CSAA, Transparency AA (Multisampling), & Texture Sharpening (MIP LOD Bias) (w/pics)

The proof is in the pudding, no? Maybe ISI's TS @ 0 (off) isn't what you're thinking it is, or maybe it's not working correctly?... I find it odd that ISI calls TS @ 0 "off" rather than "0". I mean they state a number for the other ones (+2, +1, -1, -2), maybe what ISI calls "off" isn't truly a MIP LOD of 0?... I have no idea, it probably is the same as 0, I'm just trying to think of something, anything. All I know is what the actual results tell me.

Can you try doing the same thing, with the same car and same track? Obviously maxed graphics... It's at Palm Beach (facing the wrong way, start finish line, main straight, cockpit FOV of 18). Do not adjust the chase cam except for one parameter. Move it as far back as it will go, other than that don't adjust it in any way (for exact consistency).

It's been doing this with 3 different 780 Tis, and, I don't know, 7 different NVidia drivers?...

I'm using 16xCSAA (original post).

 

Yes, and my pudding tastes like it should whereas yours doesn't.

We'll work out why.

Sent from my SM-G900F using Tapatalk

 

Your results are the opposite of what the rest of us get.

But I believe you. As I mentioned in another thread, there has to be some latent settings somewhere affecting my system because I get the heat haze bubble bug in the mirror even when I don't use Inspector and select the most basic settings. This bug comes from mixing multi and super sampling or using SGSS--at least I have never read any other explanation for it. So why/how am I still seeing it?

I think those of us who fiddle with these settings risk inadvertently SNAFU'ing things so that we are caught going in circles forever because the adjustments we make don't respond as intended or as they way they do for "everyone" else.

I may uninstall my NVIDIA driver completely and try rF2 with a generic VGA driver just to prove that the heat haze bubble disappears. Then try to do a total uninstall/clean of NVIDIA and start over. You may have to do the same

 

Hahaha, TechAde, lol

What I'm not understanding is...if TS @ 0 (off) should be sharper than TS @ 2 (+1), then how does TS @ 0 (off) also have less aliasing. Or vice-versa, if TS @ 2 (+1) is supposed to be more blurry, then why does it have more aliasing? Wouldn't the lower the number, the sharper but more aliased the image gets?... Why would a higher MIP bias (+1) have more blur AND more aliasing than a lower MIP bias (0)?.... Isn't it, the higher the number the blurrier but less aliased, and the lower the number the sharper but more aliased? It seems to be that way with all the MIP biases (+2, +1, -1, -2) except with 0, (which ISI calls "off", rather than putting a "0").

[HR][/HR]
Marc Collins, can you do the same thing I asked TechAde (post #42)?...

 

Marc, I don't know if this holds true for rF2, but in the FSX flight sim the general rule is:

1) If in-sim settings are modified (or cfg files), delete all shader files so that they can be rebuilt
2) If settings are changed on the GFX card, no need to do so

 

GIFs of "Texture Sharpening" - direct 1:1 comparison

(I deleted this post by mistake since my tablet doesn't give me a confirmation button if I press "delete" by accident; it was originally posted about 7 or 8 posts back.)

Ok guys, I have 2 GIF images that switch back and forth. One is for the external view, one is for cockpit. It is only between TS @ 0 (off) and TS @ 2 (+1), everything else seems irrelevant since the rest were just way too aliased, well, except for TS @ 1 (+2), but that just destroys the entire scene with blur.

I am using 16xAF - 100% sure and confirmed.

Read the following first:

A. The following part of the image shows such a tiny difference that it's basic irrelevant, in-fact, some people may not even be able to see the difference. It's basically almost imperceptible, but I'd thought I'd mention it anyways just for further proof that TS @ 0 (off) does blur the scene more than TS @ 2 (+1):
- The sections of grass on the left and right side (highlighted)

B. The following parts of the image show such a small difference that they also make almost no difference to gameplay, however, they are more noticeable than "A" (above) and it may matter to some very picky people (not me)
- the sharpness/detail of the painted white lines, especially as they get closer to you (highlighted)
- the trees, some of them loose "fullness/thickness" as if some of their leaves/branches fell-off (highlighted)

C. The following part of the image shows a massive difference and is extremely noticeable from both the cockpit cam, external "chase" cam, and also from just about every and any trackside/TV cam (unless it's angled in such a way where you cannot see into the distance)
- most of the road in front of the car and continuing-on forward (highlighted)

D. UPDATE: I forgot to highlight the fence on the left side in the external pics. With TS @ 0 (off), the fence is so blurred that it actually looks it's completely missing - gone. With TS @ 2 (+1) it's still visible (and not aliased like with TS @ 3, 4, and 5). By far, my main concern is "C" (driving road surface), but a completely missing fence may still be important to some (eg. screenshot/video makers, super picky people).



NOTE: the GIF images loose quality and look "dotty", however, they still manage to easily show the differences, especially with regards to "C". That being said, I have uploaded the original pictures if you would like to view and compare them in much better quality (I recommend right-clicking on the 2 external or 2 cockpit pics and opening both up in a new tab respectively in order to easily switch between the two for a perfect comparison, and viewing the pictures on a "regular" monitor, not something small like a tablet or phone).

GIFs
External (chase) - http://makeagif.com/i/4VAota
Cockpit - http://makeagif.com/i/9BQojq

 

Is the image in the mirror affected by any settings at all?
Don't the mirrors just show like default bog standard settings regardless of any image settings used?

 

@Spinelli and anyone else that wants to attach pics

your screenies may well be 1600x900 but you are losing a lot of detail reducing them for attaching.
So quite possibly anything you are trying to show us is lost in the reduction.

 

TS @ 0 doesn't have less aliasing than TS @ 2, at least not on any system other than yours.

What's going on with your system I'm not sure yet, other than some of the bias settings you're applying aren't actually taking effect properly.

When I'm back at my PC and can replicate your shots I will work out what's going on.

Could you post a shot of your complete Inspector profile please?

Sent from my SM-G900F using Tapatalk

 

Why is rfactor2 the game most difficult to configure graphics of the history?

 

rF2 is very easy to configure using in-game settings. The guys here, for the most part, are talking about absolute optimization of GFX which is something else entirely (advanced HW/SW tweaks).

 

Nothing was reduced. Original size. Original quality. They are original 1600x900 pics. It's native resolution of the original image. Even if they were reduced, the difference is so massive that to not see it would be ludicrous. However, rest assured, the images are native size. I made sure not to resize, convert, or do any sorts of stuff to the pics.

The only ones reduced are the GIF images, however, even then, the differences are so massive that you'd have to literally be blind to not see the difference. Seriously, if you're a human being, and not legally blind, then there is no way for you not to be able to see the difference.

 

Don't the adjustments do just that? Adjust? So +2 just increases the current bias by 2? -3 would become -1, for example. I don't know if that's relevant at all, but thought I'd throw it out there.

 

2 Blurry seems to suit me best thanks for the info.

 

Nope, the TS options set a specific bias globally, except for 5 which uses the bias set in the materials.

Sent from my SM-G900F using Tapatalk

 

Ok. I suppose that's actually worse for different scene comparisons? Depending on objects and textures the track maker can adjust to suit, whereas forcing a single setting could produce mixed results?

 

I don't think a 1600x900 pic equates to 500kb size?
But that's irrelevant if you say differences are easy to see still.

 

They are identical to the original pics straight from rFactor 2. Absolutely 100% identical. All i did was crop the original triple screen pic (around 5800x1080), and I cropped both pics on the exact same pixel so as to make them identical. And i cropped an area 1600x900.

They look identical to the original pics, identical. If you don't believe me you can see for yourself. The original post has those EXACT same 4 pics linked to, only difference they are uncropped (triple screen wide), so you can easily see this for yourself by comparing the cropped and highlighted pics to the uncropped, unhighlighted, and origjnal pics.

Finally, it's not even a matter of whst I say or trusting me, just click the pics and see for yourself, the difference is ridiculously obvious. The difference is so obvious that you can even see the difference in the GIF pics and the GIF pics do loose quality because of the GIF website converting the 2 pics into one GIF.

 

"the TS options set a specific bias globally, except for 5 which uses the bias set in the materials."

Ah, thanks TechAde. This is what I surmised, but wasn't sure. One other question, does a setting of zero completely disable TS? If so, how does this differ from a setting of 5?

 

0 forces 0, it doesn't disable anything. 5 uses track creator's settings, as you already understood. All settings force a certain level except 5.