TEST: CPU Bottlenecking In A Gaming System

[PooPipeBoy]

Grab your lab coats and safety glasses, because it's time to do some SCIENCE.

BEFORE POSTING...at least read the Conclusions section, so that you at least know half of what's going on.

 

WHAT IS BOTTLENECKING?

Bottlenecking is the term used to describe the decrease in computer system performance if the CPU is not powerful enough to support the graphics card at maximum performance, or vice versa, limiting overall system performance dramatically.

 

THE AIM OF THE EXPERIMENT

1. Test if bottlenecking actually exists and if it's worth bothering about.
2. Determine whether increasing CPU performance produces any meaningful results when gaming.
3. Calculate if increasing CPU performance increases FPS stability when gaming.

THE COMPUTER SYSTEM

My Core 2 Duo E8400 is quite obviously a bottleneck in my gaming system, and it overclocks like a boss so this is the perfect opportunity to test out some theories.

At this point in time, I will test two configurations:

1. Stock speed at 3.0GHz
2. Overclocked to 4.2GHz

Soon I will be upgrading to a quad-core processor, so I also want to test this upgrade eventually as well.

 

GAMES THAT WILL BE TESTED

• The Elder Scrolls - Skyrim
• Test Drive Unlimited 2
• Dirt 3
• Minecraft 1.8.8

All games are tested at the maximum possible settings at 1080p with no V-sync.

 

METHODOLOGY

For each game and system configuration, I will:

1. Run a Fraps benchmark for 60 seconds while I complete the same in-game task (driving, fighting, running, whatever).
2. Take the Fraps results and calculate the average FPS, minimum FPS and standard deviation.
3. MSI Afterburner will be used to check whether the games I'm playing are CPU or GPU-limited.

 

BENCHMARK RESULTS

 

Stock E8400 @ 3.0GHz

Skyrim

Average = 36.4 fps

Minimum FPS = 27 fps

Standard deviation = 4.31

Test Drive Unlimited 2

Average = 53 fps

Minimum FPS = 41 fps

Standard deviation = 9.14

Dirt 3

Average = 74.3 fps

Minimum FPS = 62 fps

Standard deviation = 7.78

Minecraft 1.8.8

Average = 101.7 fps

Minimum FPS = 28 fps

Standard deviation = 20.79

 

Overclocked E8400 @ 4.2GHz

Skyrim

Average = 37.7 fps

Minimum FPS = 30 fps

Standard deviation = 6.6

Test Drive Unlimited 2

Average = 57.7 fps

Minimum FPS = 47 fps

Standard deviation = 3.81

Dirt 3

Average = 87.5 fps

Minimum FPS = 70 fps

Standard deviation = 9.39

Minecraft 1.8.8

Average = 129.6 fps

Minimum FPS = 76 fps

Standard deviation = 15.89

 

MSI Afterburner Results

• Skyrim = GPU-limited
• TDU2 = GPU-limited
• Dirt 3 = CPU-limited
• Minecraft = CPU-limited

 

CONCLUSIONS

1. Average FPS increased dramatically (17 to 27%) in CPU-limited games with the CPU overclock, and increased marginally in GPU-limited games (3 to 9%).
2. Minimum FPS increased in all games by about 13% in the overclocked configuration, except for Minecraft with a whopping 2.7x increase in minimum FPS.
3. Increased CPU performance does not conclusively improve FPS stability, as results didn't show any consistent trend in terms of standard deviation.
4. Any games with an average CPU usage of 90% or higher before overclocking (Dirt 3 and Minecraft) showed the best gains in FPS after overclocking.
5. Since my monitor runs at 60Hz, I only saw a maximum of 60fps, so I actually didn't notice any real change in the gaming experience from my perspective.

[Minsekt]

tldr to check if your cpu is bottenecking: set your resolution lower, if you still get the same fps, it does. 

[manikyath]

pretty sure this ligns up perfectly with what jayztwocents said on tech talk quite a while back.

 

in short: its complicated.

even when the gpu seems to be at 100% load, sometimes there can still be a decrease in game performance if the cpu cannot keep up well.

 

that said, people are driving it WAY too far lately, seeing "bottlenecking" as something that should be avoided, but "bottleneck" means that one component in your system is the weakest link, and there'll always be a weakest link.

 

1 minute ago, Minsekt said:

tldr to check if your cpu is bottenecking: set your resolution lower, if you still get the same fps, it does. 

theoretically yes, but theres too many variables to do this consistently, as well as games that dont really change gpu usage with resolution much.

 

i consider the best method to just monitor GPU usage, and CPU usage per core, if the cpu caps out, its necking.

[Moonzy]

Just now, PooPipeBoy said:

 

why bother doing all of these (even statistics calculation blek)

when you can just use MSI afterburner, play games and tab out to see which one hits 100% full time, and thats your bottleneck

 

Just now, manikyath said:

but "bottleneck" means that one component in your system is the weakest link, and there'll always be a weakest link.

and its normally the gpu bottlenecking the system these days lel

[PooPipeBoy]

3 minutes ago, manikyath said:

pretty sure this ligns up perfectly with what jayztwocents said on tech talk quite a while back.

 

in short: its complicated.

even when the gpu seems to be at 100% load, sometimes there can still be a decrease in game performance if the cpu cannot keep up well.

 

that said, people are driving it WAY too far lately, seeing "bottlenecking" as something that should be avoided, but "bottleneck" means that one component in your system is the weakest link, and there'll always be a weakest link.

 

theoretically yes, but theres too many variables to do this consistently, as well as games that dont really change gpu usage with resolution much.

 

i consider the best method to just monitor GPU usage, and CPU usage per core, if the cpu caps out, its necking.

It's complicated, yeah.

Bottlenecking does clearly exist, but the reason I did this is because nobody has ever bothered to test exactly to what extent it exists and how much it actually affects game performance.

It has to be wondered whether bottlenecking is serious enough to warrant spending an extra $100 to get a better CPU, but the jury's out on that one.

2 minutes ago, huilun02 said:

E8400?

 

Uh... ok

Small-scale testing helps amplify the effects, and....I'm kinda broke.

[manikyath]

49 minutes ago, Moonzy said:

nd its normally the gpu bottlenecking the system these days lel

usually, unless chosen otherwise purposely.

 

on my first gaming rig i had to purposely bottleneck applications to the cpu because otherwise the gpu would fall apart.

(vsync was broken)

[ThomasD]

The other thing this test shows is that the application in question is as important as any particular hardware specifications.  

 

Discussing "bottlenecks" absent any mention of the actual applications involved is pointless if not ignorant.