phazer11 Posted April 16, 2014 Posted April 16, 2014 The less complicated the mesh the less intensive it is to process. There is a reason SMIM causes slowdowns.
Kuldebar Posted April 16, 2014 Posted April 16, 2014 Kuldebar - I'm wondering why you opted for the medium resolution meshes, since Ethatron states he's not aware of any machine that can't handle the hiigh resolution meshes. I was waiting for some clarification, but my understanding is that the more complex the mesh, the more work your system has to shoulder. Also, I wasn't really comprehending his statement: As no current hardware I know of is triangle-setup limited the effect of the med. resolution meshes is purely psychological. If the difference is only "psychological" why use a higher res. mesh?
thommaal Posted April 16, 2014 Posted April 16, 2014 (edited) (...)I stopped trying to force LOD distance loading as a method to fight "popping" and I sure as hell never mess with uGrids. I can tell you I am getting much better (read immersive) play experience by using modest LOD settings (fade distance, etc):(...)As someone stated earlier in this thread, I'd much prefer lower quality distant views with no popping, then higher quality LOD's and crazy popping. With a balanced approach based on your PC specs, the game engine will handle things much, much more smoothly, any popping that will occur will usually be at the very edge of your reasonably set range and not at random or sporadic intervals.I also don't touch uGrids - but, so far, I have tried to reduce "popping" by adding .. well, more and more. But it leads to less immersion, not more and I am currently getting around to a similar point of view. The experience needs to be tweaked toward the way we actually use our visual sense, how we focus and what we dismiss. The information we are actually aware of while seeing is minimal compared to the data we process. Of course, that is the result of an overwhelming amount of (potential) data provided by reality all the time for free. Visual virtuality is still the opposite. But when we add as much information as possible while leaving the data out, the game should run much smoother even on midrange machines. The question should not be how we can cram more data into the stream but what data should be in there and what can be left out. Edited April 16, 2014 by thommaal
keithinhanoi Posted April 16, 2014 Posted April 16, 2014 The less complicated the mesh the less intensive it is to process. There is a reason SMIM causes slowdowns.Sure, but I don't see any chain link, intricate lantern or woven fence -shaped mountains, hills or valleys in the landscape. I think Ethatron was trying to say the perceived performance gains of medium resolution meshes is purely psychological, and maybe that was a pre-SMIM statement. Anyhow, only one way to find out - in-game testing!
Kuldebar Posted April 16, 2014 Posted April 16, 2014 (edited) LOL, always so many opportunities for "more testing" -I'd like to get a character passed 35 one day. The more triangles, even though they are simple triangles, still would add to the overhead in rendering load. This assumes that the High and Medium Res. version are significantly different in quality, of course. Mesh before: Mesh After: Also, you bring up the ENB options in game and see the polygon counts. Here we go: Medium Resolution Mesh: https://i.imgur.com/kTTCTdP.png High Resolution Mesh: https://i.imgur.com/ZkgYAuW.png Just realized, the FPS is cut off in the screen :/ but that's a transient thing anyways and not conclusive. NVM, it's there. /blind Edited April 16, 2014 by Kuldebar
Aiyen Posted April 16, 2014 Posted April 16, 2014 It is basic game design 101..... use as few vertices and triangles as possible for your models! Those are what really eat performance!If you can do something using parallax, tesselation or other texture effects rather then model effects then do that, since they cost less then the extra triangles do. Anyone who has played around with a 3d program will also know how much of a slow down you get when you start to subdivide just a few times!
keithinhanoi Posted April 17, 2014 Posted April 17, 2014 Kuldebar - another question of curiosity: Why did you decide against Soolie's HD Distant Terrain? His take is interesting because the different sets of the landscape LODs based on distance have different resolution levels. Although he doesn't provide normals or meshes, certainly Ethatron's could be "mix and matched" with Soolie's textures. Oh dear - even more testing!
Kuldebar Posted April 17, 2014 Posted April 17, 2014 (edited) HD Detailed Terrain by Soolie I don't have it active in my current game, I think I unchecked the mod when I was running a test prior to my current actual play through. My ENB sharpens textures, that combined with HD Distant Terrain made it overly sharp, I prefer it to be a little "blurrier". Also, if I recollect, I was trying to keep VRAM usage in check, I had the 50% and the full version...both seemed to peg the VRAM up a bit. I am using Detailed Terrain and Tree LoD by Rennn with Realistic Aspen Trees by AceeQ (compatibility provided by Detailed Tree LOD for Realistic Aspens by Kidhedera) I'll enable HD Detailed Terrain - 50 Percent Version and have it overwrite HQLODs - Textures Original. Edited April 17, 2014 by Kuldebar
Spock Posted May 23, 2014 Posted May 23, 2014 @Kuldebar: Have you tried fTreesMidLODSwitchDist=14000.0000? I would be interested if you like it, this is probably a more controversial setting as tree LOD switching is more noticable. How did you get your actor fade values? 7 seems fairly low at first glance, I haven't tested them though. What does fLODFadeOutMultSkyCell fade? Another question for people with ugrids 7 or 9: Does lowering the actor fade remove the problems with events triggering on cell load?
Kuldebar Posted May 23, 2014 Posted May 23, 2014 @Kuldebar: Have you tried fTreesMidLODSwitchDist=14000.0000? I would be interested if you like it, this is probably a more controversial setting as tree LOD switching is more noticable. How did you get your actor fade values? 7 seems fairly low at first glance, I haven't tested them though. What does fLODFadeOutMultSkyCell fade? Another question for people with ugrids 7 or 9: Does lowering the actor fade remove the problems with events triggering on cell load?I have tried that in the past, like pre-Sheson Fix past. I think I dispensed with it then and only briefly used it somewhat later. My operating theory is that using such settings on a graphically limited PC will slow texture loads down and ultimately cause more popping. But, I'm going to change my current fTreesMidLODSwitchDist=5000.0000 to 14000.0000 and see how it goes.
Spock Posted May 23, 2014 Posted May 23, 2014 Well that setting definitely comes with an FPS loss in foliage heavy areas compared to a setting of 5000. 14000.0000 costs almost as much FPS as setting the value past your ugrids and loading all the foliage right away. I do think that higher fTreesMidLODSwitchDist (up to ugrids borders) provides noticable visual improvements though.
Kuldebar Posted May 24, 2014 Posted May 24, 2014 Been running fTreesMidLODSwitchDist=14000.0000 since yesterday...it helps the trees be less twitchy, I some really nice LOD textures for trees and a nice foggy/misty ENB that help mitigate the pops, but the setting clearly improves things visually. I think (not 100%) that my load times are longer when I zone into the world. Also, I'm a little bit leery when I leave the Whiterun area and go to more forested climes.
Spock Posted May 26, 2014 Posted May 26, 2014 (edited) Glad you like it! I will test similar settings for ugrids 7 when I get a working GPU again. Until now I just multiplied the settings with 1.4 (7/5) which worked reasonably well. Popping is less of a problems with ugrids 7 anyway. I would also be interested if some of the downsides of ugrids 7 (events triggering on load, CPU overload) can be avoided by lowering actor distance. Edited May 26, 2014 by Spock
Kuldebar Posted May 26, 2014 Posted May 26, 2014 I would also be interested if some of the downsides of ugrids 7 (events triggering on load, CPU overload) can be avoided by lowering actor distance.I suspect that scripted events will still occur despite conservative actor distance values with a higher uGrids because scripted actor will override normal actor values....that's my prediction. Obviously, I would like to have my cake and eat it too. :p You know, there's still those people out there that post things like, "I've been runnin' uGrids 26 for perk-near 45 years and haven't had a lick of trouble with quests!" Now, I don't necessarily believe them, and also...they may not even realize what's being borked up. For me, sometimes I have to see the vanilla behavior of things to refresh my perspective on certain mods, in Skyrim it is way too easy to get tunnel vision and adjust to a "new" normal.
Kuldebar Posted June 1, 2014 Posted June 1, 2014 (edited) fSkyCellRefFadeDistance:This variable effectively controls the distance at which clouds are seen, particularly around mountain tops. A higher value will increase the amount of cloud, adding to realism and eyecandy. In combination with the above values I’ve found a value of 250,000 works in nicely. -Xtreme forum postThis seems like a nice tweak. Edited June 1, 2014 by Kuldebar
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now