Aircraft performance comparison in 2D and VR

Hey guys,

the recent discussion about the level of detail of some modules in DCS got me thinking, and since I have all the modules currently available, I decided to make a little performance comparison in DCS 1.5.6. This thread is not an absolute performance benchmark, but rather to show relative performance of various modules in a quick and reproducible way.

1.5.6. The Setup:

Each aircraft was parked at Sochi runway ready to takeoff at noon in clear sky weather. You can download the mission files at the bottom of this post. After the mission loaded and FPS stabilized, I measured an average FPS for 10 seconds using fraps. I used default view from cockpit, but adjusted the FOV in 2D so that all aircraft were measured on roughly equal terms. I didn't make any changes to displays or aircraft systems. So i measured FPS in a view like this:

The results I got were pretty consistent, with very little FPS fluctuations. Below are DCS settings that I used for the test both in 2D and VR. In 2D the resolution was 2560x1440. In VR the pixel density was set to 1.7, and ASW was obviously disabled.

Hardware specs (those that matter):

i7 6700k @4,5 GHz

32 GB RAM @ 3000MHz

GTX 1080 (mildly overclocked)

SSD dedicated for DCS

Windows 10

And here are the results:

A note about VR performance: the maximum achievable FPS in the Rift is 90 FPS, so any aircraft with a high average (85+) is hitting this limit. That doesn't mean they're equal, quite often you need only to move your head for the FPS to drop, and some of those aircraft still give you less performance hit in such situations.

Another note: turning on the aircraft systems can have significant effect on FPS. For example the radar in MiG-21 or Viggen, TGP in A-10C or ABRIS map view in Ka-50, all eat up between 5 and 10 FPS on my system.

Conclusions:

* As expected, the best looking or the most complex modules are heavier on FPS, but there are some exceptions.

* There's something significantly wrong with the Spitfire, its performance is much worse than other modules in both VR and 2D. Yeah, I know, it's beta (7 months and counting).

* C-101 also seems to have some performance issues.

* While the Hawk is one of the best performers in 2D, its one of the worst in VR. I've no idea how VEAO managed to achieve that, they probably don't know either.

* WWII planes in general have worse FPS than more complex aircraft. Not sure if it's because of the propeller or something else. It is significant in VR.

* Early jets from Korean War offer the best FPS, even better than FC3 level aircraft.

* Honorary mention to Razbam M-2000C as it offers almost FC3 level performance in a complex ASM/EFM aircraft.

* Gazelle also fares quite well on my PC, despite some complaints about performance on the Polychop forum. But of all Helicopters, UH-1 gives the best FPS.

1.5.7

https://forums.eagle.ru/showpost.php?p=3273609&postcount=19

2.1.1 The Setup:

Aircraft in take off configuration at Nellis AFB:

Configuration as shown on screenshot. It's similar to the one in 1.5.6 except for MSAA off, as enabling it cuts FPS amost in half

The results:

The one undeniable fact is that 2.1.1. is an unfinished alpha. The models are in various state of transition to the new PBR textures. Some have missing gunsights or propellers, which may explain that they give better FPS than in 1.5.6 (P-51, Bf-109). Some have crazy speculars or other graphical glitches. Don't put too much attention to those results, as those aircraft that show worse performance may actually be more up-to-date, with at least partially implemented PBR materials and textures.

Another obvious conclusion is that VR performance in 2.1 is terrible. While in 2D with MSAA off I get comparable results to 1.5.6 with MSAA 4x, I get roughly half of the FPS in VR compared to 1.5.6 using the same setup!

A note about VR performance: while the older ATW smoothing algorithm in Oculus software allows for different FPS than 90 and 45, it still tries to maintain 45 FPS if the game generates roughly that amount of frames per second. That's why you see 45 next to most aircraft from the mid of the pack. But you can still see which aircraft perform better than average, and which are below the threshold.

benchmarkSochi.miz

benchmarkNevada.miz

Thanks for this perfromance review, could you try the same test in 2.1 with Nevada? +1

Very nice comparison and conclusions some1, thanks. Interesting results and good for someone like me who doesn't have many modules but is 'looking'. (Who am I kidding..like I am going to fly anything else once the Tomcat is released!)

Would be interested in results in NTTR as well, as that is the only terrain I fly on. I imagine there would be quite a difference between say Nellis and Groom thou.

On another note - I saw the title and first thought was A/C performance is up 300% in VR....on the 'fun performance' scale.:joystick:

Sure, I can test NTTR, but I'll wait at least for the first update to 2.1. Though I don't expect much difference between aircraft, just the overall FPS to be lower.

Thanks! Great comparison :) When the first dust has settled, I would also be interested in something like this in 2.1.

Hi,

useful information :thumbup:, thank you for sharing the measured values.

Will you be (re)testing and editing your original post now that 2.1.1. is out? Im curious to see whether there are any performance differences between these two versions? Thanks!

Sure, I can test NTTR, but I'll wait at least for the first update to 2.1. Though I don't expect much difference between aircraft, just the overall FPS to be lower.

Cheers. Looking at the Hawk in 2D you would also expect it to do fairly well in VR. The thing with DCS is it's like a Christmas cracker...you are just not sure what you are gonna get lol. :xmas:

Will you be (re)testing and editing your original post now that 2.1.1. is out? Im curious to see whether there are any performance differences between these two versions? Thanks!

Since his test was done on a different map and engine, there is little reason to compare them. The Open Alpha is a completely different animal than v1.5.x.

As he stated, the difference between aircraft will, likely, be about the same but overall FPS will probably be lower. An update to the Open Alpha is not likely to change that.

Since his test was done on a different map and engine, there is little reason to compare them. The Open Alpha is a completely different animal than v1.5.x.

Not quite seeing the logic here. The point would be a comparison of the performance of the aircraft relative to each other for the new engine (and map).

As he stated, the difference between aircraft will, likely, be about the same but overall FPS will probably be lower. An update to the Open Alpha is not likely to change that.

Similar argument as above. It would be interesting to see, if any aircraft is more affected by the new engine/map in terms of FPS compared to others. Saying that there will be the same change/decrease in FPS for all of them is a hypothesis, which needs to be proven. Maybe we see something interesting - maybe not. But as you said yourself, 1.5 is a "completely different animal". So why should we assume that things are the same in 2.1?

(For example I had the feeling(!) that the radar of the Viggen is heavier on my FPS in 2.1 than in 1.5. But I never thoroughly tested it. Maybe we could shed some light on questions like this here.)

Well, I did the test in 2.1.1 because I was interested myself in what the changes are, but after seeing the results, and how the game looks, I think I won't waste much time in the nearest future on this piece of unfinished software. Just stick to 1.5.6 after the issues get sorted out. The relative performance of various aircraft in DCS 2.1.1 is indeed quite different than in 1.5, but I guess it's mostly because at least half of them hasn't been updated to PBR and they show various graphics glitches and artifacts. Missing glass, gunsights and propellers are the most obvious ones.

Also, VR performance in 2.1.1 is very bad, even though in 2D I get quite good results on par with 1.5. At least in the middle of the desert.

Anyway, first post updated.

(For example I had the feeling(!) that the radar of the Viggen is heavier on my FPS in 2.1 than in 1.5. But I never thoroughly tested it. Maybe we could shed some light on questions like this here.)

In both of my tests, Radar in the Viggen was off. But the Viggen is near the bottom of the list either way.

Thanks some1! Surely, 2.1.x is unfinished/alpha. But your results might give the developers (and us users) some hints where there is potential for improvement. The developers might knew already but I did not ;)

 

Anyway, first post updated.

Thanks :thumbup:

Anyway, first post updated.

Cheers for that. I sort of want to just go back to 1.5.6 for VR, but can't be bothered getting all sorted there for them to bin it and move to 2.5 anyway, and having two completely separate instances of DCS etc.

Will soldier on with 2.1 I suppose...:(

Thanks, much appreciated!

I've been asked about it, so I added mission files I used for the tests.

For 2D, every aircraft has slightly different FoV, so you have to match zoom level manually after the mission loads. It's easiest to do that using some buildings near the edge of the screen asa reference.

Very good thread/test , thank you.

I have repeated the test in DCS 1.5.7.10872 Update 4. I tested only in 2D, as it is much easier for me to do than VR. Looks like I gained a few FPS in most aircraft, either thanks to some DCS improvements or other factors in the system. One notable change is much better FPS in all warbirds (15% increase, around 20 fps). C-101 also gained quite a lot. I suspect this is because of some changes in transparency rendering between these two versions.

Spitfire still lags behind the other warbirds but at least it isn't the worst performing aircraft any more. That title has been taken by Heatblur AJS-37. Fortunately, the differences between aircraft aren't that big.

Quick comment, your 2.x test had visibility range at ultra vs high for 1.5. That does make a massive difference as visibility range is the primary hit on CPU usage in the game. Radar, flight models, AI, shadows, it's all trivial compared to the visibility and tree draw distance as far as the CPU is concerned. This is true in both branches of DCS.

As for my own experience, there is no noticeable difference between any of the aircraft in terms of VR performance. The only exceptions being the MiG-21 and Viggen with their radar on, that causes it to drop a few fps in both VR and 2d.

Quick comment, your 2.x test had visibility range at ultra vs high for 1.5. That does make a massive difference as visibility range is the primary hit on CPU usage in the game. Radar, flight models, AI, shadows, it's all trivial compared to the visibility and tree draw distance as far as the CPU is concerned. This is true in both branches of DCS.

It is also on a completely different map, in a completely different place. And with MSAA off. The main purpose of this test is to compare airplanes, not DCS branches. When the big merge happens, if I'll still be around, I'll try to make a more direct comparison between old and new DCS.

Regarding VR performance, you may want to read what I wrote about measuring FPS in VR, especially if you use ASW.