DDR4 Memory Scaling on Intel Z170 – Finding The Best DDR4 Memory Kit Speed

Jump To:


Metro Last Light and GTA V


Metro: Last Light is a first-person shooter video game developed by Ukrainian studio 4A Games and published by Deep Silver. The game is set in a post-apocalyptic world and features action-oriented gameplay with a combination of survival horror elements. It uses the 4A Game engine and was released in May 2013.


Metro: Last Light was benchmarked with very high image quality settings with the SSAA set to off, Tesselation on normal and 4x AF. We used the game titles built-in benchmark (seen above) and ran it 3 times at a 1080P screen resolution to ensure accurate results.


Benchmark Results: In Metro: Last Light with the NVIDIA GeForce GTX 980 Ti discrete desktop graphics card installed in the system we saw a jump in performance between DDR4-2133 and DDR4-2400 and again between DDR4-2666 and DDR4-2800. The performance between DDR4-2800 and DDR4-3466 was flat and then we saw 0.5 FPS increased with the move to DDR4-3600 and DDR4-3733 clock speeds. We were shocked to see performance gains at the end after no gains for several clock speeds, but we’ll take it and it was repeatable. 


Grand Theft Auto V, currently one of the hottest PC games, was finally released for the PC on April 14, 2015.  Developed by Rockstar, it is set in 2013 and the city of Los Santos.  It utilizes the Rockstar Advanced Game Engine (RAGE) which Rockstar has been using since 2006, with multiple updates for technology improvements.



In Grand Theft Auto V we set the game to run at 8x MSAA with 16x AF and normal image quality settings as we didn’t want the GPU to bottleneck the system too bad, but wanted a setup that your average gamer would actually play on. GTA V estimated that we’d use 2GB of our 6GB frame buffer with these settings, which isn’t too taxing and should give us a good frame rate. V-sync was of course disabled.


Benchmark Results: In GTA V we found hardly any difference between running DDR4-2133 and DDR4-3733 as there was less than a single frame per second difference found after running the games built-in benchmark three times and averaging the numbers. The DDR4-3000 result of 106.0 FPS on average was about as good as you’ll get as we only got 0.1 FPS higher by going up another 733MHz. 

Jump To:
  • David Keller

    Excellent review, thanks!!

  • emtee

    Sigh, more useless ‘memory’ benchmarks… you guys shouldnt even benchmark anything if you don’t know what you’re doing… memory bandwidth will only make a significant difference in CPU limited titles and/or scenarios, something simply never tested by anyone…

  • John Truckasaurus

    DDR4 2133MHz: dirty RAM peasants.
    Looks like I will be struck down by the tyrannical mainboard manufacturers and their ITX-caprices.

  • Ray Koopa

    So I recently bought the Ripjaws F4-3400C16D-16GVK, categorized as DDR4-3400, and the Asus Z170 Pro Gamer Edition, but if I look into the BIOS, it says it runs at 2133 MHz? I’m a bit new here… shouldn’t it say it runs at 3400 MHz? Or is the categorization unequal to real speed?

    • AtheO

      I think you need to enable XMP in the BIOS.

      • Ray Koopa

        I tried that, it asked me to use some optimized settings or Intel stock ones… I tried with both, but with both my PC didn’t boot up anymore and I had to reset back to the previous settings…

        • Jared Henderson

          Dude I have the same issue. Manually changed it to 2400mhz and still won’t boot. Xmp profile 1 doesn’t work either

        • Ray Koopa

          I couldn’t get my system to boot at all with XMP on. It always just turned on and black screened, not doing anything, and reset itself after 1 minute.

        • Kevlar

          I know these comments are really old but I had the same issue on my ASRock board. I had to set the XMP and then manually set the ram’s rated voltage. After rebooting it was running at the XMP speed. Also make sure your bios is up to date for the best compatibility.

        • Mikołaj Łajming

          My computer needed about 3-5 minutes to boot after setting on XMP and after every change in RAM frequency, voltage or so, but after that it booted up in no time as always

    • Robcabob1

      Visit support for motherboard and see if it has compatibility with that specific RAM. If not, you may not get full use. MSI has complete specifications for it.

      I personally bought the wrong RAM and could only OC up to one below mac (3200mhz RAM could only boot one choice under max ;30??mhz)

  • Dave4321


  • Justifiable187

    For Handbrake test, it has been my experience that the read from the source drive is the bottleneck. With that, I would really like to see someone create a ram drive, put the Big Buck Bunny source into it, and see those results.

  • Aron_1

    Do these results hold on a non-overclocked CPU on a Z170 board and is this still possible without a K series CPU? 🙂

  • THU31

    You should not test the average framerate, you should test the minimum framerate, because that is what is mostly affected by system memory speed. In CPU-bound scenarios you can often get 5-10 more FPS with faster memory, which is significant.
    You do not change your system memory too often, it might last you until the next memory standard comes (so probably way more than 5 years), which means investing in faster memory is not a bad idea.

    • Iluv2raceit

      You make a good point, but you forgot to mention that system memory will only become a factor with frame rates based upon the game’s coding as well as the resolution that you play the game.

      • Johan Krüger Haglert

        “only become a factor with frame rates based upon the game’s coding”

        lol what? And what do you mean isn’t included in that?

    • Ian Barlow

      “…you should test the minimum framerate…”

      Agreed. Plenty of others have done this and the benefits are always shown to be significant. No disrespect intended, but these tests tell us almost nothing about the benefits of faster RAM for fairly obvious reasons.

  • Faiakes

    Could the article be updated with figures using the just released HyperX Savage line?
    It seems to be 1 CL less than the G.Skill up to 2800.

    • legitreviews

      Re-do over 225 benchmarks for 1CL less? The performance might be a tad higher and the scaling should be about the same, so that’s days of work for basically the same results!

      • Faiakes

        Hey, a guy can hope! 🙂

        Thanks anyway

  • Faiakes


    I see your RAM runs at 1T.
    How is that even possible!
    Can you share how you managed it (BIOS settings, etc)?

    • legitreviews

      We manually set the timings and you can see where the command rate value is set in slide 5 in the last gallery on the first page of the article!

      • Faiakes

        Thanks for the prompt reply.

        So all it took was manually setting DRAM Command Rate to 1?
        Nothing else was needed to get the RAM running at 1T?

        • legitreviews

          Nope, just change the setting manually from 2T to 1T!

        • Faiakes

          Well…I thought there was more to that.

          Excellent. I will try it and get back to you.

  • DinamykPT

    The older Ripjaws 4 memories aren’t compatible with z170-based motherboards?

    • Nathan Kirsch

      The XMP and SPD profiles likely were not set for Z170 based boards, so you’ll need to manually adjust the settings on some kits.

  • Noshuru

    Would it be possible to repeat the GTA V test with these settings: http://www.eurogamer.net/articles/digitalfoundry-2015-intel-skylake-core-i7-6700k-review

    (All Ultra, no AA, 980 Ti or Titan X clocked as high as they go, advanced tab maxed out)
    I’m really quizzed over these results. My current working assumption is that Skylake performs so well because even the lowest clocked DDR4 RAM available is 2133MHz. One of the settings in the advanced tab, namely Extended Distance Scaling, is hugely CPU dependent. With that setting off I get a whopping 40FPS more.
    Personally, I have a 2500K running at 4.8GHz, DDR3 RAM at 1600MHz. I get 64.8FPS with that setup.
    Hope to hear from you!

  • ramonzarat

    Love those scaling article! I would like to see i7-920 to 6700K @ 4GhZ scaling!

    • legitreviews

      Thanks, next time I’ll be sure to give 7-zip a shot!

  • musef1

    Interesting article.

    Just one thing, I noticed that in your chart on the final page, it says lower = better. Surely it should be higher = better?

    • Nathan Kirsch

      Good catch and a new chart has been inserted!

      • Ryan Snaca AniloHectycle

        Without a stock memory controller to a typically achievable overclocked comparison this is a very silly benchmarking/conclusion to be sure.

        • David Keller

          I’ve done a lot of benchmarking myself with my older 4790K with uncore between x44 and x48 and the only really benefits I ever saw were faster cache speeds and some latencies, up to a point. At one point cache speeds actually started to suffer. Which, as with most hardware, there is a sweet spot that is most efficient. OCing the uncore hasn’t produced any real world benefits that I’ve noticed. Although, that is one configuration and it’s DDR3. Can’t claim it’s universal. I have seen improved performance in Crysis 3 with faster memory speeds. One example that sadly, is pretty uncommon.

  • Anton

    Did you measure the power usage for the different frequencies? As the performance is quite similar throughout the test the temperature and power would be an interesting consideration when choosing the modules.

    • Nathan Kirsch

      From DDR4 2133 to 3733 we used 1.35V across the board, so it would likely be another flat line on a chart to look at. Temperatures would have been interesting to look at, but it’s a non-issue at these speeds as long as you have airflow in your case since we were only running 1.35V.

  • Shadowdane

    Wish you would of tested SLI configurations with the various memory kits. I’ve found on Haswell at least faster memory kits have a rather big impact with Multi-GPU configurations. For example I tested GTA5 on my system. Going from DDR3-1600 to DDR3-2400 saw a 8% performance improvement.

    • Nathan Kirsch

      I’ll toss in another GeForce GTX 980 Ti into the system and see if I spot a big jump up like that!

  • Selena Gomez

    Need better benchmarks for games. Higher speed RAM generally shows some (small) gains in situations that are CPU bound/bottlenecked.

    • MewkaKazami

      I’m pretty damn sure if you’re buying Skylake thats not gonna be a problem considering only i5-6600k and i7-6700k are out

      • Selena Gomez

        Go look up benchmarks, Skylake isn’t some huge advancement, it’s only slightly faster than Haswell so there’s obviously still going to be games that are more CPU bound than GPU.

        • MewkaKazami

          It’s actually not about performance but stability. Check Eurogamers review and you’ll see a lot of games get a huge improvement in minimum FPS over the old CPUs

        • shadowhedgehogz

          Look further down the page to the OC gfx card results, pretty much the same perf on both Haswell and Skylake. x99 is the better buy.. Also that small improvement is nowhere near “huge” as you say, a high end GPU upgrade would make a much much bigger difference.

        • Dan Eagen

          If you don’t plan on overclocking (99% of the population won’t) then Skylake presents massive improvements.

        • Mirko Opacic

          Like always when he throw out new cpu…few % better and more expensive then previous.

          Fucking Intel pussy…go amd kill them all

    • Nathan Kirsch

      Have any suggestions as to what better benchmarks would be for game titles? I always want to test using the best benchmarks possible.

      • Selena Gomez

        I’d like to see Arma 3. There are lots of other games although most only become CPU bound when you’re already at very high frame rates so the performance differences won’t be noticeable in the real world. Arma 3 on the other hand (all I can think of right now) is heavily bottlenecked, even a Skylake @ 5 Ghz probably wouldn’t be able to get 60 fps consistently in populated servers.

        Just make sure that whatever you test, since you’re not focusing on the GPU performance, ensure that the GPU is never near 100% load and that the bottleneck is clearly the CPU.

        • Nathan Kirsch

          For memory testing I hate using FRAPS as often you are seeing just tenths of a FPS difference between clock speeds or timings. The human element adds the PEBKAC issue, so I prefer to use built-in benchmarks for testing like this. I see some tools are out there for ARMA 3, but the results are scores and not FPS from the looks of it – https://forums.bistudio.com/topic/142875-arma3mark-benchmark-your-arma-3/

          Do you know of a tool for ARMA 3 that gives results in FPS?

        • lol

          There is a benchmark existing which shows its results in FPS (instead of a score): https://forums.bistudio.com/topic/177406-yet-another-arma-benchmark/ would be great to see Arma 3 included in your benchmarks in the future 🙂

          I am using a souble-sided (IC’s) G.Skill F3-2400C9D-8GTXD Kit (@ 2600 Mhz CL10-12-11-31 1T) + SSD in my 4,5 Ghz Haswell System because Arma requires that much bandwidth…

        • lol


      • JeffreyQC


        It’s possible to run the X264 HD benchmark? This Skylake dual channel DDR4 come close to the Quad Channel Intel Core i7-5960X. Skylake is faster for write speed, not bad!
        (For a cheaper CPU)

  • anon

    Very interesting article indeed.
    I would like to point out two things,
    In freq-timing vs. anything charts it would be interesting to use the y-scale as per unit and add at each measured point the actual value. This is very helpfull to keep a proportional scale and give perspective of the real performance impact.
    Also it would be interesting to see, in the first part of the article, a brief comparison of similar memories used in other platforms.
    These to points are meant to add perspective to the article.