About Performance/Single Thread

[Dankx]

Hi , i have been playing PR for quite a while now and i know the game doesnt run more than 1 or 2 cores on your CPU, so i bought new hardware having this in mind , i got a G4560 3.5ghz CPU (2.1 single thread rating on the CPU pass mark page ) which is considered pretty decent for single thread performance. Problem is on full servers 50vs50 when im looking into the side of the map where the action happens ( where most players are gathered ) i get 30 fps , if i look back into my main base it jumps back to 100fps, yh literally 100. My question is what can i tweak to fix this ? I tried toggling on the " No Lods " option on launcher settings , i tried Pr.exe high priority /cpu affinity in the task manager , i also tried unparking the CPU and nothing seems to change this. If my single thread performance is 2.1 in the pass mark score and i only get 30 fps when shit goes down do i need a 4.0 single thread score to get 60 fps lol ? Does anyone even run this game at 60+fps ? I feel something is terribly wrong here since going from 100fps too 30 because i look in the direction players are is ridiculous. Also in main base when everyone is gathering to start the game my frames stay rock solid 100fps , only when its enemies mixed with allies in the battlefield ahead of me. Help please , i think this has been an issue for alot of people and we really need some clear answers since i cant find any on the forums.

PC specs
cpu: G4560 3.5 ghz
gpu: 750 ti 2gb ( shouldnt matter anyway )
ram: 8 gb

[DogACTUAL]

You are not alone my friend.

[AlonTavor]

If you ever figure it out let us know.
You should try disabling hyperthreading on that dual core.

[qs-racer]

"No lods" option active means that every buildind/texture will be display at maximum resolution what ever the distance, so avoid it with a small configuration (even with my 1050ti i cannot use no lods)

I had a similar configuration on my previous laptop (gt650m, i7-3630QM @3,4Ghz) and it works fine, even if FPS drop to 30 on some maps, it is playable (Kamishiyah with the burning oil pump kill the FPS).

I recommend to put anti aliasing option to 2 or 4 maximum, then other settings to high or medium depends how your system can handle it.
Be carfeul about the cooling of your system (clean it if dusty) to avoid overheat self restriction on frequency), especially that you seems not have issue at the begining of the game.

[Danesh_italiano]

Not much you can do. With a 5Ghz last gen CPU, gave me 35~40 fps in some instances on op. marlin.

I did a "BenchMark" with best CPUs/RAM speed/timming on a map that usually give you low FPS:



Disabling hyper threading helps but since you have a DUAL core, you gonna struggle to play PR with others software opened at same time (TS/Discord for example).

[Dankx]

Oh my even with those machines u get that result , insane . How many frames do u think disabling hyper threading will give in those scenarios where everyone is bunched up?

[DogACTUAL]

Yes.

The 'no lods/Pr.exe high priority/cpu affinity/unpark cores' advice from devs and support staff is completely useless and was just them throwing shit onto a wall to see what might stick.
I read all those 'low FPS' threads myself and tried all their advice when i was asking the same questions you are asking now.

The truth is just the game is terribly optimized, so far that beyond a certain point hardware-wise you won't even get better results even if you upgrade your hardware. For example my mid-range build gets virtually the same FPS in these urban maps with many players as another guys extreme high-end build i talked to.
Which suggests that the game itself has become the ultimate bottleneck for FPS. Like the game engine has a hard cap that doesn't allow for further performance despite having hardware capable of doing so.

[PatrickLA_CA]

The only thing I noticed to help while I was playing on a laptop was cooling. Without a cooler, the game became rather unpleasant quickly. On my desktop I don't really notice any FPS issues except in some rare cases where there's a lot going on, but hey, it is a 15 year old engine that has gone way past what it was designed for. 30 FPS is not even bad, I mean, imagine playing on a console.

[SemlerPDX]

I don't think I saw FPS drop below 100 on an Intel 4690K with a GTX 970 on a 1080p 144Hz display.

Pinned at 100FPS now with a system designed for very modern games (3900X w/ RTX 2070 Super).

Not unable to be achieved, but throwing money at the problem is not always the most fiscally responsible method for a number of reasons. Either way, a purpose built gaming rig will demolish older games like PR.

[PeppeJ]

Like the game engine has a hard cap that doesn't allow for further performance despite having hardware capable of doing so.

Well in 2004 there were barely any multi core CPUs out, so the only way to make a game engine perform well was to make sure it performed well on a single core. Thus throwing more cores doesn't solve the issue. Also we've kinda reached max speed on a single core today the solution to stronger CPUs is to just throw more cores at the problem, which is why a lot of modern games scale well.

For instance on passmark a 2005 processor (Amd Athlon 64 x2 3800+) has 650 score and the strongest modern one has about 3000. So today we're only ~4.6x stronger on a single core. So you could make the assumption that the game should run about 4.6x faster. However PR compared to BF2 has a lot more detail, higher player count, more gameplay logic, much higher view distance and a bunch more stuff. So in the end we're not really gaining many frames compared to a processor running BF2 back then.

Tl;dr we don't really have the right hardware to push the game to higher frames.

[SemlerPDX]

Well in 2004 there were barely any multi core CPUs out, so the only way to make a game engine perform well was to make sure it performed well on a single core. Thus throwing more cores doesn't solve the issue. Also we've kinda reached max speed on a single core today the solution to stronger CPUs is to just throw more cores at the problem, which is why a lot of modern games scale well.

For instance on passmark a 2005 processor (Amd Athlon 64 x2 3800+) has 650 score and the strongest modern one has about 3000. So today we're only ~4.6x stronger on a single core. So you could make the assumption that the game should run about 4.6x faster. However PR compared to BF2 has a lot more detail, higher player count, more gameplay logic, much higher view distance and a bunch more stuff. So in the end we're not really gaining many frames compared to a processor running BF2 back then.

Tl;dr we don't really have the right hardware to push the game to higher frames.

Yea, no it's hard capped at 100, don't matter the hardware, when the game engine limits it, it's not possible to get past.

You can use current hardware to pin it at 100FPS at max GFX settings regardless of number of players around you or the amount of action, period. But again, for PR, building a modern gaming PC just for PR would be extreme overkill...

[AlonTavor]

Yea, no it's hard capped at 100, don't matter the hardware, when the game engine limits it, it's not possible to get past.

You can use current hardware to pin it at 100FPS at max GFX settings regardless of number of players around you or the amount of action, period. But again, for PR, building a modern gaming PC just for PR would be extreme overkill...

game.lockfps 0
I get 1200fps when high in the sky. Its obviously something with the rendering.

[Outlawz7]

From a mapper's POV, it depends on the map.

danesh's benchmark video is on Operation Soul Rebel which is one of the worse performing maps in the game.

[Danesh_italiano]

Funny thing is.... FPS goes doum when you look at where the majority of people are, right? (You can test yourself. Play online on op.marlin a find a place where you get 40Fps. Disconnect and create local server without bots and go to the same place. Your fps will probably be 100).

But then, found this video that has high quality textures and ?2?5?6? 255 freakin bots:



So.... something on the soldier/kit/weapon models that is breaking the game?! Maybe someone (i cant, no desktup at dis moment) could replace all PR soldier/kit/weapon models with BF2 models and see wut happen?

[PatrickLA_CA]

Well, view distance seems to be the same as in BF2, the map is the same as in BF2, it doesn't have plenty of detail and objects. TBH I get less FPS in Forgotten Hope 2 than in PR on some of their maps with a stunning amount of detail.

[DogACTUAL]

He is right. Devs should certainly look into the connection the player models or players in one place have in this.
Fallujah also runs at much higher FPS looking at the city centre if there are no players on it.

It's also weird how other maps won't have the same problem with a lot of concentrated players though.

[ALADE3N]

He is right. Devs should certainly look into the connection the player models or players in one place have in this.
Fallujah also runs at much higher FPS looking at the city centre if there are no players on it.

It's also weird how other maps won't have the same problem with a lot of concentrated players though.

its the bunch of statics that is causing fallujah to lag so much

[Rhino]

He is right. Devs should certainly look into the connection the player models or players in one place have in this.
Fallujah also runs at much higher FPS looking at the city centre if there are no players on it.

It's also weird how other maps won't have the same problem with a lot of concentrated players though.

Oh ye, that should take what, 10mins to do and fix? Sure no problem, why didn't you ask before?

I'm sorry DogACTUAL, but you act like we do nothing nor have any interest in fixing this game, which simply isn't true, but even to make the tiniest improvement takes a huge amount of work and is normally totally unnoticeable. Even when I fixed up the Lynx with some vast optimizations and improvements, that one helicopter took a huge amount of work to rework all its faults: https://www.realitymod.com/forum/showth ... p?t=148783

PR players should all be very well acquainted with the Lynx AH.7 (Army Helicopter Mark 7) since it has been ingame as the British Army's Light Utility Helicopter since v0.85 (2009), but since then I have optimized and improved somewhat and they were actually quietly implemented into v1.5 last year, just other than the changelog it wasn't mentioned as the other versions of the Lynx were not finished enough for me to make this post.
Most of these changes are invisible or hard to notice but most of the changes have either decreased the overall load on the client's system to draw and load, while some have just made the vehicle look better with little or no extra load on the client's system. One of the most notable changes is the main texture resolution which have been increased from 1024x1024 to 2048x2048 (four times the texture size/detail) while only slightly increasing the overall memory usage from before with moving all the transparent parts off the main sheet and onto their own unique small sheet with only transparent parts on it.
New optimized & higher quality "Rotor Blur" textures were also made since the old rotor blurs used the main rotor's rotor blur for the tail rotor, missing the tail rotor's iconic red and white stripes from the blur.
There is a whole range of other small improvements and optimizations made but most of them I can't remember and are too small to be worth noting but the bottom line is that the Lynx does not only have a smaller impact on your hardware but also looks better than it did before v1.5, and all these optimizations have of course been passed on to the other mew versions below!

As for why "lots of players in one location causes FPS drops" (at least on some systems), that's most likely, for most systems, because the client has so many draw calls to process and it's worse on maps like Fallujah that have lots of lots of draw calls already from all the individual statics around the client is having to process on top of all the players and the weapons they are carrying etc.

Fixing this thou isn't like changing one line of code in PR I'm sorry to say, it requires reworking and optimizing all our models.

Back in around 2011/2012, I identified our handheld weapons, mainly their 3rd person models being a huge issue for performance in how many draw calls they have, with every attachment having their own texture, sometimes multiple textures and multiple materials, each classing as a new draw call. This is ok for a 1st person model as you only ever see one of them at any one time and only have to process that for each weapon. But for 3rd person models, anyone in view means that if their weapon has say 10 draw calls, then if 30 players are in your view distance with no other way of them being culled out, you have 300 draw calls to process just from the handheld weapons they are all holding alone.

As such, I tried to introduce a method of creating separate 3rd and 1st person models and textures, with them having as few draw calls as possible, as well as being much lower poly and smaller textures etc, which you can see the unfinished tutorial on it here: https://www.realitymod.com/forum/showth ... p?t=111845

1st person, 20,135 tris: https://media.realitymod.com/tutorials/ ... tut_02.jpg


3rd peson LOD0, 5,858 tris: https://media.realitymod.com/tutorials/ ... ut_222.jpg


You can see there the 3rd person models had almost 1/4 the tris(/detail) as the 1st person models and everything grouped up onto one texture atlas. There were some big things I would do very differently if I was going to do it again, like separating the alpha/transparent textures out onto their own texture sheet to save on wasted texture memory and because they require their own material/draw call that's separate from the others (something I did for the Lynx series when optimizing it, before it had all it's alpha textures in the main sheet, wasting loads of texture memory), and do common attachments differently as well but my knowledge/skill of optimization and modelling has come quite some way in 8 years so take that tutorial/method with a big pinch of salt.

The biggest problem with this whole concept, however, was mainly the huge amount of work it took to do. It took me a long time to do just the SA80 series. Then on top of that, you had the problems of not many people understanding it or able to do it, and its inflexibility which also lead to Chuc accidentally reverting all of this when he updated the SA80 series models/animations and would have been another huge amount of work to fix which I never got round to.

So the method was pretty much abandoned for the most part, some weapons do have optimized 3rd person models where possible but most still have the same models and textures for their 3rd person weapons as they do their 1st since its simple, easy and looks by far the best.


If DogACTUAL, you really care about this as much as you say, I could update the tutorial/method and you could go about updating all our weapons with proper 3rd person models but we are realistically talking a few years of work to do that and that would be even for just someone like me doing it who already has the skills but I will happily point you in the right direction if you want to do all the work that's required

For me and most of the other Devs/Cons thou, we simply don't have the time or the motivation to go over all the old work other people have made, optimizing them with only the tiniest bit of performance improvement for everything you do. For now, me and the other devs are just focusing on making anything new we make as optimized as possible (The L4A4 Bren is just one of many new weapons with proper separate 1st and 3rd person models/textures), or optimizing any old models we are updating for any other reason, like what I did with the Lynx series, I optimized it fully before making the new versions, even thou the original Lynx AH.7 won't be seen very much in PR with the new Lynx AH.9A taking its place on most maps.

Cheers!

[Danesh_italiano]

As for why "lots of players in one location causes FPS drops" (at least on some systems), that's most likely, for most systems, because the client has so many draw calls to process and it's worse on maps like Fallujah that have lots of lots of draw calls already from all the individual statics around the client is having to process on top of all the players and the weapons they are carrying etc.

So, just confirming, weapons that has same 3rd and 1st models is the mainly reason for FPS drop when looking at players direction?

Doing it manually is basically impossible unless someone get paid to doet

[Outlawz7]

So, just confirming, weapons that has same 3rd and 1st models is the mainly reason for FPS drop when looking at players direction?

No, it's not that straightforward. Like Rhino said, some weapons are better optimized than others.