G.SKILL Ripjaws 8GB 2133MHz DDR3L SO-DIMM Memory Kit Review – F3-2133C11D-8GRSL

Jump To:

DDR3L 2133MHz Memory Kit Benchmark Results:

SiSoftware Sandra 2014 w/ DDR3L 1600MHz:

sandra-1600 

SiSoftware Sandra 2014 w/ DDR3L 1866MHz:

 sandra-1866

SiSoftware Sandra 2014 w/ DDR3L 2133MHz:

 sandra-2133

SiSoftware Sandra 2014 showed that the NUC went from having 17.33 GB/s of memory bandwidth at 1600MHz up to 21.73 GB/s with the 1866MHz kit and then it dropped down to 21 GB/s with the 2133MHz memory kit. We were expecting to see an increase in performance from 1866MHz to 2133MHz, but we actually saw a small decrease in memory bandwidth.

AIDA64 v4.00 Beta w/ DDR3L 1600MHz:

 aida-1600

AIDA64 v4.00 Beta w/ DDR3L 1866MHz:

 aida-1866

AIDA64 v4.00 Beta w/ DDR3L 2133MHz:

 aida-2133

AIDA64 v4.00.2752 Beta showed the memory read went from 19525 MB/s to 23573 MB/s, which is a 14% increase in memory performance by going from 1600MHz to 2133MHz. Also notice the latency went from 90.7 ns to 83.6 ns, which is a 7.8% decrease. The performance differences between 1866MHz and 2133MHz are fairly minor once again.

Futuremark 3DMark w/ DDR3L 1600MHz:

 3dmark-1600

Futuremark 3DMark w/ DDR3L 1866MHz:

 3dmark-1866

Futuremark 3DMark w/ DDR3L 2133MHz:

 3dmark-2133

In the latest build of 3DMark we found solid performance with a score of 43,624 in Ice Storm with 1600MHz memory, 45,481 with 1866MHz memory and finally 46,804 with 2133MHz memory.  The scores went from 702 in Fire Strike with 1600MHz memory, 710 with 1866Mhz memory and 758 with 2133MHz memory. This is a 7.2% performance gain in Cloud Gate, which is the ideal 3D benchmark for a desktop system like the NUC with Intel HD 5000 Graphics. The performance between running the memory at 1866MHz and 2133MHz was found to be 2.9%.

Cinebench R15 w/ DDR3L 1600MHz:

cinebench-1600 

Cinebench R15 w/ DDR3L 1866MHz:

cinebench-1866 

Cinebench R15 w/ DDR3L 2133MHz:

cinebench-2133

Moving along to Cinebench R15 we found the Intel NUC with the Intel Core i5-4520U processor scored just a tad higher on the CPU test, but the Intel HD Graphics 5000 jumped up from 25.06 FPS to 28.40 FPS on the OpenGL benchmark. It should be noted that single core CPU performance was significantly degraded with the 2133MHz memory kit installed and we have reached out to Intel to see what could be causing this and they are looking into it as well. We’ll take a 13.3% performance gain on the graphics side by switching out a memory kit any day!

TrueCrypt w/ DDR3L 1600MHz:

truecrypt-1600 

TrueCrypt w/ DDR3L 1866MHz:

truecrypt-1866 

TrueCrypt w/ DDR3L 2133MHz:

truecrypt-2133

The last performance test that we wanted to run was TrueCrypt and we found the mean score of the AES algorithm benchmark test went from 1.3 GB/s to 1.4 GB/s. This isn’t a huge increase, but it’s still a 7.7% performance gain on a test that isn’t really memory bandwidth limited to begin with.

Print
Jump To:
  • Dom

    Hi
    I have also a D54250WYK NUC and bought the G.Skill F3-2133C11D-8GRSL Kit after reading your interesting article! ;-). After a weekend of testing i can say the following: My Firestrike Benchmarks are not so good like yours:
    @1600mhz: 635
    @1866mhz: 665
    @2133mhz: 684
    I am using Windows 8.1 with all the newest patches and drivers. NUC Bios is also the newest… I have 2 kits of this memorys and getting the exact same result in both.

    To compare: With Corsair Value Select CMS08GX3M1C1333C9 (2 x 8 GB) i am getting 665 @1333mhz and 692 @1600mhz! (in faster setting it will not start)
    Do you have an idea what’s wrong here?
    Regards, Dom

  • Olly

    Nathan, for stability and heat load would you prefer the G.Skill 1866MHz or this new 2133MHz memory? I was just about to buy when I saw this review.

    • Nathan Kirsch

      Flip a coin… I run the 2133MHz kit in my NUC as it does everything the 1866MHz kit can do and run at 2133MHz, which my 1866MHz kit couldn’t do!

  • DougInOK

    The KO in the chip part number says it is a DDR-1600 chip. So perhaps G.Skill just re-binned Samsung’s already binned chips

    Would only having 8 chips per stick have less heat output than having 16? Or does 16 chips per stick give more surface area to cool itself?

    • Nathan Kirsch

      Yup, they are sorting this chips and saving the best clocking ones for these modules. As for heat…. That is a great question and I’ll have to see if I can see any differences with my temp probe.