Record DDR3 Memory Overclock on ASRock Micro-ATX Mobo

By published

ASRock has managed to reach an effective DDR3 memory speed of 4285.6 MHz on a Micro-ATX motherboard.

A team of overclockers consisting of Nick Shih, John Lam, and Splave has managed to set a world record for overclocking DDR3 memory. Nick Shih is ASRock's own overclocker. The motherboard that the team has used is the ASRock Z87M OC Formula and, clearly, it is living up to its name. This motherboard is a Micro-ATX motherboard that is targeted at overclockers and gamers.

The overclock that the team managed to achieve is, behold, a staggering 2142.8 MHz (effectively 4285.6 MHz). This feat was accomplished at the overclocking show at Computex. The memory module that was used was a Team Group Xtreem LV-2666 module, 4 GB in size. Timings were set at 14-31-31-50.

Niels Broekhuijsen
Niels Broekhuijsen

Niels Broekhuijsen is a Contributing Writer for Tom's Hardware US. He reviews cases, water cooling and pc builds.
13 Comments Comment from the forums
  • bigshootr8
    at this point though what does it really do performance wise. I get it processing wise but with ram speed I'm just a bit confused with how applications are now days are you really going to improve things a ton with memory speeds at this point I know the number is nuts but still question remains.
    Reply
  • brandonjclark
    This is all about marketing. They literally pay these guys to OC in hopes of a ROI.
    Reply
  • Sonny73N
    Hmm... How fast does 4285.6 MHz 14-31-31-50 compare to my 1333MHz 7-7-7-21?
    Reply
  • maxiim
    Cool...now let me see the cpu clock
    Reply
  • slomo4sho
    With those timings, this overclock will likely perform around the level of a readily available case 11 2800MHz kit.
    Reply
  • Evolution2001
    10961647 said:
    at this point though what does it really do performance wise. I get it processing wise but with ram speed I'm just a bit confused with how applications are now days are you really going to improve things a ton with memory speeds at this point I know the number is nuts but still question remains.
    It's not about usable performance. There's two things here.
    1) Knowing that it's possible to do with a specific configuration. Think of it more along the lines of Proof of Concept. Proof of Concept doesn't have to have an immediate practical use.
    2) It's about doing it for the sake of doing it.

    It's the reason why we drive cars easily capable of exceeding 100mph/160kph, yet very few of us ever go there on a daily basis. :)

    Reply
  • Amdlova
    when amd will uses 256bit interface on apu ??? maybe with 4000mhz can do 3850 job
    Reply
  • InvalidError
    10962137 said:
    Hmm... How fast does 4285.6 MHz 14-31-31-50 compare to my 1333MHz 7-7-7-21?
    That depends a lot on whether your software is more bandwidth-bound or latency-bound.

    Since this memory is running at over 3X your memory's clock speed, the latencies are equivalent to 5-10-10-16 at 1333MT/s which is not that bad. The much higher bandwidth allows filling cache lines much faster and would likely make up for most of the loss on timings.

    In general, the only type of code where lower latency systematically wins over bandwidth is very branchy and unpredictable code like compilers. For most mainstream software though, CAS latency and speed grades can be traded for each other with little to no net effect on performance.
    Reply
  • Mike Honcho
    Yay LN2. Move along folks, nothing to see here.
    Reply
  • wiinippongamer
    cool story, bro. There've been 2400mhz (4800mhz effective) kits on the market for some time now.
    Reply