Cossacks 3 - performance benchmark

Cossacks 3 is a very interesting and unique real time strategy (RTS) game. Sadly, due to its bugs it is not very popular.

There are many guys making benchmarks of different games with different GPUs, but none of them measuring performance of the Cossacks 3 game. Since I have access to multiple GPUs, I have decided to measure performance myself and share my results with you.

How I tested the game

For benchmarking purposes I have created a 1 vs 1 fight using the game editor. The fight is very similar to what you can see in the 30 minutes peace time games between two players. The benchmark battle includes the following units from both sides.

  • 400 rundashirs in a formation.
  • 400 18th century musketeers in a formation.
  • 160 + 90 18th century dragoons in two formations.
  • 100 monks from the cathedral.

For measuring game FPS the FRAPS tool was used. Here are the benchmark steps I performed five times for every test setup.

  • Load the benchmark saved map.
  • Move game focus so it covers rundashir formations from both sides.
  • Press F11 to start FRAPS benchmark.
  • Press Ctrl + A and order an attack to the left bottom corner via mini map.
  • Switch team by clicking 2 on the numpad.
  • Press Ctrl + A and order an attack to the left top corner via mini map.
  • Observe the battle until no more shooting units left.
  • Press F11 to stop FRAPS benchmark.

You can download the save file for the benchmark battle and try it yourself - benchmark.map.

Hardware tested in the game

Motherboard: MSI Z170 PC Mate
CPU:
Intel Core i7-6700k @ 3.9Ghz with hyper-threading disabled, simulation of i5-6600k
Intel Core i7-6700k @ 3.8Ghz with two cores disabled, simulation of i3-6300
Intel Core i7-6700k @3.5Ghz with two cores and hyper-threading disabled, simulation of G4500
RAM: Corsair Vengeance Lpx 32GB 2400MHz DDR4 @ 3000Mhz, CL16
SSD: Samsung 850 EVO 500gb
Power supply: EVGA SuperNOVA 750 G2
Operating system: Microsoft Windows 10 Pro x64
NVidia drivers: GeForce Game Ready Driver 381.89 WHQL
AMD divers: Crimson ReLive Edition 17.4.4
GPU:

Graphic cards test results

Honestly, I did not expect to see such results. It looks like Cossacks 3 game is strictly CPU limited, since the performance between GTX 1060, GTX 1070, and GTX 1080 almost identical. The performance is not only identical between the cards, but also between 1080p and 1440p resolution. Even much cheaper RX 460 shows almost the same minimal frame rates.

CPU performance benchmark

I was going to test two GTX 1080 in SLI configuration, but the previous benchmarks show, that adding extra GPU power will not improve performance in this game. So, instead I am going to test some CPU configurations.

Unfortunately, I do not have access to wide range of CPUs to test, but I can play around with CPU frequency and cores to try to simulate Intel Core i5 and i3 level of performance.

Two core with hyper-threading disabled running at 4.3Ghz brings us the best minimum value of 26 FPS, while fully enabled i7-6700k at 4.2Ghz shows the best performance overall with 50 FPS average. The G4500 configuration shows the worst result across all metrics, not much surprise here. Two cores clocked at just 3.5Ghz are producing only 33 FPS on average.

As we can see, both number of cores and the core frequency matters for this game. More cores rises up the maximum FPS number, while stronger core helps with minimum FPS number. Since minimum FPS value is much more important for pleasant game play, I would recommend to pick a CPU with strongest possible single core performance and overclock it, if that is possible. For now I can only hope, that developers will optimize the game and it will utilize all of the cores available in a CPU.

* - Simulated CPUs. Performance of real i5 and i3 might be lower due to different cache sizes and turbo clock modes.

RAM performance benchmark

Much to my surprise, the memory clock actually matters for this game as well.

DDR4-2133Mhz lags behind DDR4-3000Mhz for more than 10fps on average, but wins with 1fps on minimums due to lower CL15 latency, compared to CL16 for DDR4-3000Mhz. DDR4-3200Mhz does not bring extra performance over DDR4-3000Mhz, but rises up the minimal frame rates by 1fps and matches lower latency DDR4-2133Mhz.

System performance benchmark

In the end I compare performance between some systems which are actually being used, and not assembled for this test.

  1. Extreme gaming PC - i7-6850k @ 3.8Ghz, 64gb DDR4-3000Mhz, NVidia GTX 1080, Samsung 960 EVO 500gb, 1440p
  2. Old extreme gaming PC - i7-4930k @ 4.5Ghz, 32gb DDR3-3200Mhz, NVidia GTX 1070, Samsung 850 EVO 500gb, 1440p
  3. High end gaming PC - i7-6700k @ 4.2Ghz, 32gb DDR4-3000Mhz, NVidia GTX 1060, Samsung 850 EVO 500gb, 1440p
  4. Dell XPS 15 9550 - i7-6700hq @ 3.5Ghz, 32gb DDR4-2133Mhz, NVidia GTX 960m, NVMe SSD 1T, 1080p

As expected, higher clocked i7-6700k in pair with GTX 1060 outperforms much more expensive, but lower clocked i7-6850k accompanied by GTX 1080. Previous generation i7-4930 with DDR3 RAM lags behind even if overclocked to 4.5Ghz.

Conclusion

The Cossacks 3 game is very not optimized for multi-threaded workflow, even high end CPUs are not able to keep minimal frame rates above 30fps mark. For the best possible performance in the game you have to look for the most recent CPU clocked as high as possible. The CPU can be paired with GTX 1060 or RX 480/580 graphics card. More expensive GPU such as GTX 1070 will help to raise the average frame rates a bit more, but it does not justify the price difference.

If I get my hands on some more hardware such as AMD RX 480/580 GPU and AMD Ryzen CPUs then I will update the benchmark charts. Stay tuned.

I hope you found these benchmarks interesting and you have learnt something. Remember to download the benchmark file, run it on your machine and share results with me.

Written by on
Views: 489 Comments: 0