Highest possible graphics?

[Rick]

Been playing around with cvars and JA++ all day.. trying to get my graphics just right before I get into the game itself again.

 

r_textureMode GL_LINEAR_MIPMAP_LINEAR

r_simpleMipMaps 0 

r_textureBits 32

r_stencilBits 24

r_depthBits 24

r_detailTextures 1

r_lodBias 0

r_ext_compress_textures 0

r_picmip 0

r_subdivisions 80

r_ext_texture_filter_anisotropic 16

 

My biggest query is on r_subdivisions and r_simplemipmaps. Some places say having r_subdivisions at 80 and r_simplemipmaps at 0 delivers better graphics, while others say it's in the reverse, where r_subdivisions at 4 and r_simplemipmaps make the game look smoother. To my understanding, simplemipmaps at 0 goes well with having anisotropic filter on max since it essentially loads the entirety of the view on the map, while simplemipmaps being activated (set at 1) makes the distance looks somewhat blurry, though at the same time improves performance. r_subdivisions is the map detail I think?

 

I think I have the rest of the cvars I posted above already on max, not entirely sure though.

[Rick]

Alright, I figured out that r_subdivisions 1 is basically max map detail, still not sure on the difference between simplemipmaps, but I have a question about r_lodcurveerror. Is there a maximum number for it? I have it set to like 999999999999 for as much quality as it can generate, but i'm assuming q3 engine cvars all have a limited scope before it resets to default.

[Raz0r]

r_znear 1
cg_drawScreenTints 1
cg_renderToTextureFX 1
cg_shadows 2
r_flares 1

r_subdivisions is not related to r_simpleMipmaps

r_subdivisions 1 will produce the highest quality curves.

r_simpleMipmaps 0 doesn't have much of an effect, but it sounds logical to not use a "simple" approach to mipmapping if you're aiming for high quality. It will likely aid the effect of trilinear and anisotropic filtering.

Even after looking at the code for r_lodCurveError, I'm not entirely sure what it does, or if a higher or lower value is better. I don't imagine it will look that different. Best way to check is comparing screenshots.

[Rick]

 

Even after looking at the code for r_lodCurveError, I'm not entirely sure what it does, or if a higher or lower value is better. I don't imagine it will look that different. Best way to check is comparing screenshots.

 

r_lodCurveError is basically how you want things in the distance to look. The smaller the value the more edges you'll see on curved areas and so on. Setting it to 99999999999999 (or whatever the highest value actually is) will make it so the curves are seen normally in the distance as well.

[Didz]

r_lodCurveError doesn't have a cap, it goes up to the practical maximum limit for a floating point number. I think 99999999999999 is a bit excessive though. Stand far away from a curved surface with r_showtris 1 and fiddle about with r_lodCurveError, and you'll see there's no changes after 3000. It's safe to say 100000 is a good value if you absolutely always want curves to be high quality and never be optimized for speed.

 

r_simpleMipMaps 0 is a good bet. You'll pay a small time cost while loading a map, but it'll mean that mip maps are generated using linear filtering instead of a more rough approximation.

[Rick]

I figure that 999999999999 just sets it at whatever the limit is. A lot of people with Q3 engine experience say the limit is at around 10000, so even if it's put at 999999999999, it'll just take up the 10000 value anyway. I don't seem to be getting much slow-downs in terms of FPS. The biggest thing causing the occasional lag on some heavily detailed areas is dynamic glow and volumetric shadows. I've turned of the latter, kept the former, but it still gives it an FPS drop. Not very noticable, but can be annoying at some points. I've considered some alternatives to dynamic glow though, like JA++'s bloom options seem like a good option. I think Dynamic Glow's just poorly optimized for this game, with some GPUs not being able to run it at all.

 

But the gain is that the glow on your lightsaber looks good so there's that too. 

[eezstreet]

The real limit is 3.40282347e+38F, but you'll never be able to see the difference past whatever the viewcull is (if not less). So yeah, 10000 seems like a safe bet.

 

Also: my understanding of r_lodCurveError is that it functions similarly to r_lodbias, but patches have their own LODs based on removing vertices, I think?

[Rick]

Yeah, no change in performance whether I'm doing 1, 500 or 9999999999999 here. The big FPS drop is caused by dynamic glow, it depends on the area but usually isn't too bad. I've been trying to use JA++'s bloom options to replace it, but it doesn't seem to work for me.

[Futuza]

Yeah, no change in performance whether I'm doing 1, 500 or 9999999999999 here. The big FPS drop is caused by dynamic glow, it depends on the area but usually isn't too bad. I've been trying to use JA++'s bloom options to replace it, but it doesn't seem to work for me.

That's what we want rend2 so badly for.

[Raz0r]

JA++ hasn't had bloom since I added OpenJK compatibility, because it relied on engine modifications.

The menus are years out of date. Bloom would be best added in a new renderer for OpenJK

[Xycaleth]

If your GPU is newer than 8 years old then it will be able to do dynamic glow without any problems whatsoever. What GPU do you have?

[Jango40]

If your GPU is newer than 8 years old then it will be able to do dynamic glow without any problems whatsoever. What GPU do you have?

It lags on my GeForce 710m. Certainly not the best card out there, but dynamic glow seems to eat way more FPS than it should.

[Rick]

JA++ hasn't had bloom since I added OpenJK compatibility, because it relied on engine modifications.

The menus are years out of date. Bloom would be best added in a new renderer for OpenJK

I am running it with OpenJK though. OpenJK + JA++.

 

 

If your GPU is newer than 8 years old then it will be able to do dynamic glow without any problems whatsoever. What GPU do you have?

ATi 5700 series. It's honestly not that big of a drop like I said. Instead of 170 in some areas, you'll have 130-140 instead and so on. The FPS still doesn't drop below 70-80 except in maps with really extensive areas, like mirrored floors/objects. So it's not really a big deal and it's very much manageable. It's why I'll keep using dynamic glow but if a better alternative comes up, like rend2, I'll take that and use it instead.

[Raz0r]

I am running it with OpenJK though. OpenJK + JA++.

Yes, and when I added OpenJK support I had to remove the bloom feature until it's added in a new renderer for OpenJK.

I just haven't updated to menus to reflect that change.

[Futuza]

I am running it with OpenJK though. OpenJK + JA++.

 

 

ATi 5700 series. It's honestly not that big of a drop like I said. Instead of 170 in some areas, you'll have 130-140 instead and so on. The FPS still doesn't drop below 70-80 except in maps with really extensive areas, like mirrored floors/objects. So it's not really a big deal and it's very much manageable. It's why I'll keep using dynamic glow but if a better alternative comes up, like rend2, I'll take that and use it instead.

Unless you have a fancy 120hz monitor anything more than 60fps is going to be excess and cause tearing.

[Didz]

I guess you can count screen tearing as being something that decreases graphics quality too. So if your "highest graphics" target is intended for actual game playing (not recording or screenshotting), use r_swapInterval 1 (vsync) in fullscreen mode to prevent screen tearing. There's reasons not to have vsync on because of some quirks regarding movement in the Q3 engine, but it's a tradeoff you need to consider.

[Futuza]

I guess you can count screen tearing as being something that decreases graphics quality too. So if your "highest graphics" target is intended for actual game playing (not recording or screenshotting), use r_swapInterval 1 (vsync) in fullscreen mode to prevent screen tearing. There's reasons not to have vsync on because of some quirks regarding movement in the Q3 engine, but it's a tradeoff you need to consider.

I'm intrigued.  What might be some of those reasons not to use it?  (In the past I've only ever turned vsync off if I wanted to fps test things, etc...)

[Raz0r]

I'm intrigued.  What might be some of those reasons not to use it?  (In the past I've only ever turned vsync off if I wanted to fps test things, etc...)

Pretty much, http://jkhub.org/topic/1669-how-to-boost-the-fps-in-jedi-academy/?p=19939

I'm sure I've explained it in detail in other places, but I can't seem to find these places

[Rick]

I don't seem to be getting any screen tearing. Only thing I get is an occasional glowly outline since I got this new monitor, but that might be a 'hidden feature' or something.