Graphics stretch issue with latest nVidia drivers

[tsunami2311]

use the nvidia hack or run the game in progressive mode

[refraction]

Well well, long time no post from me

Anyways, it's late, i have to be up early then drive nearly 600 miles tomorrow, but just because I love you guys, I think I have found the definitive fix.

The problem seems to occur when performing a triangle stretch with the geometry shader set to NULL (so disabled in the pipeline).

I have modified the nvidia hack to always use a geometry shader for triangles in the stretchrect function and it seems to work...


If EVERYBODY could test games with the nvidia hack enabled, including AMD users, if all is stable I will just make it a permenant thing and remove the hack option completely.

Happy testing (btw its the grey Download This File button ;p) http://www.filedropper.com/gsdx32-sse2-nvidia

Settings required:

DirectX 11
Native x2 or higher (native doesn't suffer)
hw hacks enabled
nvidia hack enabled

[shinra358]

How bout dat AVX version mang

[Karashata]

Did some cursory testing with The Legend of Spyro: The Eternal Night (and by cursory, I mean ran through the intro splash screens a few times with this plugin and the SSE4 plugin that shipped with the latest GIT build) and was able to confirm that the stretching issue happened in the GIT build plugin but did not occur with this one each time.

Did encounter an interesting issue when I first switched from this plugin to the one from the GIT build for this quick test. The splash screens all ended up as white boxes, and the title screen background was improperly sized and would flicker back and forth from displaying a small section of the background and a bunch of characters (looked like a section of the text sprite sheet, maybe..?). Had to restart PCSX2 to fix that, going to see if I can replicate it now, maybe capture some video footage of it happening in case anyone needs/wants it...

EDIT: I can fairly reliably replicate that issue if I switch from the patched plugin to the SSE4 plugin from the GIT build, switching between GIT plugins without touching the patched plugin doesn't seem to cause it, and it didn't seem to trigger when I switched from the patched plugin to the AVX2 plugin from the GIT build (but it did trigger when I switched back to the patched plugin). Might just be a problem from using the patched plugin then using one that isn't patched, don't know if it could end up being a problem switching between plugins even if they're all patched.

[refraction]

A big caused by switching plugins is pretty low on the priority list, but thanks for checking

Awaiting more tests!

[gregory]

Ref did you include the fbw detection (snow engine stuff) in your build?

[refraction]

Nope, this is purely the master branch, there may be a couple of things i edited and forgot to change back, but i think i got them all

[refraction]

Gonna sticky this for a minute while people test the fix.

[Devina]

Well well, long time no post from me

Anyways, it's late, i have to be up early then drive nearly 600 miles tomorrow, but just because I love you guys, I think I have found the definitive fix.

The problem seems to occur when performing a triangle stretch with the geometry shader set to NULL (so disabled in the pipeline).

I have modified the nvidia hack to always use a geometry shader for triangles in the stretchrect function and it seems to work...


If EVERYBODY could test games with the nvidia hack enabled, including AMD users, if all is stable I will just make it a permenant thing and remove the hack option completely.

Happy testing (btw its the grey Download This File button ;p) http://www.filedropper.com/gsdx32-sse2-nvidia

Settings required:

DirectX 11
Native x2 or higher (native doesn't suffer)
hw hacks enabled
nvidia hack enabled

It works, but then again, the previous Nvidia Hack already worked before.

[vsub]

How exactly this is different than the old hack...I mean,I never used it to begin with and I don't know how to determine if this one is better than the old one.
Does the old one have some negative side or\and don't work somewhere?