banner-pcinside-v6

If you are gaming on your pc then your frame rate will usually be limited by either your cpu or your gpu. If you do not know how to check this then I recommend the following article: How to check your frame rate and gpu usage in any game. If you do know the difference then the following article might interest you.

Detail settings

First let’s take a look on how the different graphic settings affect the in-game visuals.

Level of detail

Very HighHighMediumLow

Cities Skylines with the graphic settings detaills on very high.

Cities Skylines with the graphic settings detaills on high.

Cities Skylines with the graphic settings detaills on medium.

Cities Skylines with the graphic settings detaills on low.

Shadows Quality

HighMediumLowDisabled

Cities Skylines with the graphic settings shadows quality on high.

Cities Skylines with the graphic settings shadows quality on medium.

Cities Skylines with the graphic settings shadows quality on low.

Cities Skylines with the graphic settings shadows disabled.

Shadows Distance

Very FarFarShortVery Short

Cities Skylines with the graphic settings shadow distance very far.

Cities Skylines with the graphic settings shadow distance far.

Cities Skylines with the graphic settings shadow distance short.

Cities Skylines with the graphic settings shadow distance very short..

Texture quality

This might be subtle but the texture quality can be noticed on the car models and the buildings. Check the white office on the right side and the black hearse.

LowMediumHigh

Cities Skylines with the graphic settings texture quality low.

Cities Skylines with the graphic settings texture quality medium.

Cities Skylines with the graphic settings texture quality high.

CPU scaling

Before we begin testing I need to note that Cities Skylines has a frame rate limiter linked to your display’s refresh rate (Hz). Therefore on my  hardware Cities Skylines was running at the maximum frame rate  (6) at the maximum graphics quality. I tested this game on an Intel i5 2500K that was overclocked to 4.2 Ghz with 8 GB of DDR3 1600 Mhz ram.

How many cores do you need for Cities Skylines?

So how much CPU power does Cities Skylines need? For this test I zoomed out maximum, kept all details at maximum and decided to disable cpu cores and see what happened.

Cities Skylines fps graph with one, two, three and four cores enabled.

On decent hardware Cities Skyline only requires two cores. There is no added benefit of owning a quad core processor on the highest settings. Keep in mind that this was tested at a city of 51.000 inhabitants. Larger cities require more computational power. But this does indicate that a modern cpu will have little trouble with the highest details.

How well does Cities Skylines do multithreading.

For this test I checked the cpu usage with different cores disabled. I started with 1 core and then moved my way up to all 4 cores enabled.

Cities Skylines with four cores enabled

As demonstrated Cities skyline actually scales awesome on multi-core. Meaning that each time you add another core the load on the previous cores will drop. This is good news for all AMD owners that are limited in single threaded performance like the AMD FX 8320. It is also good news for Intel owners with HyperThreading (like the i7 on desktop and the i5 on mobile platforms).

Cities skyline actually scales awesome on multi-core.
If we lower the CPU speed to 1.6 GHz the game will fully utilize all cores!

 

How do detail settings effect frame rate in Cities Skylines

I would like to remind you that my gaming rig was always running at the maximum frame rate of 60 fps. To simulate a slower cpu, or processor, I ran the game on a single 4.2 Ghz core and bench marked the different graphics options.

Level of Detail

Cities Skylines fps graph with detaills on low, medium, high and very high.

 Shadows Quality

Cities Skylines fps graph with shadows disabled, low medium and high.

The level of detail has a considerable impact on cpu usage: FPS was increased by 35%. Shadow quality also has a significant impact on cpu usage. FPS was increased by 46% . The other settings such as Texture Quality, Anisotropic filtering ant Anti-aliasing had no effect on the frame rate during this test.

Recommendation: If your city is getting bigger and your simulation starts drawing more cpu cycles you should consider lowering details and shadow quality. Personally I would start with disabling shadows and leaving the details as high as possible.

What sort of cpu do I need for Cities Skylines

The fact that I was able to hit 60 FPS on maximum quality with only 2 cores is a good indication that this game does not require the most expensive cpu ever made. In fact I was able to run Cities Skylines on my Windows 8.1 Surface Pro 2 tablet that is equipped with mobile dual core i5 4300U. Although it only did 14 fps due to the slow integrated graphics it did so with only 30% cpu usage on low detail settings.

Therefore I am pretty confident that you can run this game on any i3, i5 or i7 cpu from 2011 or newer. This includes mobile cpu’s. For AMD things get a little bit more complicated because I only have an AMD E-350 available. However I can ballpark this by comparing the single threaded passmark score  with AMD alternatives. My best guess is that even the FX 4300 quad core should be able to handle this game on the highest detail settings. Disclaimer: You can run into issues after passing the final 80.000 inhabitants milestone when the simulation calculation becomes more significant.

Gpu scaling

Again I would like to remind you that on my hardware Cities Skylines would run at the maximum frame rate of 60 FPS. In order to test the impact of different settings I lowered my video cards core clock by 500 Mhz. This way I ensured that my rendering was limited by the GPU and that I could bench mark the impact of different options on the frame rate.

Texture quality

As expected texture quality does not affect frame rate. However it does affect the graphics memory used by the game. Most video cards should be able to handle high texture quality.

Cities Skylines graphics low, medium and high memory usage.

Level of Detail

Level of Detail has a large impact on frame rate. Medium and Low are both limited to 60 fps, however the low setting uses considerable less resources. The difference between low and high is about 63%.

Cities Skylines fps graph level of detaill, low medium, high and very high.

Shadows Quality

As expected shadow quality has a reasonable impact on frame rate. The difference between disabled an high is about 18%.

Cities Skylines fps graph shadows quality disabled, low medium and high.

Shadows distance

This test left the Shadow Quality on high but instead lowered the draw distance of these shadows. Basically you need to change  this setting to very short to have any effect.

Cities Skylines fps graph shadows distance very short, short, far and very far.

Anti-aliasing

Anti-aliasing did not seem to do much on my downclocked videocard. The quality was not perfect either though.

Cities Skylines fps graph anti aliasing disabled and enabled.

Recommendation: If you are gpu limited then start with changing Shadow distance to very short and lowering level of detail to high. If you need more frames then disable shadows completely. If you need even further then start lowering level of detail all the way back to medium or low. Texture quality and anisotropic filtering have little effect.

Questions?

Feel free to post questions or other remarks in the comments.

GameplayInside recommends gaming on:

Comments:

  • Jakkar

    Thanks for the analysis, Sebastiaan. May I ask – were you able to achieve a stable 60fps while *zoomed in*, with all cores enabled and no experimental throttling? At any quality settings.

    I’m attempting to analyse a cross-platform performance issue with the game, and while innumerable users have implied stable 60fps performance under any circumstances on adequately powerful systems, I’ve found no proof that this is the case for any user.

    • Sebastiaan de Kooter

      Glad you liked it 🙂

      Yes zoomed in the game is really “light”. I am getting 60 fps at all times. I have made screenshots for you here: http://imgur.com/a/Zlx80

      If you look closely you will notice the GPU is in a power saving mode. This is why I had to “nerf” my hardware in order to do any meaningful test. I have heard severall users complain about performance issues on Mac and Linux, however most of them also used inferior hardware (IGP’s etc).

      Obviously there is a point where the game “breaks” because your city is so large (150.000+) and simulation starts to slow down the game. On 89K inhabitants I am still getting a stable 60 fps when zoomed out.

      • Jakkar

        You’re most helpful. Thank you for the data. The core group of those collating data and reporting the problem are Windows and Linux users on powerful i5 or i7 processors, with at least 8 gigabytes of RAM and either Geforce 970s, or 980s, and we all drop to 20-30fps if we zoom in on our cities – even small towns, without ever bottlenecking or making full use of our CPUs or GPUs.

        Even some of the major youtubers sharing their ongoing builds admit to suffering huge performance drops during zoomed, lateral views of their creations.

  • Steve

    Thanks for this testing and analysis… I was looking for something like this.

    However, this conclusion is not the correct one, IMO:

    “As demonstrated Cities skyline actually scales awesome on multi-core. Meaning that each time you add another core the load on the previous cores will drop. This is good news for all AMD owners that are limited in single threaded performance.”

    The operating system will automatically bounce threads around to different cores giving you the effect you see here. Software that scales well on multi-core will use 100% of all cores regardless of how many cores you throw at it and will provide a proportional benefit for the extra processing power used (in games, often frame rate, but it can also be extra effects, physics, etc). Conversely, a thread that is seeing its execution bounced between two cores averaging 50% usage is no performance benefit over a thread running on a single thead with 100% usage.

    • Sebastiaan de Kooter

      “Software that scales well on multi-core will use 100% of all cores regardless of how many cores you throw at it and will provide a proportional benefit for the extra processing power used”

      Although I appreciate the effort this is incorrect. This game has a framerate limiter (like most games) meaning that there is no need for an endless amount of processing power. To base the definition of proper multi-core scaling on 100% load on all cores makes no sense. There is simply no demand for that much processing power.

      Basically a game that has “bad” multi-core properties will feature a heavy main thread. This thread controls the other threads such as the sound and ai threads. These games can be recognized by having a single core show a very high load, whereas the other cores have a very low load. For example 75%/25%/25%/25%.

      If you do not find this convincing then I could lower the multiplier of my cpu and demonstrate this :).

      • Steve

        A framerate limiter that prevents the need for more processing power means, if anything, that multi-core scaling simply cannot be meaningfully measured. What does a race between a Prius and a Mustang tell you about the performance of their enginies if both are goverened at 55mph? A little at the start of the race, maybe, but otherwise not much.

        Bottom line is if CPU usage is not expanding much beyond 25% on a 4 core machine either: 1) The game does not scale well for multi-core machines, 2) There is no benefit to more CPU power. The only way to find out which is to find ways that should stress the machine more. Bigger cities?

        What cores a process is using tells you very little about the number of threads a process is using unless affinity is being used to lock threads to a particular process.

        Here’s test to try out:
        1) Download SuperPI (single-threaded benchmark) as well as Microsoft’s Process Explorer in order to monitor thread usage.

        2) Start SuperPI on 32M.

        3) Notice CPU is not limited to a single core in task manager; it is spread across different corers. You will find depending on what else is going on on your machine different tests will yield different distribution of the CPU usage amongst cores… but typically it will be pretty even.

        4) Per this post:
        http://superuser.com/questions/462969/how-can-i-view-the-active-threads-of-a-running-program
        …open up the threads property tab for SuperPi.exe. Notice that despite 4 cores being used to some degree, SuperPi has only one thread with meaningful CPU usage.

        5) Right-click on the pi.exe child of SuperPi.exe and select “Set Affinity”. Uncheck all CPUs except CPU 0″.

        6) Now look at the CPU graphs in task manager…

        7) Re-select all CPUs in the affinity dialog and check task manager again.

        See?

        • Sebastiaan de Kooter

          Hah! That is actually an unlikely scenario but a valid point indeed! Even though I was pretty confident there was no thread bouncing here I decided to test this regardless.

          Luckily I own a 2500K cpu that can adjust its speed at will. So I changed it from x40 down to x16, making it a 4-core running at 1.6 Ghz.

          I ran SuperPi and checked the cpu load: Maximum of 25%. So indeed thread bouncing there. Now I fired up Cities Skylines: 92% cpu usage (100% for the system)! http://www.gameplayinside.com/wp-content/uploads/2015/03/cities-skylines-performance-reviews-test-for-stevie.jpg

          So clearly this game has very good multicore properties just like the article said :).

          ps: The core bouncing was sort of taken care of somewhere between 2007 and 2010. Most games tend to “stick” to a certain core. For example the MMO rift has a config that by default sticks to core 3 for it’s main thread. You can edit the rift.cfg to change it.

          • Steve

            Awesome! Good multi-core scaling indeed… Thanks!

  • Thomas Wootten

    Would you have got different results on

    • Sebastiaan de Kooter

      Would have gotten different results for sure. However I tried to demonstrate the impact of the different graphics options. To that regard the results will be very similar (just with lower framerates). I think this test shows that level of detail and shadow quality have a rather large impact.

  • lordmogul

    But it can be very CPU demanding when your city is growing:
    (run in a 3570k @ 4,3 GHz)

  • Jens Speicher

    Nice performance test! I can play Cites:Skylines on my 3 year old notebook with a Sandy Bridge i5 at 2.6 Ghz Turbo with Low details and about 30fps until i hit 100k population. Then it gets unplayable due to cpu limitation. Good, that I’ll build a nice gaming rig for 1000 bucks soon^^

  • Juliana Cardona Valencia

    I wanted to ask, i really want to play this game but i am just unsure if my laptop has the right specs, it is a i7-5500 @ 2.4GHz 2.4GHz RAM: 12 GB 64-bit Nvidia Geforce 940m could you please help me, thank you so much.