Low CPU utilization & Low FPS

[Vigil Vindex 64]

Snarky comments won't get the game fixed.

I apologise if it seemed snarky, it's just frustrating how this issue seems to get danced around all the time. I am genuinely concerned though that it seems BIS don't do any testing with large numbers of AI, if they did, we wouldn't have to submit repro missions for such basic problems.

[PuFu 4591]

we already have the mission (maddox sent me that) , it will be analyzed ... ;)

although i commend maddogx for taking the time to reformulate the same, i doubt it this mission is something BI couldn't have sorted by themselves based on this single thread.

The mission is nothing special; it literally took <5 minutes to make and test. Still nice to know someone is looking at it. :D

Of course 150 AI is quite a lot, but the framerate hit is already noticable (though not as bad) with just 50.

I know. still, BI requires and requests repro steps even for the obvious things.

I find it hard to believe that we are 200 pages into this thread and we are still acting like this is some sort of mysterious intermittent issue that requires a repro mission submitted. For real? I guess BIS never try to play with 100+ AI because they know it cannot be done without crippling fps.

No one (or at least not me), is acting as if this is something that is not well known. It is just that BI blames it on anything but their own engine. Still, no matter how bit the frustration is, not gonna help by making it visible here, quite the contrary

[Predator.v2 10]

Upgrading to an i7 4770k solved my problems. :)

It was hell to play A3 MP on a oced Q9550..

[Speederr 1]

So we have to pay a new proc to hope play ARMA 3?

LOL

[MItchMayne 10]

I was considering to upgrade my GPU from GTX 570 to GTX 770 but I changed my mind. Since the CPU bottlenecks the GPU it make nosense to upgrade my card.

As soon as the AI is involved I would get the same FPS as with my old GPU.

Hardware:

i5-2500K @4.3GHz, GTX 570 , Samsung SSD EVO 250GB, Ram 8 GB

Yea i'm rocking i7-2600K @4.8GHz, 2x GTX 580's SLI, intel ssd 250GB, 16GB corsair vengeance ram....I have been having a lot of issues. Having a monster PC does not seem to matter because of bad utilization or coding within it.

[antoineflemming 14]

It is just that BI blames it on anything but their own engine. Still, no matter how bit the frustration is, not gonna help by making it visible here, quite the contrary

This is my biggest gripe with them (why months ago I along with others got into an argument about BIS acknowledging the problem). Pretty much want BIS to acknowledge that there's a problem (not a limitation, not nuance) with their engine. Because there is. And it's not particle effects or view distance or clutter. Because changing those things after a certain point has negligible impact on performance.

Think it has something to do with AI and maybe even ballistics calculation. Because, I made a simple mission of BLUFOR and OPFOR groups fighting on the Northeast corner of Stratis. It was just them shooting at each other really. I get about 40-50 FPS just standing around generally with settings on standard/low (most things on standard, all particle effects and object detail set on low), about 35-40 when this mission starts, but that doesn't matter. The change in FPS is what matters. When the shooting starts, the FPS drops down to like 10-15, occasionally 8 or 9, when the firefight starts and you have a large volume of fire being exchanged, then it goes back up anywhere between 17-25FPS. So I started tweaking settings. Turned up and down particle effects, negligible change. Turned up and down view distance and textures and object detail. Sure, over standard it starts to impact pretty significantly. I then started changing the sampling rate, dropping it down until it was a blurry screen. And that probably gave me 3-5FPS. Then I changed the resolution, and that seemed to be the largest way to gain FPS (about a 15-20FPS gain). I know none of this is backed up by video, and it's subjective for others. But what I noticed was that the dip was the same for pretty much all standard settings (with some high) and below, regardless of any of these tweaks, even the resolution. It'd still drop down to 20FPS or below. Not really sure why though, as I don't really understand how the engine is calculating ballistics or handling AI, but it seems to be inefficient in my opinion. Not sure if there's a way for BIS to make it more efficient, and from what has been said in the past, AI coding is much more efficient than in previous titles. But just from observation, those two factors seem to have the most effect on my performance. I've even noticed slight FPS drops when shooting with no AI around. Not sure if anyone else has observed this or come to the same conclusion.

And I'm sure that some of what has been said might be incorrect or the wrong assumptions, but just wanting to see if anyone else sees the same effect, where FPS also drops when shooting.

[wtornado 0]

This is my biggest gripe with them (why months ago I along with others got into an argument about BIS acknowledging the problem). Pretty much want BIS to acknowledge that there's a problem (not a limitation, not nuance) with their engine. Because there is. And it's not particle effects or view distance or clutter. Because changing those things after a certain point has negligible impact on performance.

Think it has something to do with AI and maybe even ballistics calculation. Because, I made a simple mission of BLUFOR and OPFOR groups fighting on the Northeast corner of Stratis. It was just them shooting at each other really. I get about 40-50 FPS just standing around generally with settings on standard/low (most things on standard, all particle effects and object detail set on low), about 35-40 when this mission starts, but that doesn't matter. The change in FPS is what matters. When the shooting starts, the FPS drops down to like 10-15, occasionally 8 or 9, when the firefight starts and you have a large volume of fire being exchanged, then it goes back up anywhere between 17-25FPS. So I started tweaking settings. Turned up and down particle effects, negligible change. Turned up and down view distance and textures and object detail. Sure, over standard it starts to impact pretty significantly. I then started changing the sampling rate, dropping it down until it was a blurry screen. And that probably gave me 3-5FPS. Then I changed the resolution, and that seemed to be the largest way to gain FPS (about a 15-20FPS gain). I know none of this is backed up by video, and it's subjective for others. But what I noticed was that the dip was the same for pretty much all standard settings (with some high) and below, regardless of any of these tweaks, even the resolution. It'd still drop down to 20FPS or below. Not really sure why though, as I don't really understand how the engine is calculating ballistics or handling AI, but it seems to be inefficient in my opinion. Not sure if there's a way for BIS to make it more efficient, and from what has been said in the past, AI coding is much more efficient than in previous titles. But just from observation, those two factors seem to have the most effect on my performance. I've even noticed slight FPS drops when shooting with no AI around. Not sure if anyone else has observed this or come to the same conclusion.

And I'm sure that some of what has been said might be incorrect or the wrong assumptions, but just wanting to see if anyone else sees the same effect, where FPS also drops when shooting.

I have Win 7-64 bit with 12 Gb of ram and a I5 quad at 2.67 GHz.My video card is a GTX 660OC-2GB card.

What you just put up sums it up perfectly for me.

I am worried now that i won't be able to play the offline campaigns and at least have some fun with this game.

[Covert-OPs 10]

Upgrading to an i7 4770k solved my problems. :)

It was hell to play A3 MP on a oced Q9550..

I have the same processor and I still experience issues. As I've stated previously I can run the game at 80FPS but when we get 16 guys on and in a village/town the frames drop to 35. I'd be willing to match up specs and even get on our server together to see what our performance comparisons are.

[bez 10]

Why won't they implement a network bubble like Dayz standalone?

it seems like a smart idea, that at least I think will have a nice performance boost.

-EDIT-

In general I have a feeling BI always start things from the beginning (in a bad way, not things like a brand new engine)

But it seems they don't use their past experience, fixing stuff again, that they already fixed in the past.

Mods and addons for example, why are they not utilizing addons that already fix thing? legally it can be solved easily. so please, don't say there is a legal a problem.

Also, for example they are working closely with the Dayz standalone devs, but it seems they don't take any advantage of what the Dayz team is doing and completely ignore it, why?

This reminds me of the work ethics (or lack of) of a little kid who is not organized enough to save his work and use it in time of need,

instead doing every thing from scratch over and over. I just don't get it.

Edited September 18, 2013 by bez

[antoineflemming 14]

I have the same processor and I still experience issues. As I've stated previously I can run the game at 80FPS but when we get 16 guys on and in a village/town the frames drop to 35. I'd be willing to match up specs and even get on our server together to see what our performance comparisons are.

Once again, don't use multiplayer as a basis for judging performance gain/loss. MP performance depends heavily on the server. Test out the editor, and see how your performance is.

[jumpinghubert 48]

performance gain with latest dev branch, altis-benchmark from 44 to 48fps and stuttering when flying low in a jet over altis is completely gone :)

[bvrettski 10]

Once again, don't use multiplayer as a basis for judging performance gain/loss. MP performance depends heavily on the server. Test out the editor, and see how your performance is.

Thats all fine and dandy but I'm not playing in the editor and I'm not playing single player missions. For me Arma is a multiplayer game and needs to be optimized for online play. Yes I understand that using the editor will show you some bottlnecks in the system but its also being compounded by BI's multiplayer architecture.

In the laundry list of things to do when you developing a great game I'm pretty sure making sure your multiplayer game plays well should rank near the top or the list. Yes the dedicated fans will stick with the game for a variety of reasons but if BI hoped to pick up new players I'm afraid that goal is quickly slipping away because they released a half finished and poorly optimized game..maybe so they could beat BF4 out the door.

In our gaming community we had a a few people pick it up by now they aren't recommending others purchase it until the major issues are fixed. What do think the chances are of any of them saying in four months..ohh by the way Arma 3 is fixed...runs great..you should get it. Pretty slim...

[t4nk83 10]

performance gain with latest dev branch, altis-benchmark from 44 to 48fps and stuttering when flying low in a jet over altis is completely gone :)

Any idea when this Dev branch will be released?

[jumpinghubert 48]

@t4nk83

the current branch gives me this performance gain: v. 103.110181

[antoineflemming 14]

Thats all fine and dandy but I'm not playing in the editor and I'm not playing single player missions. For me Arma is a multiplayer game and needs to be optimized for online play. Yes I understand that using the editor will show you some bottlnecks in the system but its also being compounded by BI's multiplayer architecture.

In the laundry list of things to do when you developing a great game I'm pretty sure making sure your multiplayer game plays well should rank near the top or the list. Yes the dedicated fans will stick with the game for a variety of reasons but if BI hoped to pick up new players I'm afraid that goal is quickly slipping away because they released a half finished and poorly optimized game..maybe so they could beat BF4 out the door.

In our gaming community we had a a few people pick it up by now they aren't recommending others purchase it until the major issues are fixed. What do think the chances are of any of them saying in four months..ohh by the way Arma 3 is fixed...runs great..you should get it. Pretty slim...

But it depends on the server, so you can't say that the game runs better by going to the same or even different server. The server runs better, but not the game itself. The game hasn't really improved much. That said, Arma is not "a multiplayer game". It's a game designed for single player that features a multiplayer, imo. I'm not saying only play single player, or to play single player at all. I'm just saying that you cannot use any perceivable performance gains in multiplayer to assess whether or not the game performance-wise has improved.

[Covert-OPs 10]

Once again, don't use multiplayer as a basis for judging performance gain/loss. MP performance depends heavily on the server. Test out the editor, and see how your performance is.

Currently we're utilizing Gameservers.com. Do you have any specific recommendations about quality hosting companies for arma servers?

[windies 11]

Currently we're utilizing Gameservers.com. Do you have any specific recommendations about quality hosting companies for arma servers?

It's not a question of good server company vs bad server company. It's that trying to base game performance off multi player is a very bad thing to do because of all the changing factor's involved with how the engine differs between single player and multi player. That's not to say that multi player performance should be terrible, like it is now consequently, but that it's a bad way to measure whether performance improves or degrades based on engine changes or graphics settings because there are a lot of placebo effects to be seen in multi player, like server fps changing or people leaving the server or clean up routines running in a mission etc...

[sawe 10]

If the problem is that rendering takes the biggest slice in CPU time, let us lower those settings.

Proposal: add couple more graphical setting presets very low and ultra low to the game ?

In that way we could get the multiplayer part to playable framerates, and would buy you time to optimize the game.

Those presets could benefit together the poorest of the CPUs and GPUs.

Rendering is GPU based. CPU is for AI calculation, scripts, etc. The problem isn't that there aren't enough graphical settings, it's that changing them between Low and Ultra often has little to no effect on performance, because there are deeper issues with CPU utilization.

I'm sorry that I didn't explain my proposal more precisely.

I base my proposal on article in the developers blog: http://www.bistudio.com/english/company/developers-blog/91-real-virtuality-going-multicore

Article clearly defines that "Rendering is still the part of the game which takes most of the CPU time in ArmA, more than simulation or AI." And by that the developer means that the renderin part before the data is passed on to GPU for rendering as people know it.

My proposal was to ease the burden of the CPU rendering side, and if the overall burden would come down by simplified models or dropped ones, it would also help the weaker GPU's.

[Covert-OPs 10]

It's not a question of good server company vs bad server company. It's that trying to base game performance off multi player is a very bad thing to do because of all the changing factor's involved with how the engine differs between single player and multi player. That's not to say that multi player performance should be terrible, like it is now consequently, but that it's a bad way to measure whether performance improves or degrades based on engine changes or graphics settings because there are a lot of placebo effects to be seen in multi player, like server fps changing or people leaving the server or clean up routines running in a mission etc...

Understood. But the question remains, is there a better performing Arma hosting company that anyone would recommend? I've already spent a ton of money on a new rig to try and make this game run flawlessly, might as well go for the gusto.

[hapek 10]

Seems like CPU & GPU scaling is pretty good based on article from Techspot com.

In my case, usage of my HD7870 paired with ancient e8400@4.4ghz is often maxed. High specs requirement by arma is justified.

However it might be hard for average user to run modern cpu at 4 ghz+. Seems like i5 is the best choice performing just like i7 but its the same in every other game so nothing new here.

[SunshineLOV1N 10]

Bohemia just announced a new patch is being released that will completely fix the FPS performance problem. Should see gains of 100-200% FPS increase with the patch, as long as your hardware meets the recommended requirements.

This is what I hope to be reading sometime in the near future. I paid $60 for this game and I want to experience what I know its capable of.

Sorry to get everyone excited, just expressing my expectations of the developers.

[t4nk83 10]

Bohemia just announced a new patch is being released that will completely fix the FPS performance problem. Should see gains of 100-200% FPS increase with the patch, as long as your hardware meets the recommended requirements.

This is what I hope to be reading sometime in the near future. I paid $60 for this game and I want to experience what I know its capable of.

Sorry to get everyone excited, just expressing my expectations of the developers.

Troll rating 4/10

[bvrettski 10]

Arma is not "a multiplayer game". It's a game designed for single player that features a multiplayer, imo.

Taken directly from the arma3 website:

http://www.arma3.com/features/multiplayer

ARMA 3 WILL PLAY HOST ONCE MORE TO THE ARMA SERIES' MASSIVE MULTIPLAYER PLAYERBASE WHO, WHEN COMBINED WITH THE ABILITY TO CREATE CUSTOM MISSIONS AND CONTENT, ARE PLAYING A VAST ARRAY OF OFFICIAL AND USER-GENERATED MISSIONS AND GAMETYPES DAILY. FROM 2-PLAYER CO-OP MISSIONS TO 60 VS 60 COMPETITIVE WARS, ARMA 3 FEATURES SOMETHING FOR EVERYONE.

COMPETE

Should you choose to strike out on your own and wage war against your fellow players, you can. Arma 3 comes with no capped player limit, allowing for massive player vs player operations across the vast expanse of The Republic of Altis and Stratis. In the new Seize and Defend game modes, players can choose to compete against each other in a race to the finish.

I'm pretty sure its multiplayer.....it just doesnt run well and while I dont disagree with you that much of the issues we see may revolve around the AI and the client side / local engine there is certainly some portion of the issue that comes from the BI's multiplayer architecture. We see desync, and performance degradation across all servers and all missions unless they are short lived. Major in game events like vehicle and building destruction, airstrikes etc echo across the server in lag spikes and desync.

Try playing that 60 vs 60 mission for a few hours and you will watch the performance fall as the mission wears on..I dont care how well the mission is written or what server its on. Eventually it will degrade and become unplayable.

Bottom line BI didnt stress test it or they would have seen this coming whether it was client side or server side. My guess is its both.

All I want is some commitment from BI that they are going to fix it. I want a benchmark to reach. A performance level to achieve before they are satisfied with the game and we have a game that is playable. That is the game they seem to be saying they were going to deliver.

Edited September 19, 2013 by Bvrettski

[MavericK96 0]

I'm sorry that I didn't explain my proposal more precisely.

I base my proposal on article in the developers blog: http://www.bistudio.com/english/company/developers-blog/91-real-virtuality-going-multicore

Article clearly defines that "Rendering is still the part of the game which takes most of the CPU time in ArmA, more than simulation or AI." And by that the developer means that the renderin part before the data is passed on to GPU for rendering as people know it.

My proposal was to ease the burden of the CPU rendering side, and if the overall burden would come down by simplified models or dropped ones, it would also help the weaker GPU's.

I see. Keep in mind this was from 4 years ago...this may have changed by now.

[Redphoenix 1539]

That said, Arma is not "a multiplayer game". It's a game designed for single player that features a multiplayer, imo.

Wait.

Did I understand that right, or didn't I get the point?!

Did he just said that Arma is a single-player game?!