Does Anisotropic Filtering work in the emu?
#41
Are these the z values you're talking about? I'll try reverting this rev and see if it works then.
Reply

Sponsored links

#42
(09-23-2011, 04:42 PM)lightchris Wrote: Of course I could also do more testing. Just make suggestions.

I did more testing.

Graphics card I used at the time I took the screenshots: GeForce 8800 GTS, driver probably FW ~160-180.
Graphics card I am using now: GeForce GTX 460, FW 285.27.
As I was using Windows XP back then, I did all the testing with the same OS.

I tested all versions of GSdx from GSdx9 v0.10-1 (11.08.07) to GSdx v0.1-14 (01.12.09) from this archive. The screenshots must have been taken between these versions, because versions prior to GSdx9v10-8 there were relatively huge graphical errors (such as the character being transparent). I took new screenshots of every version and compared them.

Result: Forcing AF does not have any impact anymore, no matter the version of GSdx. This is proof that something must have been changed outside of GSdx, most likely the graphics driver.


Of course this knowledge won't help us much, as we can't change what Nvidia does with it's drivers.
I did this purely out of interest and in the hope that now someone will believe me.
Reply
#43
AF behavior has not changed in the driver, and nvidia does not specify a profile for pcsx2.

you do the math. (or meth if you prefer)
Reply
#44
Squall:
Countless things have changed between now and then.
It's entirely possible that a change in DirectX causes this. Or an optimization on Nvidias side regarding specific textures, or surfaces.
Only a deeply involved DirectX coder with GSdx knowledge could judge this fairly.
Reply
#45
I remember forced AA used to work only on a specific setting over the nvidia CP and only with the D3D9 renderer and only when using the GSdx fullscreen option when using a custom resolution set to your monitor's native res. Maybe AF needed some specific settings like that to work too.

Anyway I think anisotropic barely made any difference in just a couple games and actually higher res/AA looked much better anyway so bleh Tongue2
Core i5 3570k -- Geforce GTX 670  --  Windows 7 x64
Reply
#46
(09-29-2011, 06:56 PM)Shadow Lady Wrote: I remember forced AA used to work only on a specific setting over the nvidia CP and only with the D3D9 renderer and only when using the GSdx fullscreen option when using a custom resolution set to your monitor's native res. Maybe AF needed some specific settings like that to work too.

I do remember this too. However AA is a different thing and usually much more problematic, especially in quite some "modern" game engines of the last ~6 years (keywords: full screen effects, hdr rendering, deferred shading/rendering).
In contrast to this, forcing AF worked in practically any PC game, regardless of whether it is selectable in the games menu. Now PS2 emulation is a different thing of course, with normally no use of mip mapping. And here is where I think the driver went in: It automatically generated mip maps and applied AF. There even was a "force mipmaps" option in the driver's control panel at one time (though I think this disappeared a longer time ago).

Anyway, I am certain that forcing AF worked under many conditions and not a single special one. I used it in more than one game and played through FFX entirely for example; with different settings regarding the internal resolution, no special tweaks and nothing.

(09-29-2011, 06:56 PM)Shadow Lady Wrote: Anyway I think anisotropic barely made any difference in just a couple games and actually higher res/AA looked much better anyway so bleh Tongue2

Well that's subjective of course.
In my opinion it's very different. I found it very much noticeable. In order to fight texture shimmering you can always go for a higher internal resolution too, right. However the performance drop is a high multiple in comparison. You could force AF with really no noticeable drop at all.
Reply
#47
(09-29-2011, 06:51 PM)rama Wrote: Squall:
Countless things have changed between now and then.
It's entirely possible that a change in DirectX causes this. Or an optimization on Nvidias side regarding specific textures, or surfaces.
Only a deeply involved DirectX coder with GSdx knowledge could judge this fairly.

Its not a change on nvidia's side. I can guarantee it. AF stopped working moving from 0.9.4 to 0.9.6
Reply
#48
I dunno what AF is good for these days. top notch PC games do DOF or lame blur or other screen fx and the sh!t like hell anyway, making that AF depth precision absolutely pointless. I do understand why graph devs don't optimize it anymore.

nuff said
Reply
#49
Distance LOD particularly on floors and walls is still easily perceived to take a blur hit with AF off on most engines.
Reply
#50
(09-30-2011, 02:15 AM)Squall Leonhart Wrote: AF stopped working moving from 0.9.4 to 0.9.6

Not true. I just checked it. It doesn't work in 0.9.4 either.

Just to make this clear: I am not saying that Nvida deactivated something for PCSX2 in particular. What I am assuming is that they changed the general behaviour when an application is not using mip mapping.

(09-30-2011, 02:43 AM)xstyla Wrote: I dunno what AF is good for these days. top notch PC games do DOF or lame blur or other screen fx and the sh!t like hell anyway, making that AF depth precision absolutely pointless. I do understand why graph devs don't optimize it anymore.

nuff said

What Squall said.

And what I want to add is that the reason for using AF in PCSX2 was mainly to fight the extreme texture shimmering. This has always been one of the most annoying things in PS2 graphics, a disadvantage especially in comparison to the other consoles of this generation which usually did proper mip mapping and bilinear/trilinear filtering.
Technically, the PS2 could do this too as far as I know. Yet it is not used because it requires additional VRAM which the PS2 is very short on.
Reply




Users browsing this thread: 1 Guest(s)