Topic on User talk:BONKERS

FFXIII AA hair artifacts

8 comments • 01:08, 19 October 2014 01:08, 19 October 2014

8

Anonymous (talkcontribs)

Hi - about this edit about avoiding block-like hair artifacts: does this happen by default or only in a certain case (and what is that case if so)? I probably misunderstood the linked post, but it seems from the post it only happens when combining the in-game MSAA with additional forms of AA or replacing the in-game MSAA? I've only tested it on an AMD rig so far (IIRC Inspector uses a NVIDIA Fermi architecture flag to reduce color thresholds for the "Antialiasing fix" to remove black edge aliasing?), but I can't tell the difference between that, the linked demonstration video of the problem, and the PS3 version (I spent a few hours last night gazing between the 3).

In fact, on a tangent, the "shadow acne" looked the same too if you did it from the right angles. The only real difference I noticed was the controllable brightness on the PS3 version, as it seemed brighter and the hair highlights looked more pronounced (I'm not sure why the PC version doesn't have a brightness slider).

That said I'm just an ignorant layman on the subject with very limited knowledge. Sorry if I rambled a bit - the edit caused me to read up more on the subject than I usually would.

00:04, 19 October 2014 00:04, 19 October 2014

BONKERS (talkcontribs)

It doesn't happen on AMD cards as far as I know. (And you are right, that the AA Fix is for Fermi and above, the description of the function within the driver is "FERMI_SETREDUCECOLORTHRESHOLDENABLE" It doesn't give any description of what it exactly does, but it has been known to fix several issues and ALSO CAUSE several issues depending on the game. Which is why it's not a good idea to enable it globally.) (There are also around ~7 or 8 other Fermi color threshold functions with additions at the end specifying a particular texture format. But there is no way to enable these on a user basis to see if they do anything)

It does happen with the in game MSAA as well. I mention in my reference/post that it does happen even with in game MSAA.

The linked video shows it happening with only in game MSAA. (If you look at the GPU usage, it is only ~10% that's on a GTX 570) It does not happen on PS3.

Shadow Acne on PS3 versions in some scenes is far less severe. If you saw my post on NGaf about the opening scene with Light and Sazh for example.

Increasing the shadow resolution with GeDoSaTo curtails the issue somewhat.

I understand what you mean, though others have done as I recommended and confirmed the AA Fix does indeed fix the block issue on hair with just using in game MSAA besides my own testing.

00:56, 19 October 2014 00:56, 19 October 2014

Anonymous (talkcontribs)

Interesting, and sorry if my sounded combative in any way - I was just trying to learn some stuff. Normally I don't get into these subjects at all, but how they relate to FFXIII and Crystal Tools fascinates me. BTW I've read both NeoGAF threads several times, but they are sort of long by now (I've been trying to get to full membership on NeoGAF for years, but my ISP keeps changing before the moderation queue comes around).

Anyway, back on topic. I wonder which cards exhibit the this? Is it just the Fermi+ that the "AA fix" is for (I ask if it involves a GeDoSaTo setting but I gather you already answered that)? I couldn't reproduce it (blocks, mostly in the middle of the head) on my NVIDIA laptop with a pre-Fermi card (9800M GTS), but I have a modern mid-gaming desktop rig with a 750ti at another place that I gather I should be able to.

As for shadows, it's an annoying problem, and the GeDoSaTo shadow resolution setting seems to be a YMMV thing. On my AMD laptop (4650 Mobility HD) it just expanded the shadows and made the problem worse even at 2x (I should SS that...).

Ultimately I want to be able to perhaps heavily summarize that post here, or rather what to avoid, since this is about fixes and such (while still linking to it of course, I did a somewhat similar thing with the GPU-going-to-sleep explanation (even if it a bit long, and hopefully it's technically correct)). Or would that not be a good idea?

I'm probably rambling, so thank you for humoring me! FFXIII is just the kind of weird port that I'd expect from a JP company that isn't Falcom. Especially its official "XInput controller" recommendation (it doesn't support XInput at all or use XInput drivers of the controller, it just uses DirectInput and assumes the internal layout is the same as the XBox controller spec). Thanks again for the explanations!

04:51, 19 October 2014 04:51, 19 October 2014

BONKERS (talkcontribs)

I didn't take it as combative or in any negative way. Don't worry, I was just frustrated about important information being over generalized,removed or simplified when it didn't need to be.

I've been meaning to try and write more comprehensive stuff for AA with Nvidia cards here, but is just a lot of stuff to cover.

"I've been trying to get to full membership on NeoGAF for years, but my ISP keeps changing before the moderation queue comes around). "

I can understand the frustration, it took a long time to get on there too. At this point I can't remember if it was from waiting or from someone I knew. >_>

As far as the AA Fix is concerned. That is interesting, it probably does only apply to Fermi and greater, same as that SGSSAA only works on Fermi Cards and greater as far as I remember. Because I know the issues and fixes from the AA Fix DO persist to Kepler and Maxwell.

On the shadows, I forgot what 2x looks like, but 4x and beyond look leagues better, even if issues remaining are simply from how stuff works in the game engine. (Like they did in FFXIV 1.0), after checking Dragon Quest X since it also uses Crystal Tools the issues don't exist to lack of self shadowing and the stylized rendering approach.

"Ultimately I want to be able to perhaps heavily summarize that post here, or rather what to avoid, since this is about fixes and such (while still linking to it of course, I did a somewhat similar thing with the GPU-going-to-sleep explanation (even if it a bit long, and hopefully it's technically correct)). Or would that not be a good idea? "

I think adding things like the fact that it's Fermi and forward specific are ok. Simplifying something like this too greatly isn't a great idea I think. My link post has an in depth step by step process to get Nvidia Inspector set up to the point to fix it. Simplifying it to "Follow the guide in the reference" like I have it set up I think is the bare minimum it can get to.

"I'm probably rambling, so thank you for humoring me! FFXIII is just the kind of weird port that I'd expect from a JP company that isn't Falcom. Especially its official "XInput controller" recommendation (it doesn't support XInput at all or use XInput drivers of the controller, it just uses DirectInput and assumes the internal layout is the same as the XBox controller spec). Thanks again for the explanations! "

It is a weird port! I mean.. it's like they didn't quite think it through. I was so excited but skeptical when it was announced and I was right to be I guess. I mean, they can't be that out of touch. They are on the ball when it comes to Final Fantasy XIV ARR on PC. (Except for when it comes to AA and UI scale. Which hampers downsampling)

The controller nonsense makes literally no sense! I mean, if this is based off the Xbox 360 jp code branch, it'd HAVE to be Xinput. Being direct input would mean using BetterDS3 to map the PS3 controller to direct input would get it to work just fine. Same with the PS4 controller ,because isn't it just DirectInput??

Why change to DirectInput?

>_>

08:28, 19 October 2014 08:28, 19 October 2014

Mirh (talkcontribs)

Well, in fairness I never understood all that Xinput crap. I mean.. it's just the API used by x360 controllers.. ok
But it was pointless and it has only downsides when compared to old DirectInput.
You could say that triggers are messed up and force feedback does not work with it.. but this was microsoft fucking up with drivers (call it purposeful limitation). It's not API own problem.

Besides.. this may even be really weird (bad?) port. A sop, definitively.
Though personally, I was going to buy it for cheap for my PS3. Now I won't have :D. So it's still an improvement over the previous situation after all imo

10:00, 19 October 2014 10:00, 19 October 2014

BONKERS (talkcontribs)

If they don't fix the issues with the port. Even with amazing IQ and AA available. I may end up sticking with and replaying on PS3 afterall personally.

20:43, 19 October 2014 20:43, 19 October 2014

Anonymous (talkcontribs)

"I think adding things like the fact that it's Fermi and forward specific are ok. Simplifying something like this too greatly isn't a great idea I think. My link post has an in depth step by step process to get Nvidia Inspector set up to the point to fix it. Simplifying it to "Follow the guide in the reference" like I have it set up I think is the bare minimum it can get to."

I moved a variation of the guide reference itself to the essential improvements section, while also keeping the macroblock issue below that keeps a trimmed down version of my previous explanation, referring back to the guide section. Let me know if it looks OK now (I might explicitly cite the fermi architecture in addition to the 4XX card mention).

14:41, 26 October 2014 14:41, 26 October 2014

BONKERS (talkcontribs)

I would also GREATLY appreciate if you didn't remove my AA Reference for Nvidia cards and AA. The game needs VERY specific settings to function correctly.

Your generalizations do not help end users

01:08, 19 October 2014 01:08, 19 October 2014