Answers probably a "No, not possible" but I thought I'd ask anyway.
Since clouds themselves are not huge FPS "hogs" to render, is it possible to tell the sim to just render all the clouds - including those not currently in view - all the time so that they don't pop in and out of existance when they get partially out of view?
Answers probably a "No, not possible" but I thought I'd ask anyway.
Since clouds themselves are not huge FPS "hogs" to render, is it possible to tell the sim to just render all the clouds - including those not currently in view - all the time so that they don't pop in and out of existance when they get partially out of view?
Oh, how I have wanted that issue resolved as well! Any cloud rendering guru's listening? Maybe we have to wait for DX11 porting, if that is possible.
Last edited by Robert_Wiggins; 07/10/1809:49 PM.
(System_Specs)
Case: Cooler Master Storm Trooper PSU: Ultra X3,1000-Watt MB: Asus Maximus VI Extreme Mem: Corsair Vengeance (2x 8GB), PC3-12800, DDR3-1600MHz, Unbuffered CPU: Intel i7-4770K, OC to 4.427Ghz CPU Cooler: Cooler Master Seidon 240M Liquid CPU Cooler Vid Card: ASUS GTX 980Ti STRIX 6GB OS and Games on separate: Samsung 840 Series 250GB SSD Monitor: Primary ASUS PG27AQ 4k; Secondary Samsung SyncMaster BX2450L Periphs: MS Sidewinder FFB2 Pro, TrackIR 4
The basic game engine has some cloud related hard coding in it that gets in the way of doing this in any simple way. Perhaps if AnKor can drop the clouds completely, and go with all new volumetric DX11 clouds then...
I have never found a magic bullet to resolve this issue. At present, the only way to reduce cloud popping is to avoid using certain cloud types, or reduce the overall number of clouds used for the different weather conditions. Unfortunately, both result in a less pleasing and less realistic set of clouds--unless you like mostly cloudless skies all the time.
“With Major Lawrence, mercy is a passion. With me it is merely good manners. You may judge which motive is the more reliable.”
Personally, I think that straight cloud line thats moving from side to side is a far more an game breaker. Think that most not that difficult to stop this from moving.
Interesting idea. Clouds are not exactly free, but it is indeed not a big deal to render more of them. My DX9 shaders keep track of visible clouds to render clouds shadows. And even after a cloud disappears from view the shaders still remember it for a few minutes. The problem is that this approach is not very suitable for on screen rendering - clouds are flat sprites and should be re-aligned depending on camera view. For cloud shadows I just keep them aligned to the sun, but for on screen clouds I would need a more sophisticated algorithm to avoid constantly rotating clouds (which would look silly), but still re-align occasionally to avoid seeing "flat" side of a cloud. I tried to find what part of cfs3 code is responsible for cloud visibility, but without success yet.
Originally Posted by MajorMagee
Perhaps if AnKor can drop the clouds completely, and go with all new volumetric DX11 clouds then...
That's the goal, but it is really difficult. I more or less know how to make volumetric clouds work, but I'm currently struggling to make them work fast.
Here are some very early cloud tests from January (still based on DX9 by the way). Not really good looking, but just a proof of concept.
AnKor, If you are well on your way to 3D volumetric clouds as shown by your proof of concept pictures, I heartily endorse moving in that direction!
If it turns out 2D clouds will be remaining, is the following of any use?
If the shaders remember a cloud for a few minutes after the cloud is removed from screen view, then that should be more than enough time to keep rendering it until the cloud is actually 100% out of view of the the screen. The popping effect seems to be happening when the the rendering engine thinks the cloud is out of view of the screen but is in fact only partially out of view. Continuing to render for as long as it is "remembered" would essentially mean that the clouding popping would happen off screen, out of the players view.
As far as keeping the clouds from rotating, is it possible that the same algorithm that currently decides if a cloud is out of view of the screen could then use a simple fixed "last known position" coordinate. Given the relatively slow speeds of the aircraft and the fact that the cloud is already partially out of the view screen, the rotating effect would seem likely to be minimal (and in any case still less jarring than clouds popping in and out of view on screen) before the cloud was indeed actually totally off the screen and out of view of the players.
Thanks for chiming in here and voicing your experiences. You have given me hope for great things to come.
Your effort and the time you take to explain things are greatly appreciated.
Best Regards
(System_Specs)
Case: Cooler Master Storm Trooper PSU: Ultra X3,1000-Watt MB: Asus Maximus VI Extreme Mem: Corsair Vengeance (2x 8GB), PC3-12800, DDR3-1600MHz, Unbuffered CPU: Intel i7-4770K, OC to 4.427Ghz CPU Cooler: Cooler Master Seidon 240M Liquid CPU Cooler Vid Card: ASUS GTX 980Ti STRIX 6GB OS and Games on separate: Samsung 840 Series 250GB SSD Monitor: Primary ASUS PG27AQ 4k; Secondary Samsung SyncMaster BX2450L Periphs: MS Sidewinder FFB2 Pro, TrackIR 4
Didn't someone once say that using 2 x Nvidia cards in SLI mode cured the cloud popping? If true, that suggests the CFS3 hard coding may be able to take advantage specifically of SLI... Ankor, do you know if this could be the case?
Didn't someone once say that using 2 x Nvidia cards in SLI mode cured the cloud popping? If true, that suggests the CFS3 hard coding may be able to take advantage specifically of SLI... Ankor, do you know if this could be the case?
I want to say that was RAFLouvert who found that to be true on his system. Apologies if I have mis-identified the individual.
I searched and found this (it was indeed RAF Louvert who posted this)
OK, so here is what I've discovered at this point.
With both cards and the HB SLI bridge installed again the cloud popping has gone away. Swapping out the HB SLI bridge with a single soft bridge yields occasional cloud popping though from what I can see it does not appear as prevalent as it was with a single card. Installing two soft bridges, thus utilizing both SLI ports on each card, seems to cut down on cloud popping frequency somewhat further but does not eliminate it entirely, (though to be fair I am not 100% sure about this as it starts to get rather subjective after watching clouds pop in and out for three hours and trying to determine if the frequency is now significantly different than it was earlier). Removing the soft bridges and reinstalling the HB SLI bridge gets me back to where I was in the first place, i.e., no cloud popping. I tested each combination flying QC from St. Omer in October of 1918 with the DH.4 and tried the different cloud options available. There was one instance with the HB SLI bridge, on the large fluffy cloud setting, where one particular cloud well out in front of me would pop in and out as I moved my head slightly in TrackIR. However, when I went out of QC and came back and tested it again with the same settings it was no longer an issue.
I'm not quite sure what to make of this. I am wondering if the HB SLI bridge is somehow allowing data to share between the two cards in such a way that it just happens, out of pure dumb luck, to fix this age old CFS3 issue. Perhaps as one card starts to lose track of the cloud rendering along the screen edges the other card picks up the data stream. I honestly don't know. I also honestly don't know if the issue will stay away as time goes by. Will a Windows update and/or a driver update suddenly make it reappear? It could. Since I don't know why this has currently fixed the issue there is no way of knowing what might "break" it again at a later date.
Full thread here.... http://SimHQ.com/forum/ubbthreads.php/ubb/showflat/Number/4300768/Searchpage/3/Main/408962/Words/%2Bcloud+%2Bpopping/Search/true/re-scored-another-evga-gtx-970-ssc-card-and-am-going-back-to-sli#Post4300768