Great looking videos Rob

 

I fully agree with you. Very nice!

 

Patric

clouds and anti aliasing= FPS buster

my only fps hit in the sim

this is a quote direct from the p3d forum, which I think is hitting the nail. Hope LM will response to this. Right now fps are in high teens and in 2.2 they were perfect during overcast;

 

Had a quick look at the shader.
Oh boy... everything became much more complicated, the simple rotation matrix based on the pbh is gone, instead two new textures coordinates are monitored all the time, etc.

No wonder... with dense clouds and cloud shadows turned OFF, I get a steady 60 fps with the 2.2 shader, but with the 2.3 shader (even using Beau's uncomment solution) the fps becomes 35-38.
No need for complicated or further tests, that is a clear 40% performance loss, due to the cloud and shader changes in 2.3.

Thanks Word Not Allowed for the tip and for the pseudo-solution, that at least clearly proved that the shader and the changes behind are the culprit that cause the significant performance hit.
I have no idea who does thorough tests and beta-testing for P3D, and how they do it, but I can hardly believe that such a clear, obvious and important difference was not noticed by them.
(I don't want to get nostalgic here, but in the good old FS9&FSX days it would have been almost impossible that we, the beta-testers, wouldn't have noticed such a difference...)

I really hope LM comes up with at least a single corrected shader or hotfix, and we do not have to wait for that till 2.4, which is probably months ahead.

A last thought and sentence:
I strongly believe the 'volume increase' or adding 'noise' to the clouds is absolutely not needed and necessary.
Instead LM should design or even buy at least a single set of clouds out of the many available out there (there are very nice ones even freeware), and that in itself would solve the problem for those, who still use the elderly default clouds, which indeed lack the volumetric effect, but that can be done and corrected by using better textures, instead of trying to make complicated and structural changes, which led to this very significant performance problem.
A single and nicely done cumulus01.bmp file can do that, and the other experiments which result in such a performance hit, are absolutely ill fated...

Lapi

I wonder if I am imagining things but even though I didn't have performance problems with clouds, I decided to use that fix anyway: I don't really need a feature no one seems to notice and which MAY hurt performance.

After using the fix I, as expected, didn't see or notice a difference when it comes to performance but I THINK I noticed something else: the rotating clouds are less annyoing when flying in the clouds or through (slight) fog. Maybe this is due to the fact that the new cloud options adds another 'noise' texture, which, I guess, means there are more textures in view at any given time and hence more textures rotating when looking around. Disabling the feature removed that (or those) noise texture(s) and hence the amount of clouds that rotate when looking around has become less. If you know what I mean.

 

Anyway, during my last two flights it occured to me that rotating clouds annoyed me less (they mainly annoy me when flying through clouds and foggy situations) and I wondered if anyone else is noticing the same. It's as if a few textures (which were rather transparent with touches of very light grey in them and which always were, if they were in view,  pretty close) have disappeared now. It could be a positive side effect of this fix which may be interesting for those who don't have problems with performance but who don't like those rotating clouds.

 

Of course I could be imagining things or it might have to do with today's weather (using ASN here) but it may be worth investigating.

I found some information on the LM forum (Author: Word Not Allowed) that indicated if one replaced the P3DV2.3 cloud.fx file with the P3DV2.2 cloud.fx file that the performance of V2.2 would be recovered.  I happened to have a backup of my P3DV2.2, so, I renamed the V2.3 cloud.fx file to cloud.fx.org (ShadersHLSL folder in main P3D directory) and placed a copy of the V2.2 cloud.fx file into V2.3.  I then cleared the Shader folder so it would rebuild upon startup.

 

Bing-Go....Performance within heavy cloud cover was restored.  I haven't noticed any unwelcome side-effects as yet.  Flew for about a half hour in heavy clouds and rain and maintained solid performance.  Did see a small decrease in FPS periodically, but, nothing out of the normal, and certainly nothing significant.  I believe this is a reasonable work-around until LM issues a revision.

 

Hope this helps.

I found some information on the LM forum (Author: Word Not Allowed) that indicated if one replaced the P3DV2.3 cloud.fx file with the P3DV2.2 cloud.fx file that the performance of V2.2 would be recovered.  

 

I believe it also says that cloud shadows will be broken if you use this. Just a word of caution for those wanting to try.

I think it is safer to use the fix that was suggested by LM itself than to do something drastic as adviced by another user. If you are going to use files from a previous version, you never know what problems that may cause. These files tend to work together and changing one make break something else.

I think it is safer to use the fix that was suggested by LM itself than to do something drastic as adviced by another user. If you are going to use files from a previous version, you never know what problems that may cause. These files tend to work together and changing one make break something else.

Jeroen, I am not sure but the //volumizing trick is not the same as replacing the cloud.fx file.

I can only agree with Jeroen. I also think that's not a good idea to mix files from two different versions. For the moment I use the LM fix, too.

 

Patric

Jeroen, I am not sure but the //volumizing trick is not the same as replacing the cloud.fx file.

 

I didn't say (or meant to say) it was the same. By using that trick you selectively disable one new feature, which is okay, specially because LM says it is an option, while by replacing the entire file you are 'disabling' all new features and 'reenabling' all old features without knowing if everything will be compatible with other (new) files and (new) features.

I didn't say (or meant to say) it was the same. By using that trick you selectively disable one new feature, which is okay, specially because LM says it is an option, while by replacing the entire file you are 'disabling' all new features and 'reenabling' all old features without knowing if everything will be compatible with other (new) files and (new) features.

 

But what if that's the only way to restore performance to previous levels? Of course nobody should have to do something like this, which is why I am hoping that 2.4 will fix this cloud performance problem.

But what if that's the only way to restore performance to previous levels? Of course nobody should have to do something like this, which is why I am hoping that 2.4 will fix this cloud performance problem.

 

 

I didn't say (or meant to say) it was the same. By using that trick you selectively disable one new feature, which is okay, specially because LM says it is an option, while by replacing the entire file you are 'disabling' all new features and 'reenabling' all old features without knowing if everything will be compatible with other (new) files and (new) features.

I did the trick with //volumize, but it's not helping me, waiting for 2.4 is not an option because it would ground me for weeks / months????. I will have to use the 2.2 fx file to try to get better resluts. 

 

Would someone share this file with me?

 

 

But what if that's the only way to restore performance to previous levels?

 

 

 

I did the trick with //volumize, but it's not helping me, waiting for 2.4 is not an option because it would ground me for weeks / months????.

 

I am not saying you aren't allowed to do it LOL! But if you do it, don't expect support from LM if something goes totally wrong. In general it simply is wise to not mix files from various versions with ANY program but if it works for you, be my guest, of course! ^_^

Hi Jeroen - I noticed that the spinning cloud effect seemed less pronounced as well (still there, but mainly in the periphery).

 

Again, I don't see the catastrophic drop in frames like others in this thread, however I don't use Nvidia inspector, FSUIPC, or ASN. Just OPUS, REX4 textures at 1024 / DXT5, and cloud density at Medium (OPUS still fills the sky in full overcast, without a major hit). I also tried disabling the volumize effect, but it didn't have any real impact on frames.

 

I remember we used to have this issue in FSX DX10, but Steve partially fixed it (though even with his option to ignore AA on clouds, there was still a significant hit in clouds...).

Hi Jeroen - I noticed that the spinning cloud effect seemed less pronounced as well (still there, but mainly in the periphery).

 

Ah, nice to know I am not the only one noticing this! It's indeed still there but you have to look harder to see it, so it seems.

So those who have great performance and don't need the fix for that but who are annoyed by the spinning clouds: you might want to give this fix a try anyway.