[INVESTIGATING]Black Smoke very costly on FPS, Engine Exhaust, Smoke Stack.

What is causing stuttering for me for quite some time now, this is NOT limited to 2.5.6, it's the engine exhaust from F-18 that would cause FPS dips whenever and it took me a long time to kinda pinpoint it and actually realize it.

But it only happens randomly when playing usually, because you're actively rotating the camera in F2 and your airplane, the camera and the engine's of the airplane(x axis) have to come in almost perfect alignment and distance for this to have the maximum effect, not counting the speed and angle of all the physics of the smoke that blows it out of that alignment, so yeah you don't need that alignment, if the smoke blows into your camera you're going to get hit hard.

It may also be a configuration bug, along with the A-10C noisedive thing, perhaps smokes are that costly only for me, because I didn't heard anyone talk about it in terms of FPS, but yeah it took me long to figure it out too.

So I'll be updating GPU drivers shortly as I was planning to do so for some time around other issues too, I wanted to test things out like this first, then again after that.

White Contrail Smoke, at high altitude, does not kill FPS at all, there is some slight dip if you do get a lot of them on screen like with a 4 engines of a AI tanker, but a fraction of what the relatively newer type Black Smoke does.

I haven't tested with others than F-18C yet, but I will, but I'm sure it happens else where too, it seems like the black smokestacks may be using the same type of smoke which lags a lot.

Which was somehting I noticed way back last week, but didn't report it as I was working on it determining whether the ocean fidelity and waves have something to do with it because when you do come to a certain zoom level the oceal LOD coupled with the amounts (draw calls) causes a similar dip in FPS.

So I set the sea quality to lowest, and here's what I got with the Admiral Kuznetsov Smoke Stack.

I went on and did the other aircraft I own. these are water tests, to get most other factors out of the way. Water quality was set to LOW.

---------------------

---------------------

---------------------

---------------------

---------------------

Same for Su-25T and Mig-29S

There definitely is variation due to amount of engines and size how much smoke is produced but it does have a similar effect across the board. In some aircraft it's barely seen and it still makes a big difference.

A-10C, A-10A, do not have the Black Engine Exhaust Smoke.

Reserved for updates.

Basic Example Track: https://forums.eagle.ru/attachment.php?attachmentid=229195&stc=1&d=1583278297

------------------------

Update 1:

F/A-18C Terrain Tree Test, Maximum tree radius.

--------------------------------------------

--------------------------------------------

Update2: This is NOT limited to 2.5.6 OB AFAIK, I have used the current build to do this, but this one was on my head for some time, I just never got to the bottom of it and kept forgetting.

Updated the OP accordingly.

--------------------------------------------

--------------------------------------------

--------------------------------------------

Update 3:

Seems like a truckload of pixels ...

NO SMOKE

SMOKE SIDE

SMOKE INSIDE

Update 4:

A cheesy list of smoke types: (unofficial)

• Black Engine Exhaust Smoke (BEES) - Double draw calls. Super laggy.
  
• Gray Wing-Destroyed Smoke (GWDS) - Single draw call (so 1 for each wing/streak)
  
• High Altitude White Contrail Smoke (HAWCS) - Double draw calls. Very fast.
  

This is initial profiling me using this utility, hence just a peek, not really fair to call it analysis or diagnosis, I didn't do any of this prior ever, so this is the real deal walkthrough, but I cut up the noninteresting pieces out good, neither I did any advanced comparisons or hunting down the details, but on the surface it's quickly evidence by the BEES is the most costliest of smokes by far, so much, that it raises the question whether it's a bug or not optimized, or just design flaw or something.

Note that it's the "GPUTime" that matters for FPS AFAIK, other times I think are aggregated/summed as they work in parallel but don't hold me for my word, that's why PSTime can be so high and that's why other times could be higher than GPUTime, apparently, but I'm only beginning with this utility.

Double draw calls, which can both get very coslty with BEES was a bit of a side thing I wondered, but thought it's probably just the positioning and sue of different color channel for the pixels to get the pattern and the final effect right, and I still do, but the the big oops might have been if it's, get a load of that, rendering in stereo when not in VR Mode, potentially wasting 50% of the GPU, because I never ever had VR on this machine nor any VR utility or SDK or anything.

But that didn't look like the case as I haven't found good clues on other draw calls where it deals with models, pretty much the whole frame should have things doubled and I didn't detect that, but if these double draw call on certain smokes including BEES really are for VR then it makes BEES that much laggier. It does affect HAWCS too but that's a totally cheap smoke on it's own.

The clue is left vs right shift in these pink draw call affected area markers, you can't see the smoke in this case, these line outline a square area of the smoke object effect internally, and you can see the density of how bazillion a lot of these boxes coming out of the exhaust in a cone shape toward the camera.

Usually performance suffers due to CPU bottleneck with draw calls, in this dase so much of this is in one draw call, and since the rest of the scene is pretty (sky/model), there's no other draw calls to slow the CPU, which makes this an entirely GPU bottleneck.

-----------------------------------

Update 5:

I have got access to another machine temporairly which I installed DCS World on previously (free), older than mine but with what seems to be a better GPU, a weaker CPU tho, but since these tests are isolating everything possible except the smoke it's doesn't matter much it seems, the FPS is quite high there.

That machine more free space so I put the latest OB alongside an older 2.5.5.xxxxx Stable from somewhere in fall of 2019 (I didn't update to latest stable for this test) so I can test both there.

Initial quick analysis (without advanced tools, just the basic overlay) and the FPS drop appears to be there as well, but since it's around 150FPS when viewing sea and sky and it goes over 200 when viewing only the sky (and the aircraft model) it seems it kinda doesn't matter in practice because it's not noticable to the eye if you'd be playing for entertainment, withouth actively looking for issues. The FPS drop is like from 170 FPS to 120 FPS in F/A-18C the standard Black Smoke test with the camera some 30-50 and 100-200 meters behind.

That is why potentially affects more people but they were not aware of it due to a beefy GPU, not being in F2 View much in the precise conditions and other thing. especially I guess many of those on 1080p which is still the gaming standard AFAIK or at least many who haven't upgraded yet. That's why you need a overlay with histogram or a profiler/monitor to log the performance statistics to be able to notice this better, looking just at the SYSINFO FPS counter can be unreliable because the FPS fluctuates significantly depending on the portions of terrain-ocean-sky on the screen. Offline is also hard, there's so many dips and ups and you won't find the point in time accurately, unless you take the effort to time things to the second.

The MSI/RSST overlay you see on the right side in many of my videos is manually configured and not something out of the box, I could say it's not that hard and it's not, but fiddling simple things can take more time than it looks, to get things just right can turns into hours, it puts most people away from beta testing as it can take you whole afternoon and more, and the next day, and so on.

I will recreate the settings that were on that machine's stable version, use identical missions/scenarios, and the 1080p resolution for a 1080p monitor, so I just put mine to 1080p as well and see how it goes.

I'll be digging more tomorrow, because now I can compare totally different hardware, that machine GPU is a Nvidia, and the smoke looks less dark on the Nvidia machine (but that could just be the HW difference in how coloring looks which we know for a long time), but if it's a HW configuration bug affecting only me and certain machines, perhaps on the unaffected machines the smoke is less dense? We'll see.

-----------------------------------

Update 5:

1. Grim Reapers on youtube also mentioned they noticed this (the head guy's PC at least)

2. Okay, it's not stereo rendering lol, it's double draw call becuase it's one smoke effect for each engine. I just didn't happen to only try it with one engine, or had the chance to test a single engine'd one.

3. The difference in smoke color that I noticed on another borrowed machine is probably due to Monitor type, it's TN, and my is IPS, that would have a huge effect on it self, so nothing to worry about there.

Reminder: Those pink lines don't mean necessairly anything's wrong or too much, it's an outline of the objects in that draw call, the area being drawn, you can see the white smoke has a lot of them too the same way black smoke does, but white smoke is around 20 times less demanding if I eye ball it, if not more.

You can see the right engine contrail smoke draws being highlighted (whole area, not wireframe like in other pictures), while the left one is not, so they're to separate effects for each engine, makes sense ofcourse, but it kinda adds to a lot of smoke saturation that overlaps/shrouded that hopefully isn't being rendered, not a problem with white smoke as it never takes too much, perhaps that's what is happening with the Black Smoke, that all of the smoke objects are rendering completely, even if obstructed? But Black Smoke is more transparent seeming and perhaps that's how it is designed to make the desired effect look right depending on various conditions and amount of smoke. The Black Engine Exhaust Smoke looks totally different in design versus the White Contrail Smoke and Smoke Generator Smoke anyway, it's not just a color change at all.

However the smoke generator on F/A-18C only sprays from the left engine, I do not know if that's by design or not.

---------------------------------

---------------------------------

As I said there is not.

Allright if you say so, ... just you know it can be tricky as eyeballing it doesn't really work to good if FPS is above 100 at all times because without screenshots I can't be sure what you guys are doing.

There is a suspicious case I've got captured where the effect would not have a FPS drop, I don't know what I did different, but I was using 1080p windowed here, but I was using it before as well, and the game was in mission-start-pause for quite some time, only when I did the active pause black hole it started dropping the FPS.

Maybe some kind of a clue that it is a bug and perhaps only happens in certain conditions, perhaps the smoke is affected by environmental conditions, or that the FPS was CPU bound and GPU had enough headroom? I'll see later but I'm gonna need a break sometime soon.

Now I'm finally going to update the drivers and see what happens, I didn't want to on purpose in the middle of testing to do as much here to have enough to compare.

-----------------------------------

Update 6:

Reinstalled AMD/ATI Drivers with Factory Reset (deletes all other versions, settings, device driver history) to 20.1.3 WHQL (from 19.5.2 I think) and kept at default settings as per usual, I don't play with GPU settings much in terms of gaming tuning, as I usually don't get bothered by aliasing that much, but I was using 4x MSAA in DCS.

Well there's no difference, did a quick test (now I need a break, perhaps a day or two) same Black Smoke behavior, 4xMSAA or not. The only thing there's different is just the FPS min and max bounds shift higher up, but the big drop is still there.

Same

SAME

Reduce fps effect of the smoke I got -57%.

Hint: It's not amount of draw calls ... update soon.

I would have posted it by now but I got something in the way, got a bit out of memory, had to reboot, then another tool crashed due to it's own bug and have to wait for crash report upload, then I have to update 5 other utilities, but this is the life of beta testing :)

But it's nothing we can use, I don't have a solution to fix now or anything, it's for the devs if it helps, they should figure it out anyway.

I'll be updating drivers after this.

My rig is quite fast so I don't notice an FPS drop. I could look but I'm sure it's there.

However, I would call the black smoke an actual bug. Why? No airplane, in afterburner, generates smoke of any kind. Just intense heat. Afterburner is a complete fuel burn. Black smoke is the result of an incomplete burn of fuel. If you have ever seen a diesel truck or bus running along on a flat road, there usually is very little smoke because the engine is burning the fuel efficiently. However, lug the engine or when the vehicle climbs a steep hill, you see black smoke because a lot of fuel is being used which lots of it fails to completely finish oxydizing.

Same with jet aircraft.

In afterburner, fuel is simply being dumped into the tailpipe and ignited. It oxydizes rapidly generating a lot of heat (hence the red metal and gasses). All the fuel is burned efficiently. No smoke at all.

I think its more of a spotting aid IMO, i can actually spot other AC in VR much easier.

So did you actually try it out, you got to be sure, it may be just certain PC configs/drivers issue, it could only be an AMD issue, you need to get camera inside the cone of the smoke that travels out of the engines but also where the smoke flows but that would be due to manouvering or wind, so try to stay level.

It's important that we get people who may not have this issue at all, then it's not the game's fault so much.

I'd be surprised if your GPU eats through all that without a hitch, unless the bug doesn't affect you.

I've seen F/A-18C and Su-27 to get into pretty jaw dropping cases there, 14 FPS down from 80 :)

Update, added some findings, check out the second post.

But this is the barebone sky test, so the ratio between smoke and no smoke is huge, we'll see the diff with terrain and other activity tomorrow, I'm out for today. It won't look so dramatic against other stuff but if it cuts your fps at low level terrain from 45 to 20 FPS that's a big deal IMO.

Good find and analysis! I also noticed a severe FPS drop yesterday on the carrier deck (SP), when two AI Hornets next to me started taxiing to the catapults and they produced black smoke as they did. As soon as they stopped moving and hooked up, the FPS was back to normal. The FPS went from 55-60 to as low as 30 and back to 55-60, with my GPU going to 100% during the drops (I limit my FPS to 60, so my GPU is not 100% all the time).

Updated the OP accordingly.

-------------------------------

A cheesy list of smoke types: (unofficial)

• Black Engine Exhaust Smoke (BEES) - Double draw calls. Super laggy.
  
• Gray Wing-Destroyed Smoke (GWDS) - Single draw call (so 1 for each wing/streak)
  
• High Altitude White Contrail Smoke (HAWCS) - Double draw calls. Very fast.
  

This is initial profiling me using this utility, hence just a peek, not really fair to call it analysis or diagnosis, I didn't do any of this prior ever, so this is the real deal walkthrough, but I cut up the noninteresting pieces out good, neither I did any advanced comparisons or hunting down the details, but on the surface it's quickly evidence by the BEES is the most costliest of smokes by far, so much, that it raises the question whether it's a bug or not optimized, or just design flaw or something.

Note that it's the "GPUTime" that matters for FPS AFAIK, other times I think are aggregated/summed as they work in parallel but don't hold me for my word, that's why PSTime can be so high and that's why other times could be higher than GPUTime, apparently, but I'm only beginning with this utility.

Double draw calls, which can both get very coslty with BEES was a bit of a side thing I wondered, but thought it's probably just the positioning and sue of different color channel for the pixels to get the pattern and the final effect right, and I still do, but the the big oops might have been if it's, get a load of that, rendering in stereo when not in VR Mode, potentially wasting 50% of the GPU, because I never ever had VR on this machine nor any VR utility or SDK or anything.

But that didn't look like the case as I haven't found good clues on other draw calls where it deals with models, pretty much the whole frame should have things doubled and I didn't detect that, but if these double draw call on certain smokes including BEES really are for VR then it makes BEES that much laggier. It does affect HAWCS too but that's a totally cheap smoke on it's own.

The clue is left vs right shift in these pink draw call affected area markers, you can't see the smoke in this case, these line outline a square area of the smoke object effect internally, and you can see the density of how bazillion a lot of these boxes coming out of the exhaust in a cone shape toward the camera.

Usually performance suffers due to CPU bottleneck with draw calls, in this dase so much of this is in one draw call, and since the rest of the scene is pretty (sky/model), there's no other draw calls to slow the CPU, which makes this an entirely GPU bottleneck.

Forgot to note: That's the F-18C poofing that BEES, since this is getting a bit more sophisticated and I'll be looking at other smokes too for my perf diag practice and learning the other tools I just found, perhaps I find something more or not, hopefully it helps to get to solution faster :)

... but excuse me if some of these abbreviations step on something else, GWDS sounds awfully familiar, I might change it around if that's the case.

Okay, but you guys gonna confirm/test if you also have this bug or what? I need at least one case that doesn't have it at all, and should be on video too for proof. Thanks

First, I'm a bottomfeeder so I very seldomly encounter any F-18s or any such with their rear ends smoking Greta Thunberg out of existence anyway.

That said however, I don't see any F/S drops with either exhaust smokes from, say, L-39s, or ground units belching any kinds of smoke whether they're dead or alive. Which kind of was my original point, i.e. my particular brand of crappy old PC does not suffer from this problem at all. Sorry, no track, but I'll just repeat my observation of the PC platform being a crappy one to code anything for anyway. S!

First, I'm a bottomfeeder so I very seldomly encounter any F-18s or any such with their rear ends smoking Greta Thunberg out of existence anyway.

 

That said however, I don't see any F/S drops with either exhaust smokes from, say, L-39s, or ground units belching any kinds of smoke whether they're dead or alive. Which kind of was my original point, i.e. my particular brand of crappy old PC does not suffer from this problem at all. Sorry, no track, but I'll just repeat my observation of the PC platform being a crappy one to code anything for anyway. S!

Ohhh, the same type of smoke is present in many aircraft for this type of effect, including Su-25T so any could test this, and it doesn't matter if your PC is crappy, if there's a FPS drop or not that's the thing we're looking for. I'm not asking anything special, same thing the devs would ask in such a thread.

Good work, Worrazen! I've noticed this smoke/exhaust-FPS-killer for quite some time (even before 2.5.6).

Hope this helps the Devs iron some performance stuff out. :thumbup:

Update: I have got access to another machine temporairly which I installed DCS World on previously (free), older than mine but with what seems to be a better GPU, a weaker CPU tho, but since these tests are isolating everything possible except the smoke it's doesn't matter much it seems, the FPS is quite high there.

That machine more free space so I put the latest OB alongside an older 2.5.5.xxxxx Stable from somewhere in fall of 2019 (I didn't update to latest stable for this test) so I can test both there.

Initial quick analysis (without advanced tools, just the basic overlay) and the FPS drop appears to be there as well, but since it's around 150FPS when viewing sea and sky and it goes over 200 when viewing only the sky (and the aircraft model) it seems it kinda doesn't matter in practice because it's not noticable to the eye if you'd be playing for entertainment, withouth actively looking for issues. The FPS drop is like from 170 FPS to 120 FPS in F/A-18C the standard Black Smoke test with the camera some 30-50 and 100-200 meters behind.

That is why potentially affects more people but they were not aware of it due to a beefy GPU, not being in F2 View much in the precise conditions and other thing. especially I guess many of those on 1080p which is still the gaming standard AFAIK or at least many who haven't upgraded yet. That's why you need a overlay with histogram or a profiler/monitor to log the performance statistics to be able to notice this better, looking just at the SYSINFO FPS counter can be unreliable because the FPS fluctuates significantly depending on the portions of terrain-ocean-sky on the screen. Offline is also hard, there's so many dips and ups and you won't find the point in time accurately, unless you take the effort to time things to the second.

The MSI/RSST overlay you see on the right side in many of my videos is manually configured and not something out of the box, I could say it's not that hard and it's not, but fiddling simple things can take more time than it looks, to get things just right can turns into hours, it puts most people away from beta testing as it can take you whole afternoon and more, and the next day, and so on.

I will recreate the settings that were on that machine's stable version, use identical missions/scenarios, and the 1080p resolution for a 1080p monitor, so I just put mine to 1080p as well and see how it goes.

I'll be digging more tomorrow, because now I can compare totally different hardware, that machine GPU is a Nvidia, and the smoke looks less dark on the Nvidia machine (but that could just be the HW difference in how coloring looks which we know for a long time), but if it's a HW configuration bug affecting only me and certain machines, perhaps on the unaffected machines the smoke is less dense? We'll see.

You can create short tracks file, as a test for all, thank you.

if there's a FPS drop

As I said there is not.

Update:

1. Grim Reapers on youtube also mentioned they noticed this (the head guy's PC at least)

2. Okay, it's not stereo rendering lol, it's double draw call becuase it's one smoke effect for each engine. I just didn't happen to only try it with one engine, or had the chance to test a single engine'd one.

3. The difference in smoke color that I noticed on another borrowed machine is probably due to Monitor type, it's TN, and my is IPS, that would have a huge effect on it self, so nothing to worry about there.

Reminder: Those pink lines don't mean necessairly anything's wrong or too much, it's an outline of the objects in that draw call, the area being drawn, you can see the white smoke has a lot of them too the same way black smoke does, but white smoke is around 20 times less demanding if I eye ball it, if not more.

You can see the right engine contrail smoke draws being highlighted, while the left one is not, so they're to separate effects for each engine, makes sense ofcourse, but it kinda adds to a lot of smoke saturation that overlaps/shrouded that hopefully isn't being rendered, not a problem with white smoke as it never takes too much, perhaps that's what is happening with the Black Smoke, that all of the smoke objects are rendering completely, even if obstructed? But Black Smoke is more transparent seeming and perhaps that's how it is designed to make the desired effect look right depending on various conditions and amount of smoke. The Black Engine Exhaust Smoke looks totally different in design versus the White Contrail Smoke and Smoke Generator Smoke anyway, it's not just a color change at all.

However the smoke generator on F/A-18C only sprays from the left engine, I do not know if that's by design or not.

---------------------------------

---------------------------------

As I said there is not.

Allright if you say so, ... just you know it can be tricky as eyeballing it doesn't really work to good if FPS is above 100 at all times because without screenshots I can't be sure what you guys are doing.

There is a suspicious case I've got captured where the effect would not have a FPS drop, I don't know what I did different, but I was using 1080p windowed here, but I was using it before as well, and the game was in mission-start-pause for quite some time, only when I did the active pause black hole it started dropping the FPS.

Maybe some kind of a clue that it is a bug and perhaps only happens in certain conditions, perhaps the smoke is affected by environmental conditions, or that the FPS was CPU bound and GPU had enough headroom? I'll see later but I'm gonna need a break sometime soon.

Now I'm finally going to update the drivers and see what happens, I didn't want to on purpose in the middle of testing to do as much here to have enough to compare.

Allright if you say so...

My DCS is capped at 60 F/S and it stays right there regardless of what kind of smoke I'm seeing at any given moment. It only drops when I'm flying over big cities, i.e. where there are lots of static objects in view simultaneously.

Okay .., yet we only have your word, this effect requires to be in precise position, it's not about seeing the smoke from any given location, if I look at it from the side it doesn't do much either, look that's fine, it's not that I don't believe you, but it's not that hard to take a screenshot.

Update:

Reinstalled AMD/ATI Drivers with Factory Reset (deletes all other versions, settings, device driver history) to 20.1.3 WHQL (from 19.5.2 I think) and kept at default settings as per usual, I don't play with GPU settings much in terms of gaming tuning, as I usually don't get bothered by aliasing that much, but I was using 4x MSAA in DCS.

Well there's no difference, did a quick test (now I need a break, perhaps a day or two) same Black Smoke behavior, 4xMSAA or not. The only thing there's different is just the FPS min and max bounds shift higher up, but the big drop is still there.

My DCS is capped at 60 F/S and it stays right there regardless of what kind of smoke I'm seeing at any given moment. It only drops when I'm flying over big cities, i.e. where there are lots of static objects in view simultaneously.

If your FPS is capped, it's possible you're just not aware of the issue. Your GPU could be going from 65% to 95% and you wouldn't notice unless you ran a monitoring software.

If your FPS is capped, it's possible you're just not aware of the issue. Your GPU could be going from 65% to 95% and you wouldn't notice unless you ran a monitoring software.

Exactly.