Speeding up time in game magically makes those graphic fps issues no problem

[mordeaniischaos 3]

Here is another thing that's affected by FPS / Time acceleration

I wonder if that's why my Numpad - key acts that way... I'll have to toss vsync on and see if it works better.

We should get a benchmark mission setup, and have people post results with their PC specs.

I'm noticing just at random that faster speeds have WORSE performance, lower have better performance but it was far from a scientific test.

Edited May 30, 2014 by MordeaniisChaos

[Vigil Vindex 64]

I can't wait to read what the official BIS reply to this thread will be...

[bravo409 13]

Well let me just throw this out to you guys .... By all means my computer shouldn't be able to run this game but last night I was playing on stratis in open area no buildings and I got 33/42 fps this was without speeding up time of course cause found when you speed up time it doesnt register on the fps counter. Then when I have quite a few buildings fps lowers so I just tend to play in open areas....Anyway I digress here are my specs and again I shouldn't be able to play this really at all. Hp AMD phenom 8400 triple core processor 2.10ghz 3.00Gb Ram 32 bit vista Nvidia geforce gt 630..no over clocking...

Edited May 30, 2014 by bravo409

[MavericK96 0]

The problem is, on a system more than twice as fast as yours, the performance is more or less the same.

[bravo409 13]

Yes that is weird..........keep this thread alive

[St. Jimmy 272]

My fps only drops when I speed up the time.

[Furret 0]

Yes that is weird..........keep this thread alive

I think your fps increase is a side effect of increasing the passage of time, nothing more.

Sorry man.

[bravo409 13]

I think your fps increase is a side effect of increasing the passage of time, nothing more.

Sorry man.

Well if that's the case how come I can run through big towns without lagging if what effects fps is the overall drawing of buildings and objects ...But soon as I turn it off it lags like crazy and fps drops to 10...the odd thing is fps is already reading 10 when I speed up time but I'm running around like I have 100 fps that's why I find this very wrong and made big deal out of it..Because if we have problems with regular speed why can't they reverse roles and change the speed up time to be in the regular engine thus no one would have any fps issues...now read here that some have different things happen with speed up time and others have my problem so something is a miss here and I'm just trying to help with this fps issue to fix this by bringing this odd function to light. Thanks keep this alive believe it or not there is problem here it can be fixed...

[Furret 0]

...

Could you try using a tool called gpu-z and monitor your clock speeds and gpu usage while you run the game?

Test both with and without time acceleration. I suspect time acceleration is forcing your gpu into a higher speed state, providing you more performance.

[vegeta897 13]

Thread seems to have forgotten that the likely cause for this is the reduced AI complexity when time is sped up. Did somebody rule that out, or what? I know for a fact AI driving ability takes a toll.

[bravo409 13]

well were not talking ai, obviously the ai will be effected by this but this is about graphic fps issues not how ai reacts because when you speed time up the ai could act more dumb or have to think faster still unknown that would be a good test to try. If the ai act the same but at a faster rate this will be another can of worms to this problem I have stumbled on too.

[vegeta897 13]

well were not talking ai, obviously the ai will be effected by this but this is about graphic fps issues not how ai reacts

I'm saying the AI is causing the CPU to bottleneck performance, and when you speed up time, the AI complexity is reduced, thus less strain on the CPU, thus higher framerate.

[rory_pamphilon 16]

Somebody try this on a blank map please with no ai. Still get a boost?

[ceeeb 147]

Player alone in Kavala:

setAccTime 10.0 => 44 FPS

setAccTime 4.00 => 45 FPS

setAccTime 1.00 => 45 FPS

setAccTime 0.25 => 46 FPS

setAccTime 0.01 => 46 FPS

[maddogx 13]

OP is talking about speeding up the game with the normal time acceleration/deceleration keys. I'm not sure if setAccTime is handled differently by the engine, but it's worth trying both just to be safe.

I still have a simple AI performance test mission lying around, so I'll give this time acceleration trick a try later and see what happens.

[jumpinghubert 49]

I'm saying the AI is causing the CPU to bottleneck performance, and when you speed up time, the AI complexity is reduced, thus less strain on the CPU, thus higher framerate.

very true. In addition to ceeeb´s results we need a test with ai. If there is a fps gain with speeding up the time then you´r verified.

[jiltedjock 10]

Isn't it obvious that this is caused by AI simulation being downgraded during fast forward? => less CPU work => more FPS.

[bravo409 13]

Isn't it obvious that this is caused by AI simulation being downgraded during fast forward? => less CPU work => more FPS.

this could be part of the issue but I think it is overall drawing 50% and Ai other 50% that kills fps... So knowing this what can we do about that well nothing seeing as the game is still not really playable> I have read about countless people with really good machines having problems with this game so we know it can't be cpu,and we know GPU is has little use so what is sucking all this power and creating fps issues. My money is on the two things just mention overall drawing and AI. Look what happens when you enter a big town fps start lowering specially in Altis ,I bet the computer is having to draw all the graphics with out using much GPU is the issue with fps on Altis and the more Ai you use more power they take with all the commands and waypoints and just general thinking.

[maddogx 13]

Ok, I gave it a try with my basic AI performance test mission. (75v75 = 150 AI fighting in Kavala, player alone on the other side of the island, >15km away, way outside of render range.)

during AI fighting: ~24fps (variable)

after deleting AI: 60fps

Increasing the game speed to 2x or 4x had a detrimental effect on the framerate while the AI were fighting (dropped to 22). On an empty map with no AI I see no effect whatsoever - framerate stays the same. Same story with setAccTime 2 or 4 btw.

On the other hand, slowing the game down to 0.2x using setAccTime, I saw framerates improve to 45-50fps during the large fight. (But this was kind of expected.) Apart from proving that massive AI fights still take an extreme toll on framerates (:mad:), the test was pretty much a negative regarding the "simplified AI during speed-up = better performance" theory.

I can make my repro mission available if needed, but it's really nothing special.

[bravo409 13]

Thats a cool test so what is your conclusion do you think it is AI causing fps drop or also Overall drawing of objects and buildings? I will give this go in a town and see what I come up with. I want to take this time out and say thank you to the people that have given this theory of mine a look and testing it out.

[maddogx 13]

Thats a cool test so what is your conclusion do you think it is AI causing fps drop or also Overall drawing of objects and buildings?

Well, drawing objects is what the renderer is supposed to do when rendering the scene, so it would be kind of nonsensical to say "rendering frames is causing fps drops". ;)

In general, right now I'd say ARMA3's biggest performance problem is with the AI, especially when they are actively pathfinding and fighting. I did a test several months back with different numbers of AI, from which I drew the conclusion that greater numbers of active AI will increasingly bog down the main game loop, causing performance to bottleneck at the CPU on a single thread. Although ARMA3 is multithreaded in the sense that it is able to perform certain tasks in parallel, much of the simulation still seems to be single threaded, and the renderer is directly coupled to the simulation.

As I understand it, the simulation wants to run at a fixed rate (50 ticks per second? based on the dedi server), so if those simulation ticks are very short (<10ms) the main loop can ideally keep pumping out render calls to the GPU more than 100 times per second. In that case, the primary limiting factor that determines your framerate will be GPU power VS video settings (i.e. how fast your GPU can render the scene). However, if the simulation ticks start taking too long to calculate (>20ms), the framerate will start slowing down because simulation ticks will happen back to back, and a new frame will only be drawn at the end of each tick. When simulation ticks start taking longer than 33ms to calculate (lots going on), your framerate will drop under 30fps, then 20fps at 50ms tick time and so on.

This (again, as I understand it) is why you see GPU usage decrease as you increase the number of active AI (and possibly other simulation components) - if your GPU could manage 60fps in the current scene but is only getting 30 render calls per second, it will only be ~50% utilized. Also, since the main loop is running fewer ticks per second, it has less work to farm out to other CPU threads, so CPU usage will decline a little as well.

As other people have suggested, the game's "time acceleration" functionality might cause the simulation to simplify or skip certain tasks, which could shorten tick times and alleviate the single-thread-bottleneck issue.

Please note that I am not a professional games developer (though I have dabbled in it ;)), and I am not a BI employee so I have neither access to nor an actual intimate understanding of the game's source code. Most of this post is based on a mix of semi-educated speculation and snippets of info cobbled together over the past decade. :)

Edited June 2, 2014 by MadDogX

[vegeta897 13]

Thanks for testing that out and sharing your knowledge, MadDog. Now we're getting somewhere, kind of ;)

BIS has been talking about optimizing AI pathfinding as well as doing more parallel processing with regards to AI, right? If so, sounds like they have good targets for optimization here.

[bravo409 13]

Nice work MadDOG....One thing How do you explain the problems that high end gaming machines are having on this Game...They have the same issues and they are much faster and have better gpu/cpu so what is the deal there ?

[maddogx 13]

One thing How do you explain the problems that high end gaming machines are having on this Game...They have the same issues and they are much faster and have better gpu/cpu so what is the deal there ?

Once you hit the single thread bottleneck, only the raw processing power of a single core on your CPU matters. At that point it makes little difference whether you have a single GTX650 or multiple Titans in SLI.

Most modern CPUs do 4-4.5GHz (similar to mine) and perform fairly similarly in benchmarks, so they won't see much difference in performance. Quad-, hex- or octocore won't make a shred of a difference either.

[bravo409 13]

keep this alive...........Has anyone done any more test if so post it please.