Jump to content


Photo

ENB Guide

guide enb

  • Please log in to reply
480 replies to this topic

#226 Nearox

Nearox

    High King

  • Contributors
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 1,396 posts

Posted 28 January 2014 - 06:21 AM

Note that putting UsePatchSpeedhackWithoutGraphics=false and UseEffect=false does not enable parallax effects, regardless of whether FixParallaxBugs= set to false or true. The useffect=true parameter is necessary to enable this effect, as well as a few others in enblocal.ini. Not sure why these settings got transferred over to enblocal.ini some time ago as they need enbseries.ini useffect in order to be activated, but alas. Even if you disable all feature individually in enbseries.ini and only leave the parallax effect enabled in enblocal.ini, there is still a 5-10fps performance impact depending on the ENB version used. Earlier versions like 0.221 have less performance impact than e.g. 0.246. Some effects are always enabled when UseEffect is set to true.  Edit: Anyone know why the config tool no longer recommends to set DisableDriverMemoryManager=true for AMD users?
  • 0

#227 Aiyen

Aiyen

    Dragon King

  • Super Moderators
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 3,536 posts

Posted 28 January 2014 - 08:26 AM

Depending on preset there is also a difference in the enbeffect.fx and other .fx files... in general the more massive they are the more performance they eat. And they are loaded even if their effects are not activated I think... would have to test that at some point. As I recall then the last test I made of v.168 and the later ones only showed a difference of about 1-4FPS when all was disabled properly. which is a decent and really unnoticeable overhead. You would have to go all the way back to pre v.119 to get like 1 FPS more I think it was.... so if you have an overhead of 5-10 FPS then something is not setup properly or perhaps it is the above.. I have not tested that but would not be surprised if it was the case. As for the parallax then obviously ENB needs to be active in order for it to implement its parallax fix. This is pretty obvious I would think. If you say that the useeffect is what is required then this would explain the main difference between the two I think.. or at least one of the main differences at any rate.

#228 hellanios

hellanios

    Commander

  • Members
  • PipPipPipPip
  • 339 posts

Posted 28 January 2014 - 08:29 AM

Aiyen could you upload a set of enblocal and enbseries with ONLY the parallax effect active?
  • 0

#229 Nearox

Nearox

    High King

  • Contributors
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 1,396 posts

Posted 28 January 2014 - 08:45 AM

If I remember correctly, when it was announced that the enbseries.ini fixes were moved to enblocal.ini it was primarily so they could be enabled without having to use the effects of enbseries.ini. Also remember some discussion between PPR author and Boris @ enbdev where Boris basically said it's not possible to single out parallax effects from other effects. EDIT: Boris said

UseDefferedRendering=false will help. In enbseries.ini don't forget to turn off shadow changes and UseOriginalPostProcessing=true, UseOriginalObjectsProcessing=true, then performance will be high enough. Unfortunately i don't know yet how to turn on only parallax fix without mod itself, because recompilation of shaders required with dublicating of them and capturing of matrices from other objects, perhaps possible, but i just need to find idea

Link

Believe me, I have fiddled with ENB/ENBoost for many months trying to get rid of the overhead for parallax. Think we even discussed this or something similar in a previous thread some time ago :)

No doubt you have much more experience with ENB as I do. Nevertheless, regardless of using stock ENB, any other ENB, putting literally every setting to 'false' in enbseries.ini and having tried various combinations of false and true: on my system it costs between 5 and 10 FPS with the newer ENBs no matter what I do. Going from 58-60fps in most areas to approximately 50 fps. Results differ depending on hardware of course. I'm pretty sure by now that I have tried everything save from editing numerical value parameters in enbseries.ini as I have no clue what they do and doubt they are linked to it.

This has led me simply opt for realvision performance version, as it comes with only a relatively slightly higher fps cost (approx 15) compared to only using parallax and I get more quality in return than using enboost + parallax + sweetfx.
  • 0

#230 Aiyen

Aiyen

    Dragon King

  • Super Moderators
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 3,536 posts

Posted 28 January 2014 - 09:19 AM

Based on what Nearox says then in order to get parallax only then it should be enough to have enableeffect on and use original post processing, and then parallax... ofc. then the graphical mode needs to be enabled in enblocal.ini as well. Also when I test for overhead I do not run about, I simply stand in riverwood and look in the same direction and do a small screen pan to get the effective overhead. Running around causes too many variables for a reliable read, since everything from shadows, grass etc. to FoV will matter as well.

#231 TechAngel85

TechAngel85

    Akatosh

  • Administrators
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 12,626 posts

Posted 28 January 2014 - 12:38 PM

A note on AF... Quite some time ago I did comparisons on the AF between game, driver and ENB. The overall winner was driver AF (for nvidia at least) by the slightest margin. So unless Boris improved ENB AF since around release v.119, I would assume that drive AF is still the best. The difference performance wise was the same and the difference quality wise was only noticeable at the very farthest distances; however, with the memory patch and people upping the uGrids this might be more important than it used to be for those looking for the very best quality.

#232 hellanios

hellanios

    Commander

  • Members
  • PipPipPipPip
  • 339 posts

Posted 28 January 2014 - 01:22 PM

What do DisableDriverMemoryManager and DisablePreloadToVRAM mean? I remember that these options were recommended to be set to true. Now the utility recommends to be set to false. I have an old but trustworthy ATI 4890 1GB VRAM.
  • 0

#233 EssArrBee

EssArrBee

    Incompatibilism Manager

  • STEP Staff
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 7,721 posts

Posted 28 January 2014 - 01:27 PM

Boris changed the [MEMORY] section recently so the utility might need updating. I'll ask Phinix to take a look at it.

#234 Aiyen

Aiyen

    Dragon King

  • Super Moderators
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 3,536 posts

Posted 28 January 2014 - 01:28 PM

AF really does depend on which textures you use, since it kinda affects those. So what is best for a STEP setup and what is best if you use fancy 2k stuff from various authors make a difference.

#235 z929669

z929669

    Ixian Inventor

  • Administrators
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 9,314 posts

Posted 29 January 2014 - 12:54 AM

A note on AF...

Quite some time ago I did comparisons on the AF between game, driver and ENB. The overall winner was driver AF (for nvidia at least) by the slightest margin. So unless Boris improved ENB AF since around release v.119, I would assume that drive AF is still the best. The difference performance wise was the same and the difference quality wise was only noticeable at the very farthest distances; however, with the memory patch and people upping the uGrids this might be more important than it used to be for those looking for the very best quality.

Driver AF (AMD here) yields the same exact result as ENB AF for me, but the former gives me a 3-4 FPS gain, so I suspect that ENB AF is more costly without any gain (this is for you EssArrBee!).

Also, it doesn't make sense to user FPS limiter with vsync (Ref 1, Ref 2). And there is no reason at all to set iPresentInterval=1 in skyrimprefs.ini, as it does not work nearly as nicely as the ENB versions, which I suspect also includes D3D-compatible triple buffering, since it is soooo smooth.

iPresentInterval should always be set to zero in favor of ENB version:
EnableVSync=true
VSyncSkipNumFrames=0

... I have examined this in much detail, and find that ENB's version of vsync is the best around (at least for my card, but I would guess it is a good bet for anyone). This is particularly helpful indoors when frame rates are potentially high.

I will be playing around with RadeonPro as well to see how that does, as it has D3D-capable triple buffering.

#236 Nearox

Nearox

    High King

  • Contributors
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 1,396 posts

Posted 29 January 2014 - 04:50 PM

With temporalAA=enabled, for the first time I am experiencing what is apparently called "ghosting". Pretty annoying... Damn and here I thought I finally found a lowcost solution to this kind of AA. Also Boris said in beginning of december that it is impossible to fix  the ghosting:

I've failed with changing temporal antialiasing code in Skyrim to remove ghosting artifacts, just wasted a lot of time, this is technically impossible to fix because of external shaders. Everybody who waited of TAA update - forget about it.


DAMMIT :P
  • 0

#237 EssArrBee

EssArrBee

    Incompatibilism Manager

  • STEP Staff
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 7,721 posts

Posted 29 January 2014 - 05:00 PM

You could switch to EdgeAA + SubPixelAA + SMAA injector.

#238 Nearox

Nearox

    High King

  • Contributors
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 1,396 posts

Posted 29 January 2014 - 05:11 PM

Aye but I happen to have a 5 year old screen with a bad DPI ratio (22 inch 1680x1050 resolution) and the shimmering on flora in particular is pretty bad if I don't use temporal AA or SSAA. Seeingas the latter is no option due to performance, I may indeed have to go for the combination you mentioned. However doesn't SMAA do the same as edge AA, but a bit better?
  • 0

#239 EssArrBee

EssArrBee

    Incompatibilism Manager

  • STEP Staff
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 7,721 posts

Posted 29 January 2014 - 05:18 PM

No edge and temporal seem to be quite similar to me but smaa has bad shimmering problems and temporal or edge seemed to help a bit. It's bad on my 1366x768 laptop so your 1080 screen it would be even worse. You might see if you have the power for transparency by itself. Sent from my iPhone using Tapatalk

#240 Nearox

Nearox

    High King

  • Contributors
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 1,396 posts

Posted 29 January 2014 - 05:34 PM

I can run hardware transaprency AA at 4x (it doesnt cost much more then MSAA) but it brings with it annoying artifacts on certain textures. Depending on viewing angle, some textures switch to 100% transparency such as the moss hanging on the walls in solitude, a few trees of SFO, 1 of the 2 leaf-layers on the wooden planks in riften etc. It's all small stuff but pretty annoying when the textures keep switching to full transparency mode whenever moving.  Vurt once fixed a couple textures (I think it was meshes he fixed) for transparency AA after I put in a request, but since 1.91 the artifcats are back and I don't want to take his time from developing sfo 2.0 :P I don't know how he fixed it, but I believe he said then that it was pretty easy. If there was a way for the handful of vanilla textures to be fixed, that would be great... Keep in mind this is with AMD's Adaptiva Anti-Aliasing. I don't know if Nvidia's Transparency AA has these issues as well.
  • 0



Also tagged with one or more of these keywords: guide, enb

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users