SOS: Save our Smoke!

[Nealius]

No change in FPS if there is smoke or not.

 

Like Kameni said, not everyone is running a $1200 6GB card, and IMO it's ridiculous to do any testing with such hardware because then it completely skews the results.

 

Has anyone tried Ebs's little fix?

[SkateZilla]

I've kinda always had an issue with some missions and flying into the sunset.

[Nealius]

I've kinda always had an issue with some missions and flying into the sunset.

 

I am starting to wonder how HDR affects framerate. I noticed that when I am on the runway with landing lights on AND the sun behind me things get slow when I am looking through the HUD.

[TZeer]

Like Kameni said, not everyone is running a $1200 6GB card, and IMO it's ridiculous to do any testing with such hardware because then it completely skews the results.

 

Has anyone tried Ebs's little fix?

 

It's not. I get the same effect as everyone else regarding this issue.

 

But I agree that people should try with other hardware as well. But they might get performance drop from something that is not related to this issue at all, like lack of VRAM, etc. Or they can hit a CPU bottleneck.

 

I have moved from single 580 to 580 SLI and now a Titan, and what I have noticed is that the more power you have, the easier it is to spot performance "thieves".

 

Earlier I could maybe get 40-50 FPS in a scenario, and maybe 5 FPS drop from something in the game. At that time I would go, ok, that would be normal I guess. But then as the more power you get the less bottleneck you have in the system, and those 5 FPS, suddenly turns into 30-40 FPS.

[Nealius]

Particle effects like smoke rely on VRAM no? It amazes me that with 2GB of VRAM a single smoke plume would hurt my framerates so badly. And in that case the smoke plume can be the only possible cause, because it was the only thing present.

[TZeer]

We don't know what is causing it right now.

 

What you can do is remove all the smoke effects as described in this thread and post your findings with screenshots showing your FPS counter. That way we could see how the effect is at your setup.

 

Even if you don't see it, there is lightning effects around each wreck and building that burns. You can easily see this if you change the time to night.

 

 

Even if I have a powerful card, I should have seen a difference when removing all smoke and fire effects if they where the big performancekiller. Specially since my card is already working at 100% before I remove the smoke.

[ENO]

Correct me if I'm wrong but the biggest hit was when you lined up the smoking column of multiple vehicles? Sounded like even the Titan slammed to a halt- unless I misunderstood the fact the Gpu was at 100% and it was still rendering at 20-40 fps.

 

Also keep in mind this is quite a small scale (roughly 1/3rd) of the original mission issue resulting in 7fps on a 680.

[TZeer]

I turned off vsync, so my card is working at 100% all the time. Did it that way so I can monitor FPS instead of GPU load.

 

I thought I got the biggest hit when aligning the smoke columns... Until I turned off all smoke and fire effects, and realized it gave me no performance gain.

 

What's left to test are those light effects that are around each wreck and burning building. But I don't know which files that controls that.

[Nealius]

Turning off all smoke resulted in a performance gain for me, however I run with Vsync on because tearing makes my retinas bleed. I think with the no-smoke mod I downloaded the flames/glow were also removed, but I don't know for certain.

[TZeer]

Could you take screenshots and post your results?

 

For testing you need to take off your Vsync, as it will interfere with your FPS.

 

To see if your flames/glow were removed, change the time of the testmission to night, and see if you have any lightsource from the wrecks.

 

You also need to be very precise to how you angle your camera when looking at the wrecks/smoke. If you get the horizon in the background, you will get performance drop due to number of objects in the scene. Or you can reduce the visibility range to minimum during the test.

 

But yeah, screenshot would help a big deal as it would let us see the performance before/after and at what angles and positions your getting it.

[Nealius]

The method of testing doesn't really make sense. Testing should be done in real conditions, not ideal conditions. Vsync off is an ideal condition because it allows upwards of 150 frames per second, and any loss is not going to be noticeable until it gets below 60 or so. In reality people like myself play with Vsync on.

 

Also, how does the horizon effect performance? I have no (noticeable) framerate drop when looking at the horizon.

 

If I can stand going through the whole GAU-8 training session twice I'll do it with Vsync on and off.

Edited April 2, 2013 by Nealius

[TZeer]

Yes, it allows up to 150 FPS. But it doesn't mean you gonna get it.

 

If you lock your FPS at vsync you will have to monitor your GPU usage on a secondary screen to keep an eye on the load you get.

 

Then as more and more stuff comes into play the load will go up. Until it come to the point where the card is not able to deliver 60 FPS.

 

 

If you deactivate vsync your GPU will give out as much FPS as it possibly can. And you will easily spot when the FPS is lower then what your expect, or when it starts getting affected by whatever you are throwing at it.

 

When you are trying find out what is affecting the performance you want to get the whole specter. Not just one specific point bellow 60 FPS. Take the smoke/fire/light/whatever thing we have here. To be able to see how it behaves you need to do it from the moment it starts affecting the GPU/using it's resources. Not just when it dips below 60 FPS.

 

Example:

Let's say you have a unit, and you place it in an empty place in the editor. With any other unit you would go from maybe 130 and down to 120 FPS, but this specific one reduces your FPS to 70 FPS. With vsync on you would never notice, and you would think, why care, I can still run at 60 FPS. But you should care, as that unit are then stealing resources that could be used to keep your game running in 60 FPS when it suddenly are competing for resources with other visual effects.

 

 

Horizon:

Everywhere you look in the game there are 3d models. Every 3d model consists of objects. Those objects are dealt by the CPU and not the GPU. If you get to many objects within FOV you can see your GPU load will go down while the FPS drops. This is due to the CPU being a bottleneck by the workload produced by the objects, and not able to feed enough data to the GPU.

 

So, if you tilt your view to much and the horizon is visible, this can affect the FPS great deal and give results that have nothing to do with the smoke/fire/light effect.

 

So, instead of arguing about test-method and telling how you feel it is for you. Post your results, with screens, explanation and your system specs so we can hopefully crosstest with other with similar setup and verify.

 

 

Btw, I'm using Vsync as well when I normally fly as I can't stand tearing :pilotfly:

[Nealius]

I'll get on the testing once I get time to fly this afternoon. One thing doesn't make sense though:

 

Let's say you have a unit, and you place it in an empty place in the editor. With any other unit you would go from maybe 130 and down to 120 FPS, but this specific one reduces your FPS to 70 FPS. With vsync on you would never notice, and you would think, why care, I can still run at 60 FPS. But you should care, as that unit are then stealing resources that could be used to keep your game running in 60 FPS when it suddenly are competing for resources with other visual effects.

 

The human eye can't really see drops in framerate until it goes below 30. With Vsync on this can happen very easily. With Vsync off, and a framerate of 150, if it drops down to 120 or 70 it's not going to be very noticeable. The initial plunge from 150 to 70 might be noticed, but the game will still be smooth after the drop. Which then results in an "everything is fine, just play with Vsync off"...

[TZeer]

I'll get on the testing once I get time to fly this afternoon. One thing doesn't make sense though:

 

 

 

The human eye can't really see drops in framerate until it goes below 30. With Vsync on this can happen very easily. With Vsync off, and a framerate of 150, if it drops down to 120 or 70 it's not going to be very noticeable. The initial plunge from 150 to 70 might be noticed, but the game will still be smooth after the drop. Which then results in an "everything is fine, just play with Vsync off"...

 

Depends who you ask. I'm very sensitive to FPS drops. 30 FPS seems like a slideshow for me, specially when looking out to the sides of your plane. Seems ok when looking straight ahead, but as soon as I turn my view to left or right my eyes starts bleeding. So I have tweaked my system to stay at 60 FPS as much as possible.

 

The point isn't to get as high FPS as possible, it's to make the GPU work at it's maximum and see how much performance is lost due to specific scenarios. Take my example in previous post. That drop from 130 to 70 would indicate that there's something wrong with that unit, and it would need optimizing. Would you be able to put it into the game? Sure, but it would use resources that it would not have needed if the unit had been optimized and done properly in the first place. And in the end pulling down the overall performance of the game.

But, if someone is already having performance issues due to RAM, CPU or whatever, they might never notice this, as part of their system is already running at maximum.

 

Anyway, looking forward to your tests.

[Nealius]

Do I need to monitor my GPU/CPU usage during the tests then? I only have one screen so the best I can do is record a log with MSI Afterburner.

[Wolf Rider]

Don't confuse the "with what the eye can see", with how the monitor behaves as different (to its native) fps are displayed.

 

The eye certainly notices the difference between 100HZ and 60Hz on a CRT monitor with both a static image and motion.

[TZeer]

If you cant monitor you GPU on a secondary screen a log will be good enough. Hopefully your GPU will be working at 99-100% through the tests. If it doesn't you have a bottleneck somewhere in the system.

[Nealius]

Maybe there is a bottleneck in my system. I tried the training flight again while recording a track and I didn't experience the same performance hit as I did the last time. My framerate dropped, but it never went below 35 or so. Lowest was 22 when cannon rounds impacted the ground.

[TZeer]

Did you log the GPU usage?

[ENO]

It's funny because from what I'm reading here my CPU is a bottleneck for this kind of problem (because it's working on those small out of sight objects that are still- if magnified enough- in view... and I have a 6 core intel.

 

I know, I know- it's because the game only runs on 2... and that's all being done by 1... I'm just remarking that this issue may disappear when the load can be spread across at least 4 cores.

[Nealius]

Did you log the GPU usage?

 

No but I can run it again. I'm running an HD7850 2GB with only x2 MSAA and TSSAA off, resolution 1280x1024 and settings set to Medium. So if anything I shouldn't be pushing my card very hard at all.

 

CPU could be the bottleneck, Phenom II X4 3.4GHz.

[TZeer]

It's funny because from what I'm reading here my CPU is a bottleneck for this kind of problem (because it's working on those small out of sight objects that are still- if magnified enough- in view... and I have a 6 core intel.

 

I know, I know- it's because the game only runs on 2... and that's all being done by 1... I'm just remarking that this issue may disappear when the load can be spread across at least 4 cores.

 

If your GPU is working at 100% during your testing with the smoke effects, your cpu is fine.

 

If your GPU is throttling down it's gpu usage to less then 98-100%, there is something else then your GPU that is holding you back.

[Cobra99]

@TZEER-I really am not understanding how you are getting the TiTan to run @ 100%. The highest I can get it is about 81% and I am running a i7 980X overclocked to 4.2Ghz. So you are saying that the CPU is still holding the system back? What CPU are you currently running and at what clock speed?

 

Regards,

 

Cobra

[ENO]

Keep in mind Cobra- the game only runs on 1 core (the second is all sound)... So whether you have 6 cores, or hyperthreading with 12... or a dual CPU SR-2 with xeons... we're going to hit that hurdle.

 

We can clock to 4.2 but some of the boys with their sandy / ivy's can hit that in the bios with their monitors turned off. A modest effort gets them in the early 5s- unfortunately in this case it's speed that is king, not number or cores.

 

So while I can't answer the question above- remember OC'd 980X does absolutely NOTHING for this game. (yet)

[Cobra99]

@ENO - overclocking from 3.9Ghz to 4.2Ghz did provide a gain in testing of 16% so Overclocking any CPU is a benefit. My question is how is he managing to get 100% load on the TITAN.

 

Regards,

 

Cobra