FPS Issues as flight progresses

[ldts]

Ok, I'll it take into account. Thank you for your fast support.

[Denco]

Anyone here using W2XP and Europe Library. Today I ended up discovering the Europe Library is very badly optimized and will destroy your fps even without W2XP. I did a quick check and with Europe Library enabled and not running W2XP and I was getting around 25 to 30 fps in London City. Disabling the library and using the same settings I was getting 50+ fps in the same scenario. 

 

I'm not saying this solved my issue of stuttering and low fps after a while because I still haven't had the chance to properly test it but the fps gain I got back convinced me to not use the Europe Library, or even W2XP for that matter ( which I didn't use anyway I just had it installed and disabled ).

Edited July 25, 2016 by Denco

[Denco]

It doesn't really fix the issue for me. The sim in general works and looks better by disabling the W2XP but I still get fps dips to 20 to 25 on cloudy weather. Funny thing is that the fps usual starts to dip once I pass flight level 6000 feet and even with cloud distance set to only 5000. I see large cpu spikes but the gpu usage is usually lower. 

[archibit]

In my case change any options in smp cofiguration menu on any value in any way higher or lower and apply this change solve the problem on some time and fps increase to normal. It is like reseting some buffer or something like that...

[Richdem]

Just wondering if the issue of FPS loss during flights has been addressed in the new version 3.2.2?

Richard

[Cameron]

2 hours ago, Richdem said:

Just wondering if the issue of FPS loss during flights has been addressed in the new version 3.2.2?

Nope. We continue to look into this, as it's something not very widespread and that we can't yet re-produce.

[seth]

i disabled hd mesh, w2xplane , took out flywithlua + RTH  and i am still having in almost every flight this "weird " fps loss. Especially on LSGG - EGLL route ...in the middle of the flight fps drops to 8 - 15  and the only way to go back to 30s /40s/50s is to disable SMP . acts like an overloading buffer size which needs to be reset.

with Default x-plane Clouds + Fs global  real weather  everything is stable and smooth without any issue....but i am not having the beauty and amazing quality  of SMP...

 

[Cameron]

4 hours ago, seth said:

with Default x-plane Clouds + Fs global  real weather  everything is stable and smooth without any issue....but i am not having the beauty and amazing quality  of SMP...

Hang in there. Eventually we'll get to the root cause of it. Right now we're running a little blind since, again, we can't re-produce this ourselves.

[Jujunet]

Hello,

I also have this problem. I found a temporary workaround, which help. Not a solution at this problem, but a way to have a decent fps with a good rendering with SMP.

My configuration is X-plane 10.45, NOAA weather, RWC, SMP 3.2, RTH, flywithlua, and minimum plugin (planes, airport navigator, 124thATC).

When I had this issue, everything run smooth (20-30 fps) and after a flight time, about tens of minutes, i had a fps drop, about 6-15 fps. I seem to remember the issue started when I installed RWC. i had no issue with SMP alone. I started to think my rig was overloaded (old but overclocked, see below). When I see that other people have the same with a strong rig, I start to looking for a software solution. I notice that when I disable SMP everything run smooth with default clouds.

When I used SMP alone, I had only the tree layers of clouds of x-plane, i used to generate manually a global weather.

With NOAA there is more layers, and the weather is much more complicated (wind layers, rain, etc), and nonstop reload. there is a lot of parameters managed by the simu.

A thing makes me think that it's a graphic overload somewhere. I have to reduce drastically the graphic options of SMP to have a descent FPS during the whole flight. (area 4096). It's not realistic to have a group of clouds witch follow the plane ! But it's fast.

When I run with default clouds or clouds HD, it run fast too. A solution is have both !

The only workaround I had found to have the great environnement of SMP AND a really good visual of the world around AND a good performance is to run SMP together with default clouds of x-plane.

It may be a temporary solution, until the issue is resolved.

Then you have a benefits of SMP directy around the plane, and lots of clouds until horizon. Both of the clouds are mixed, and you have a nice rendering.

For this i use the lua script "Default_cloud_enhancer.lua". To enable the cloud mix, I add the line  set( "sim/operation/override/override_clouds", 0) in the script (open with notepad)

For cover all the sky I use Clouds HD v2.0 (free texture remplacement), but you can use default clouds.

You must have RTH and flywithlua to do this.

This workaround works only with version 3.2.0

For now I stay with version 3.2 with this config, until the reason of this issue will be identified.

The only disaventage I found, when you fly at dusk, you have black and white clouds ! and you see the limit area of SMP.

My SMP settings are :

cloud area covered : 4096

Puff rotation : 0.1

Cloud style : crisp

god rays : OFF

cirrus : med, texture 4

reflexion in water : OFF

sparse particles, lens flare : OFF, cloud shadow : 0.6, cloud/Terrain : 500

My graphic options are :

textures compressed, very high. fullscreen 1680*1050*32, fps unlocked.

trees : invaded, objects and road : tons, cars : winter. objets distance : high. airport details : default, shadows : global (high), reflexion : complete. only one box checked : "the runways..."

in the special effect box : all checked, 2x SSAA+FXAA. FA : 16x.

Clouds details : 10%

Rig : Core2 E8600 overclocked at 4.5 GHz, 6GB RAM PC8500, SSD, GTX 970 4GB.

I have 2 AI planes "XPFW" (lightened)

 

I salute the work of Sundog for this great plugin I use since version 2.0. I know that it's a really hard job to found an erratic bug like that. I hope you'll identify soon the solution of this issue, which seems to be hard to found. There is a lot of interactions parameters stake for a while (new xplane updates, new sofware RWC with interactions with many weather infos plugin, etc) and can come from anywhere. Hoping find again a total cover area with last version of SMP. Good luck.

 

I'm sorry for my English, i'm from LFOT ! :-)

 

have a good day!

[PeterR]

OK, some people are experiencing this, others are not. Some people seem not to be even VRAM constrained (I would focus on these people). As it is not easy to figure this out, it might be a worth idea to extend the discussion to new area - differences in environment people are operating in. For example, I do not recall anyone with ATI card reported this. Is anyone with Windows 10 experiencing this? Are non-default NVIDIA Control Panel 3d settings in place? Do you have other apps open at the same time as X-Plane?

Although I have the same problem, I do not qualify as good test subject right now as I have only 2GB of VRAM. GTX 1070 is expected to arrive next week, that will qualify me as good test subject. Other notable facts on my end are: sufficient CPU, 32 GB RAM, NVIDIA card, Windows 8.1 (I have suspicion this may be the key), no non-default NVIDIA Control Panel settings, X-plane 10.45, plugins are subjective, maybe X-Camera is worth mentioning, scenery/mesh/w2xp packages although resource intensive I do not consider relevant to report as RAM and VRAM meters are more important. When I sim I run Chrome and Navigraph Charts at the same time (these are quite VRAM heavy by the way).

I am looking forward to report back once new GPU installed. In the meantime I wish you luck in identifying the root cause.

[PeterR]

I've now done very first test with new GTX 1070. Great card by the way and I learned 2 things - first you can push everything much further and second you still cannot push enough it to maximum values. If you need push everything to max, probably wait for the cards that will be introduced in 2025 (if you will be still on X-Plane 10) :-).

As regarding to FPS drop during the flight, I still noticed it. Not to single digit numbers like with GTX 960, but to numbers below 20. SMP was still reporting more than 4GB VRAM and 20GB RAM available. Quick cycling of SMP settings (for example turn cloud reflections on water to OFF and back ON) resolved it immediately. However what I noticed this time was interesting - overall CPU utilization was around 25%, but when FPS dropped, one of the CPU thread (1 of 8) was saturated to almost 100% while others were underutilized. Cycling SMP settings reduced that CPU thread utilization back to 70s.

So maybe CPU is bottleneck or CPU optimization is not good enough. Not sure how much can be done here since X-Plane is limited because of its OpenGL engine.

I am still keeping an eye on it.

[sundog]

Appreciate the data. We are still working to track down what's going on and trying out some ideas.

[PeterR]

Another flight today and some new observation. GPU utilization never crossed maximum of 59%, VRAM utilization maxed at 74%, GPU temperature maxed at 63 °C (it starts throttling down only at  70 °C and above so no issue here). Aside from occasional high usage of single CPU core I didn't notice any high resources utilization. When FPS dropped I noticed that after cycling SMP settings to restore FPS, GPU utilization actually increased. It seemed almost like SMP or X-plane turned handbrake or something to slow the sim. To me it looks less and less like resources bottleneck in the end.

This reminds me something I noticed when I first started using SMP v3 (this was my first version), which may be completely unrelated, but I will mention it anyway. On X-plane without SMP, Zulu time and clock run at same rate - when FPS dropped, both clock slowed down. With SMP/RWC when FPS drops, in plane clocks slow down, but Zulu time continues to run on correct speed. Now I didn't check for this behavior for long time to confirm, but might be worth if others could look, too. Just throwing an idea here, unsynchronized clocks may not be good. I will certainly be looking at this again although I might be not able to see it as I no longer have such huge FPS drops.

[Tom Stian]

I also have perfomance issues with SkyMaxx Pro with RWC. So after using default clouds for a while I decided yesterday to reinstall SkyMaxx Pro 3.2.2. I did not have any issues with default clouds.

But after 2h I got a big FPS drop again. Reloading SkyMaxx solved this instantly. (by going to settngs and just adjust the cload area coverage by 1 sq km or so and apply settings)

I dont remember exact how much VRAM I had left. But it was at least +1GB. Having GTX980TI 6GB.

Sadly I have to disable SkyMaxx again until this is solved. But let me know if there is anything I should test etc.

[PeterR]

I did some more test flights. All of them started with draw area set to around 14 000 sq km.

Flight 1 - all plugins (except default XP plugins, SMP, RWC, Gizmo) were moved out of plugins folder. FPS drop shortly after takeoff. Cycling settings was only very short-term solution and in order to finish flight I had to reduce draw area to about 3 000 sq km.

Flight 2 - all plugins moved out of plugins folder as in flight 1 + HD Mesh V3 was disabled. Although I experienced some FPS drop, reducing draw area to 10 000 helped and I was even able to return back to 14 000 without any more FPS drops.

Flight 3 - same as flight 2, but in this case no single FPS drop and draw area was set to 14 000 all the time. Might have been influenced slightly by XP weather data problem mentioned elsewhere, but I had weather and clouds along the way.

Resource utilization seemed to be similar across all 3 flights. Only difference worth mentioning is that GPU peak utilization was actually highest on smoothest flight 3 (73%) and VRAM usage was lowest on flight 3 (51%). Seems that plugins (at least the ones I use) do not have effect on performance. However, what does have effect is HD Mesh v3. Seems that HD Mesh and SMP don't like to run together. My assumption is that HD Mesh is very memory hungry, but if memory is not a boggleneck, it should not require too much computing resources.

I have following plan for next tests:

- HD Mesh and plugins enabled but XP texture resolution slightly lowered and SMP terrain softness reduced to 0.

- Some tests with XP 10.50 when release.

- Some tests with Windows 10

[sundog]

Peter, be sure to try SkyMaxx Pro 3.3 once it's released as well. I think it might help people who are having this issue.

No ETA on its release yet, but keep an eye on your email for an announcement.

 

[frontendrob]

13 hours ago, sundog said:

Peter, be sure to try SkyMaxx Pro 3.3 once it's released as well. I think it might help people who are having this issue.

No ETA on its release yet, but keep an eye on your email for an announcement.

 

Will 3.3 be the next version, or do you plan a quick patch for the NOAA issue before that?

[sundog]

35 minutes ago, frontendrob said:

Will 3.3 be the next version, or do you plan a quick patch for the NOAA issue before that?

They'll be one and the same. We had 3.3 ready to go before the NWS/NOAA server issue arose.

[frontendrob]

Sounds good Frank, I am F5-ing my inbox already 

[Muskoka]

Still have the fps drop issue, SMP and RWC both updated to the latest, as well as XP, the latest beta. Was watching the fps on a flight yesterday and within seconds the cpu and gpu "time" rocketed from their normal's of around 0.0200, and of course the fps plummeted from a locked 30 fps to 9-10. This time I didn't disable either plugin to fix the issue, I just changed one setting in SMP and clicked apply, and all was back to normal, and my usual locked 30 fps.

Keep working at it guys, hope you soon find the issue.

[sundog]

Argh! I really thought SMP 3.3 would help with this.

This issue is basically personal now. So I spent the whole day determined to track it down.

We're dealing with an incoming tropical storm here in Florida right now, which has the benefit of providing lots of nasty weather for me to fly in and test with. Sure enough, at last I think I can reproduce what you're seeing - performance that suddenly tanks and is mysteriously fixed by adjusting the SMP configuration, however slightly.

The issue seems to have to do with cloud backdrops - the ring of 2D textures that represents distant clouds near the horizon. SMP is getting into a state where it thinks these backdrops need to be recomputed every frame, and that's the problem. It requires a rather odd set of circumstances, but if you fly long enough with a large enough cloud draw area, you can get into this state.

I believe I've fixed the underlying bug, and this fix will go out with our next update.

While I was tracking this down, I also found that due to duplicate entries in the METAR data we're getting from NOAA/NWS, we were creating way more thunderstorms than we should. I've also fixed that, which will help performance further in stormy weather.

Thanks for working with us on this, guys. I'm hopeful the next update will finally fix this. (Before you ask, no I don't know when that will be yet!)

 

[Muskoka]

1 hour ago, sundog said:

Argh! I really thought SMP 3.3 would help with this.

This issue is basically personal now. So I spent the whole day determined to track it down.

We're dealing with an incoming tropical storm here in Florida right now, which has the benefit of providing lots of nasty weather for me to fly in and test with. Sure enough, at last I think I can reproduce what you're seeing - performance that suddenly tanks and is mysteriously fixed by adjusting the SMP configuration, however slightly.

The issue seems to have to do with cloud backdrops - the ring of 2D textures that represents distant clouds near the horizon. SMP is getting into a state where it thinks these backdrops need to be recomputed every frame, and that's the problem. It requires a rather odd set of circumstances, but if you fly long enough with a large enough cloud draw area, you can get into this state.

I believe I've fixed the underlying bug, and this fix will go out with our next update.

While I was tracking this down, I also found that due to duplicate entries in the METAR data we're getting from NOAA/NWS, we were creating way more thunderstorms than we should. I've also fixed that, which will help performance further in stormy weather.

Thanks for working with us on this, guys. I'm hopeful the next update will finally fix this. (Before you ask, no I don't know when that will be yet!)

 

Thank you Frank for sticking with this. I did another flight this morning, same flight, CYYZ-CYUL and had no issues at all. It is a strange one. The weather was a "bit" cloudier with the problem flight yesterday, but nothing drastically different weather wise. Today's flight was lightly scattered cloud coverage the whole route, and everything looked, and performed great.

Edited August 29, 2016 by Muskoka

[Denco]

2 hours ago, sundog said:

The issue seems to have to do with cloud backdrops - the ring of 2D textures that represents distant clouds near the horizon. SMP is getting into a state where it thinks these backdrops need to be recomputed every frame, and that's the problem. It requires a rather odd set of circumstances, but if you fly long enough with a large enough cloud draw area, you can get into this state.

 

 

This is probably the best explanation of what I'm experiencing. Bellow in the video you'll see the low fps and stutters I sometimes face when using Skymaxxx. The clouds draw distance was set to only 10000 and the fps were bellow 20 for the most of the flight. I did a flight with The Bonanza V35 in Germany and the skies were filled with a lot  more clouds then on the video bellow and with clouds draw distance set to 30000 I experienced a very smooth flight.

Now I know that high fidelity planes like 767, the one I fly the most, tax the system a lot more than the Bonanza, but if I use stock clouds with a custom script I rarely dip bellow 30fps.

 

Edited August 29, 2016 by Denco

[PeterR]

I did several more flights, all without FPS drops. First was, as I promised, with HD Mesh re-enabled and SMP terrain softness reduced to 0. The flight was smooth, so I was able to revert it to default 500. This was however with SMP 3.2.2, so with the weather data broken (although clouds were present along the route).

The most recent flight was with SMP 3.3 and updated RWC set to Always so I get real cloud coverage at least. This flight was also very smooth. On the other hand weather conditions were good and there were not many clouds on the way.

At this point I cannot conclude if SMP 3.3 fixed performance issues I experienced, but it may very well had. I need to perform additional flight in more severe weather. I hope I will be able to do it today or tomorrow, as I will be leaving for 2 weeks afterwards. I will report back as soon as the flight is completed.

[sundog]

Denco - have you tried the new SMP 3.3 release? Your issue sounds like you just have more add-ons installed and/or high rendering settings or resolution than even your very beefy system can handle, but SMP 3.3 uses fewer resources which may help.