Welcome Stranger to OCC!Login | Register
Guides » Gaming

Does Memory Speed Affect Gaming Performance; An Intel Question


Does Memory Speed Affect Gaming Performance: Introduction

Today's topic is one of those never ending debates of "is it a waste of money" type conversations. After recently having a very long debate on another computer forum, I decided that instead of just using and regurgitating old benchmarks, I would test this theory out for myself. So I went ahead and tested the slow and fast DDR3 and DDR4 memory on three graphically and processor heavy games. The general word on the street is the fastest memory available is at most five frames per second (FPS) higher on average in games. With this type of reasoning, a lot of people recommend spending money on other components rather than getting blistering fast memory for a relatively low performance increase in the grand scheme of things. The question today is this: How true is this generalized statement I have heard over and over for the last 10 years and how much benefit is using memory above Intel specifications? This article isn't about Windows load times, the amount of memory needed, or video encoding. Simply put; if you are already in a game, how much of an increase in FPS would there be for a given memory speed? That is the question at hand.


Memory V.S Gaming: Benchmark Setup

Because there is no way to directly compare DDR3 and DDR4 on an Intel system, I had to get creative to give some semi-comparable results. AMD was left out this time around because these two types of CPU architecture will have different results and can cause confusion. I have read that Ryzen very much depends on memory for performance, however I haven't seen results and want to test it myself. Once the Ryzen platform gets the bugs worked out, I'll look into asking this same question again using AMD hardware. For this, however, I chose to use Intels last consumer socket that uses Dual Channel DDR3 and Intels current socket that uses Dual Channel DDR4. While the X299 platform will throw some things into the mix, it seems the Kaby Lake CPU I tested here is the same, but on a different socket. Therefore, results for Dual Channel should not vary. In the end you should trust the results and not what someone thinks things could or should be. The choice was to keep things relatively comparable with systems that have four physical cores. A system with 10 cores could potentially skew the results when compared to a four or six core CPU. It would also be interesting to see what the difference between Quad and Dual channel is or simply the amount of CPU cores needed for certain games to no longer become CPU bound. But that will be for the X299 platform if or when I revisit this subject.


Testing Setup DDR4:


Testing Setup DDR3:


Memory Speeds:

Performance Index is an interesting thing that I came across a few years ago. The Index gives the users a picture of what memory should perform like in comparison to other DIMMs (dual in-line memory module). However, a lot of factors are left out such as the type of processor being used or the limitations of memory speeds themselves. The chart only uses the part of the big picture with a simple equation used by dividing the speed over that DIMMs rated CAS (Column Access Strobe). CAS (or CL) is a measurement of the time it takes between when the memory controller attempts to access a column and when it is accessed. The actual time in milliseconds can be found by dividing the CL by the frequency (i.e. 9 / 1600 = .005625 seconds or 5.625ms). In theory, low is better for all timings related to memory. While I would say the CAS is one of the most important aspects of memory speed related to access times, you can have very slow memory with tight timings that is comparable to faster memory with loose timings.

This is generally where the arguments start. Would 1600 @ CL 8 be faster than 2400 @ CL 12? Well, the Performance Index says they are the same, but it can get complicated. 2400 is the faster of the two in terms of raw bandwidth, but on the flip side is that access time is a lot higher. In the end, this article isn't about what is actually faster versus theoretically, but how the memory performs in practical tests. The memory speed I chose for the benchmarks have a wide range to give an idea of what to expect for a variety of system builds.



Memory Speeds / Timings Used:

  • DDR3 1333 - CL 9-9-9-24
  • DDR3 1866 - CL 9-10-9-28
  • DDR3 2400 - CL 10-12-12-31
  • DDR4 2133 - CL 15-15-15-36
  • DDR4 2800 - CL 15-16-16-36
  • DDR4 3200 - CL 14-14-14-38


Games and Settings:

Warhammer 40,000: Dawn of War 3 (2017)

Settings: Max

  • Image Quailty: Max
  • Texture Detail: Higher
  • Resolution Scaling: 1.00
  • Unit Occlusion: Enabled
  • Antialiasing Mode: On
  • Physics: High


Metro 2033: Redux (2014)

Settings: Custom

  • Quality: Very High
  • SSAA: Off
  • Texture filtering: AF 16X
  • Motion Blur: Normal
  • Tesselation: Very High
  • VSync: Off
  • Advanced PhysX: On


Tom Clancy Ghost Recon: Wildlands (2017)

Settings: Ultra

  • Resolution Scaling: 1.00
  • Graphics Preset: Ultra
  • Antialiasing Mode: TEMPORAL AA
  • Ambient Occlusion: HBAO+
  • Draw Distance: Very High
  • Level of Detail: Ultra
  • Texture Quality: Ultra
  • Anisotropic Filtering: 16
  • Shadow Quality: Ultra
  • Terrain Quality: Ultra
  • Vegetation Quality: Ultra
  • Turf Effects: On
  • Motion Blur: On
  • Iron Sights DOF: On
  • High Quality DOF: On
  • Bloom: On
  • Godrays: Enhanced
  • Sub Surface Scattering: On
  • Lens Flare: On
  • Long Range Shadows: Ultra

  1. Memory V.S Gaming: Introduction & Setup
  2. Memory V.S Gaming Testing: DDR4 1920 x 1080
  3. Memory V.S Gaming Testing: DDR4 2560 x 1440
  4. Memory V.S Gaming Testing: DDR3 1920 x 1080
  5. Memory V.S Gaming Testing: DDR3 2560 x 1440
  6. Memory V.S Gaming: Conclusion
Related Products
Random Pic
© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.1574881077   (xlweb1)