General Discussion (dev branch)

[jumpinghubert 49]

Performance feedback on latest dev. 48vs48 pure infantry firefight @altis:

stable version: 41fps average

latest dev: still 30fps average (like the last dev update before christmas)

no change with latest dev

[dr. hladik 231]

Performance feedback on latest dev. 48vs48 pure infantry firefight @altis:

stable version: 41fps average

latest dev: still 30fps average (like the last dev update before christmas)

Thank you, will test it tomorrow.

[jumpinghubert 49]

Thank you, will test it tomorrow.

thank you very much :) I am inf in a line of 4 groups against 4 enemy groups 300m away, most each other in sight from the beginning so all around in heavy firefight. If you need the mission.sqm I can upload.

[MavericK96 0]

It introduces the ugly black halo around character though that I always hated with real time AO.

Yep, just tried it here and same thing. Performance hit is much less than SSAO and actually it looks like it makes more of a visual difference than SSAO for me.

[Bouben 3]

Yep, just tried it here and same thing. Performance hit is much less than SSAO and actually it looks like it makes more of a visual difference than SSAO for me.

I have found the original AO method (before this dev update) to be superior to what we have now on any settings. It really made some stuff look really photorealistic (mainly indoors). Now it is just fancy and somehow effected.

[Bouben 3]

Hi. Title says it all.

If my observation is valid, would it be possible to make it more performance friendly? I already have particles on low and still getting significant fps loss when firing full auto.

What do you think, guys?

[Brisse 78]

Are you sure it's the graphical effects, and not the ballistics modeling?

[Bouben 3]

Are you sure it's the graphical effects, and not the ballistics modeling?

Good point.

[Brisse 78]

Good point.

Maybe you can monitor CPU-load and GPU-load while firing? Ballistics are calculated on CPU while I guess particles and muzzle flash should show up as increased GPU-usage. Maybe particles affects CPU too but I'm not 100% sure how the game engine works.

[dazhbog 10]

I have found the original AO method (before this dev update) to be superior to what we have now on any settings. It really made some stuff look really photorealistic (mainly indoors). Now it is just fancy and somehow effected.

SSAO has not been changed at all except for renaming the options (Standard, High, Very high -> Low, Standard, High). The only change is the addition of HDAO.

[Bouben 3]

Maybe you can monitor CPU-load and GPU-load while firing? Ballistics are calculated on CPU while I guess particles and muzzle flash should show up as increased GPU-usage. Maybe particles affects CPU too but I'm not 100% sure how the game engine works.

Will try. Thanks.

[Bouben 3]

SSAO has not been changed at all except for renaming the options (Standard, High, Very high -> Low, Standard, High). The only change is the addition of HDAO.

Here comes the placebo effect. Thanks for the info.

[DarkDruid 96]

Hi. Title says it all.

If my observation is valid, would it be possible to make it more performance friendly? I already have particles on low and still getting significant fps loss when firing full auto.

What do you think, guys?

There is about 150 small particles with particles on low when you firing full auto. Performance hit should be very small. It is less particles then a single grenade explosion and approximately 5x less than effect of one destroyed vehicle (and these particles are much bigger, so performance hit is even bigger). It is very unlikely that this particle effect has significant impact on your performance.

Any performance hit is probably caused by something else. You can report it on Feedback Tracker (attach please a repro mission in that case) if the hit is really significant. :)

[Bouben 3]

There is about 150 small particles with particles on low when you firing full auto. Performance hit should be very small. It is less particles then a single grenade explosion and approximately 5x less than effect of one destroyed vehicle (and these particles are much bigger, so performance hit is even bigger). It is very unlikely that this particle effect has significant impact on your performance.

Any performance hit is probably caused by something else. You can report it on Feedback Tracker (attach please a repro mission in that case) if the hit is really significant. :)

Thank you for the explanation. Good to know.

[Fushko 59]

HDAO is a godsend. It looks and performs a thousand times better than SSAO. I can finally play the game with high quality ambient occlusion without my frames dropping like hell. GJ

[SaOk 112]

Noticed that I am getting high 58 FPS (using diag_fps) also near Pyrgos with 64 AI units & 26 Vehicles, but for some reason it stutters much at the same time. More far from big villages the FPS stays quite the same but the performance is very smooth.

Edit: Now 67 FPS near the Pyrgos with stutter. Dosent make sense. :)

Edited January 10, 2014 by SaOk

[trnapster 12]

Measure your frametimes with MSI afterburner.

[-Coulum- 35]

I don't know if it is the muzzle smoke or what, but something is definitely up when firing rapidly. I think it gets worse with the amount of zoom you are using. Here are my findings on the stratis airport looking at the north end of the runway and firing a full clip from the mx.

[TABLE=class: grid, width: 500]

[TR]

[TD]View[/TD]

[TD]Not firing[/TD]

[TD]Lowest FPS while unloading 30 shots full auto[/TD]

[TD]FPS Loss[/TD]

[/TR]

[TR]

[TD]normal naked eye[/TD]

[TD]58[/TD]

[TD]45[/TD]

[TD]13[/TD]

[/TR]

[TR]

[TD]using ironsights + naked eye zoom[/TD]

[TD]62[/TD]

[TD]45[/TD]

[TD]17[/TD]

[/TR]

[TR]

[TD]naked eye zoom[/TD]

[TD]62[/TD]

[TD]41[/TD]

[TD]21[/TD]

[/TR]

[TR]

[TD]using HAMR magnified sight[/TD]

[TD]60[/TD]

[TD]37[/TD]

[TD]23[/TD]

[/TR]

[TR]

[TD]Using the SOS first level of magnification[/TD]

[TD]60[/TD]

[TD]25[/TD]

[TD]35[/TD]

[/TR]

[TR]

[TD]Using the SOS second level of magnification[/TD]

[TD]60[/TD]

[TD]24[/TD]

[TD]36[/TD]

[/TR]

[TR]

[TD]Third Person[/TD]

[TD]60[/TD]

[TD]57[/TD]

[TD]3[/TD]

[/TR]

[/TABLE]

In addition, using the SOS at its highest magnificatjion I looked up at the sky. I had 93 FPS. I unleashed a clip. My FPS dropped all the way to 36 FPS at its lowest. A loss of 57 FPS. Something is definitely eating up FPS when you fire and it seems to be linked to magnification/zoom.

Are you sure it's the graphical effects, and not the ballistics modeling?

If it were ballistic modelling wouldn't that mean that anyone firing would eat FPS? Right now its only the players shooting that effects FPS. 2 ai can be spraying each other full auto next to me and the performance drop is minimal compared to when I fire.

Apparentlly this mod by Lao Fei Mao tries to address the problem though I haven't gotten the chance to test it.

Mao_Muzzle_Smoke.pbo ----decrased the muzzle smoke density so that the FPS won't drop drastically while you auto fire a rifle or machinegun with a high zoom rate optic.

Edited January 10, 2014 by -Coulum-

[Punisher5555 0]

So if there is a proven MOD that addresses it, shouldn't BIS take this MOD and put it in the game permenately?

[-Coulum- 35]

So if there is a proven MOD that addresses it, shouldn't BIS take this MOD and put it in the game permenately?

Have you tested the mod? Are you sure it is proven? It tries to fix the problem. I don't know if it works or not. Has anybody tested it?

If it does work I am sure BIS will take similar steps as the modder did to fix it for vanilla. But first they need a ticket on the feedback tracker.

[Sniperwolf572 758]

I don't know if it is the muzzle smoke or what, but something is definitely up when firing rapidly. I think it gets worse with the amount of zoom you are using. Here are my findings on the stratis airport looking at the north end of the runway and firing a full clip from the mx.

[TABLE=class: grid, width: 500]

[TR]

[TD]View[/TD]

[TD]Not firing[/TD]

[TD]Lowest FPS while unloading 30 shots full auto[/TD]

[TD]FPS Loss[/TD]

[/TR]

[TR]

[TD]normal naked eye[/TD]

[TD]58[/TD]

[TD]45[/TD]

[TD]13[/TD]

[/TR]

[TR]

[TD]using ironsights + naked eye zoom[/TD]

[TD]62[/TD]

[TD]45[/TD]

[TD]17[/TD]

[/TR]

[TR]

[TD]naked eye zoom[/TD]

[TD]62[/TD]

[TD]41[/TD]

[TD]21[/TD]

[/TR]

[TR]

[TD]using HAMR magnified sight[/TD]

[TD]60[/TD]

[TD]37[/TD]

[TD]23[/TD]

[/TR]

[TR]

[TD]Using the SOS first level of magnification[/TD]

[TD]60[/TD]

[TD]25[/TD]

[TD]35[/TD]

[/TR]

[TR]

[TD]Using the SOS second level of magnification[/TD]

[TD]60[/TD]

[TD]24[/TD]

[TD]36[/TD]

[/TR]

[/TABLE]

In addition, using the SOS at its highest magnificatjion I looked up at the sky. I had 93 FPS. I unleashed a clip. My FPS dropped all the way to 36 FPS at its lowest. A loss of 57 FPS. Something is definitely eating up FPS when you fire and it seems to be linked to magnification/zoom.

If it were ballistic modelling wouldn't that mean that anyone firing would eat FPS? Right now its only the players shooting that effects FPS. 2 ai can be spraying each other full auto next to me and the performance drop is minimal compared to when I fire.

Apparentlly this mod by Lao Fei Mao tries to address the problem though I haven't gotten the chance to test it.

Sounds like overdraw issues to me. Even a single transparent particle close to the camera can cause more performance impact than several distant ones. You might want to try the same experiments from 3rd person, FPS impact will probably be much less if it's overdraw.

The logic behind what I said above is that a single transparent particle that is close to the near clipping plane of the camera will result in many pixels that need to be overwritten due to the perspective, compared to a distant particle which is really small in the distance and will affect only a small amount of pixels.

Now figure, going by DarkDruids data, 150 particles, just from the weapon, that are covering each other up, right next to the camera and they are all visible and probably covering up a large portion of the screen because you are so close to them. Compared to a single grenade that might be more particles, but less overall pixels to alter per frame.

[-Coulum- 35]

I actually did test in 3rd person just forgot to record it because there really wasn't any FPS loss. About 2 or 3 FPS decrease max. Don't know much about particles but what you say makes alot of sense. If that is the problem I wonder what the best solution would be without making the smoke look bad when viewed by someone other than the shooter?

[Sniperwolf572 758]

I actually did test in 3rd person just forgot to record it because there really wasn't any FPS loss. About 2 or 3 FPS decrease max. Don't know much about particles but what you say makes alot of sense. If that is the problem I wonder what the best solution would be without making the smoke look bad when viewed by someone other than the shooter?

Then it seems quite clear what the problem is as it only affects the shooter in first person views. It would be necessary to somehow either cull the amount of particles that are close to the camera and that are stacked on top of each other while maintaining the effect.

Like a way of increasing the opacity of particles when they cover each other up in close proximity to the camera. Or using a smaller number of particles, but with animated textures that are producing similar effects as multiple particles layering on top of each other for the first person view.

---------- Post added at 21:22 ---------- Previous post was at 20:35 ----------

Anyway, to explain what I'm talking about is this. Look at the two pictures below, the left one is the area hand grenade particles affect on your screen and the right one is the area gunfire particles take up on your screen. They're ofcourse, approximations, but I'm sure someone could create some fully opaque particles that could give you an accurate representation of the area each cover.

Now, the breakdown is:

Hand grenade - 40k pixels

Gunfire - 200k pixels

Now, we'll say this was all done in one particle (it wasn't, but roll with it). This means that in a single frame, that amount of pixels need to be adjusted for the transparent particles. The difference is that the gunfire smoke, no matter how barely noticeable or faded it appears, needs to touch 400% more pixels than the hand grenade.

Now, let's assume there are 80 particles are on screen to generate the gunfire smoke, and that only half of them are intersecting, so that makes it 8 million pixels. And let's be generous and say that hand grenade also generates 80 particles and also half of them are intersecting, that's 1.6 million pixels to alter. Which basically means you'd have to have 5 hand grenades going off at the same time to generate the same amount of pixels to be altered as the gun effect because of the size of the effects on the screen.

Now this is simplified a lot and there's a lot more that goes into it, but now you can see how the amount of particles can be less of a problem than the size they are on the screen.

Now a bit unrelated, but Computerphile recently had a decent episode to explain some visibility problems while rendering a frame that also affect particles.

And you can find the playlist for this series

. They're really good and educational and provide insight into how our favorite past time gets to the screen. Edited January 10, 2014 by Sniperwolf572

[Bouben 3]

Guys, interesting finds here. I am glad that I brought here the issue.

Edited January 11, 2014 by Bouben

[Bouben 3]

Allright so I tested the mao's muzzle particle/smoke mod and it did solved the problems I described here. So it was indeed connected with graphics and not ballistics.