given the $230 nature of this board, it seems to be primarily aimed at i7-5820K users, however many of these will want to overclock as well. Our results are as follows:

Several things are quite noteworthy. Firstly, the OC Genie setting is very mild. Every CPU under the sun should be managing 4.1 or 4.2 GHz, and MSI has played it safe going with 3.7 GHz.

Next up is the Load Line Calibration, which we had to leave on auto because the option in the BIOS after several generations still does not make it clear what it should be set as. Here there is a large disconnect between setting the CPU voltage and what the voltage reading is, to the point where at 4.5 GHz, 1.225 volts in the BIOS gives 1.374 volts at load – a +0.150 jump.

Finally, it comes to performance. When we put the overclocking results of the MSI up against any of the others, at each point the MSI result is behind:

This is quite odd. It would seem the efficiency of the MSI motherboard when overclocked is somehow stunted to the point where the ASUS board at 4200 MHz can match the MSI at 4400 MHz. I am unsure if this can be fixed with a BIOS update, or if there is something fundamentally wrong with the MSI when overclocked, because its stock performance is actually rather good.

X99 In The Box System Benchmarks
Comments Locked

62 Comments

View All Comments

  • Ian Cutress - Friday, September 26, 2014 - link

    Usually some of the 16xx series have some leeway, but the larger 26xx are definitely locked down. I've managed 107 MHz BCLK from an E5 2697 v3, but YMMV.
  • halcyon - Thursday, September 25, 2014 - link

    Is this correct:

    http://i.imgur.com/3AgxLfs.png
  • bebimbap - Thursday, September 25, 2014 - link

    That's probably because the 5960x runs at 3.0-3.5ghz stock while the 4790k runs 4.0-4.4ghz stock
    so if it is single threaded MHz limited, then the 4790k can run 14%-47% faster than the 5960x can at stock settings.
  • Ian Cutress - Friday, September 26, 2014 - link

    Minimum frame rate results are always tricky. If the system software initiates something critical in the foreground and causes a single frame to falter, then the whole minimum frame rate is reduced. That's why I'm not always too keen on reporting them, but have them included for completeness.

    The single thread speed is also another aspect, also depending on the cache orientation of the CPU, it might cause a frame or two to load faster/slower than others, again causing that one frame drop.

    Given that this is more common across the Haswell-E line, compared to Haswell, it might be something that fundamental.
  • bebimbap - Thursday, September 25, 2014 - link

    I had always thought MSI was a top tier vendor of MB and GPU's but after my gtx 8800 "malfunctioned," back when they were the best available, and then my z87 mpower MSI MB headers fall apart, and OC's at higher voltages compared to my z87 gigabyte ud5h and is hotter at the same voltages. It made me think about it, and MSI is very similar in marketing style as XFX. They are both usually heavy on rebates, and very cheap for the amount of product you get. But they lack quality. None of my XFX cards perform as well as their Asus/EVGA/Gigabyte counterparts. I now put them in the same tier as ECS and Biostar.
    The MSI board OC'd performance being less seemed more of the same, and I was expecting as much. Until something drastically changes, I'll only use Asus/Gigabyte/Asrock.
  • just4U - Thursday, September 25, 2014 - link

    That's unfortunate.. however you do have to keep in mind that these are sensitive electronic components. I've had boards fail by all the major companies. It happens.. I had 3 dead boards in the Genie Rog Asus series all out of the first batch that came in (7 in total) did it stop me from using Asus? No.. again it happens. Had loose heatsinks dead chipset fans, a capacitor that fell off.. ugh.. Still if I dropped all the companies where that had happened I'd not have any companies to turn to lol.

    Msi is doing a lot of good things these days and their easily right up there with Gigabyte and Asus.
  • CFTheDragon - Thursday, September 25, 2014 - link

    Why is there not the MSI X99 Gaming in the review? Is anyone really going to buy the normal version and not the Gaming one for a X99 build?
  • Ian Cutress - Friday, September 26, 2014 - link

    For review time, these are the samples we were sourced. Not every model is available for review, depending on how the manufacturer wants to focus on different titles. We asked MSI what their most popular/consumer focused board would be in terms of numbers, and they seem really pleased internally with the SLI Plus.
  • The_Assimilator - Friday, September 26, 2014 - link

    The only people who buy "Gaming" boards are the gullible who like bling and think that the KillerNIC is something desirable to have, rather than the liability it actually is.
  • Flunk - Friday, September 26, 2014 - link

    Unless the "Gaming" board happens to be cheaper, which happens a lot because I don't think they sell that great. Damn Killer NICs, just give me Intel and be done with it. Killer started off as nothing but marketing and since they've were bought out by Qualcomm they're just remarked Qualcomm parts with tweaked drivers. MSI's "Gaming" line is a really cynical take on the whole affair, it's just their regular boards with red highlights and (only on some models) a few small IC changes.

Log in

Don't have an account? Sign up now