Integrated & Unified Benchmarking Tool for DCS

[Limaro]

Right now, everyone measures performance in completely different ways — with different missions, modules, and settings — which makes it nearly impossible to compare results or provide meaningful feedback.

With an integrated benchmarking system, we could all test under the exact same conditions.

 

Why this matters for players

We could finally find the optimal hardware and graphics settings for our rigs without guesswork.

We’d have a reliable way to compare upgrades (e.g., “Will a 9950X3D really outperform my 9800X3D in a heavy mission?”).

Squadrons could make sure everyone has a smooth experience before big multiplayer events.

Why this matters for ED

Consistent performance data across thousands of users would make diagnosing and prioritizing performance issues much easier.

It would help separate true “bugs” from cases where users just have mismatched settings.

It could even serve as a tool for measuring improvements with each patch and verifying optimization progress.

Example Benchmark Scenarios

GPU Test: Empty map, no units — great for isolating GPU performance.

Mixed Scenario: Moderate number of ground units and some AI aircraft — balanced test.

Full Stress Test: Large-scale battle with lots of AI, explosions, and scripting — perfect for CPU benchmarking and mission-load testing.

Having this tool would give ED a goldmine of telemetry data and would give us players confidence that we are making the most of our hardware. It could even be automated to submit anonymized results, helping ED continuously improve performance.

I think this would be a win for everyone — players get clarity, ED gets better data, and DCS gets even closer to being the ultimate combat flight sim.

Thanks

[Actium]

I wrote a benchmark script for the dedicated server. Exemplary results.

[SharpeXB]

Seems like a simple reasonable idea, all this needs is a standard track in the game.
Some reasons why I think it hasn’t been done:

It could only use free maps and modules and so might not provide useful results for the range of content in the sim. Personally how an Su-25 flying over the Caucasus performs is irrelevant to me. 
It would require a working track feature. This has been broken for many years, not sure if that’s been fixed entirely. 
DCS as a sandbox has the potential for such a wide range of scenarios, a “typical” one is hard to define. Multiplayer is sometimes much more demanding than single-player.
Any preset or standard track has the potential to have bugs or break. 
The game will change and evolve over time so trying to be so exact about how any particular component performs for this one game at this one moment isn’t worthwhile.

Finally all you need to test your own performance is to run your own track. I don’t know that a standard one matters so much. But it would be easy to include with the game. 
 

 

Edited September 21 by SharpeXB

[upyr1]

A bench marking tool would be a great way to provide data for a graphics configuration wizard 

 

[Limaro]

Am 21.9.2025 um 17:28 schrieb SharpeXB:

Seems like a simple reasonable idea, all this needs is a standard track in the game.
Some reasons why I think it hasn’t been done:

It could only use free maps and modules and so might not provide useful results for the range of content in the sim. Personally how an Su-25 flying over the Caucasus performs is irrelevant to me. 
It would require a working track feature. This has been broken for many years, not sure if that’s been fixed entirely. 
DCS as a sandbox has the potential for such a wide range of scenarios, a “typical” one is hard to define. Multiplayer is sometimes much more demanding than single-player.
Any preset or standard track has the potential to have bugs or break. 
The game will change and evolve over time so trying to be so exact about how any particular component performs for this one game at this one moment isn’t worthwhile.

Finally all you need to test your own performance is to run your own track. I don’t know that a standard one matters so much. But it would be easy to include with the game. 
 

 

Thanks for your feedback, I see you already thought of many of the problems it will bring, too. I think, the solution cannot be only one track - there must be at least two.

1. A pure graphic test, maybe on marianas

2. A CPU test, with lots of ground units driving on the road. (can be caucasus)

The graphic test ideally not only runs with one set of settings. But I cannot say what would be the best approach for that. I once programmed a small tool, in which I configured a set of different graphic settings, and the tool automatically set this, started DCS, run a track, recorded the performance and wrote it into a file. I had the mode to run all combinations of the configured settings (which can take veeery long time). It showed, that all settings did not interfere with each other, the drop of performance was per-setting. With this knowledge, I changed it into the way, that all settings are tested, but not all combinations (With one default from which the other change). This gave me the possibility to generate screenhots, to that it became visible, how it looks like, when changing things. My goal was to compare it with the next dcs version. Then, an update, and my tracks did not work any more and I stopped doing this.

The point is there there are infinite ways of testing. I just think DCS should go into the race with something. To find out, how settings, count of units, even some modules, weather, etc. influence GPU and CPU.

[SharpeXB]

2 hours ago, Limaro said:

1. A pure graphic test, maybe on marianas

2. A CPU test, with lots of ground units driving on the road. (can be caucasus)

You probably know this but the in-game counter will tell you if you’re CPU or GPU limited. I’m not sure what the point is in being so elaborate about performance testing is when it varies so greatly.

Edited September 23 by SharpeXB

[Limaro]

vor 19 Stunden schrieb SharpeXB:

You probably know this but the in-game counter will tell you if you’re CPU or GPU limited. I’m not sure what the point is in being so elaborate about performance testing is when it varies so greatly.

My thought was, that It allows to compare the different hardware from other players. It'll show me,

- if my system performs "normally",
- how well different components performs and influence FPS and 1%percentile

[SharpeXB]

1 hour ago, Limaro said:

My thought was, that It allows to compare the different hardware from other players. It'll show me,

- if my system performs "normally",
- how well different components performs and influence FPS and 1%percentile

It would do that. It might not give good feedback for usable gameplay settings though. And there are better ways to test whether your PC and components are functioning normally. I wouldn’t use a game benchmark to diagnose PC trouble. It would still be a very easy thing for ED to include a “Benchmark” track in the game but for the reasons above it’s only so useful. 

Edited September 24 by SharpeXB

[Limaro]

Btw. I no did the hardware upgrate I was mentioning before. And on a mission I've created, having 22 Groups of Ground units with each between 15 and 30 Units, I dropped down to ~57 FPS, which then was CPU limited. Now, with the 9800X3D it was back to 120 (vsync) limit. If I would have known this, I would have upgrated earlier. 

I still believe there are many ways to use the benchmarks. For us, and Ed.

[SharpeXB]

3 hours ago, Limaro said:

Btw. I no did the hardware upgrate I was mentioning before. And on a mission I've created, having 22 Groups of Ground units with each between 15 and 30 Units, I dropped down to ~57 FPS, which then was CPU limited. Now, with the 9800X3D it was back to 120 (vsync) limit. If I would have known this, I would have upgrated earlier. 

I still believe there are many ways to use the benchmarks. For us, and Ed.

Sure but you can use any of your own gameplay or tracks to evaluate this. A standard track seems like an easy logical solution and in many ways it is. Yet it hasn’t been done, probably for all those reasons above. It’s worth remembering that the track feature was meant as a debugging tool. Not other uses like a benchmark. 

Edited 6 hours ago by SharpeXB