still problems with amd 5xxx-series?

[manolo_]

hi,

i want to buy a new pc-system and im interested in amds 5xxx-series (maybe 5770), so are there still any bugs with ati/amd?

[imbaczek]

there are some, but hopefully none that crash the game; 10.2 drivers and spring 0.81.2 is a happy combination, at least for me.

[manolo_]

and special bling-bling liek shadows, reflection + better water-effects?

[slogic]

I don't understand what "bling-bling shadows" means. So, tell exactly what settings to test and i can test for you. Haven't tested reflective water because i'm happy with Dynamic water. Also here was told many times that reflective water should not be used with ATI cards. But, i'll give a try cause ATI raised OpenGL support to 3.x.

[imbaczek]

reflective water should be fixed as of 10.2 catalysts or spring 0.81.2.

[Tribulex]

can someone make a list of things that dont work on latests ati cards with latest drivers? I would really appreciate this.

[manolo_]

can someone make a list of things that dont work on latests ati cards with latest drivers? I would really appreciate this.

me too, sorry for the imprecise phrases, i ment with bling-bling the good looking stuff that u could enable in experts (like reflecting units, shadows, clouds,....) - but tribulex suggestion is better (like always)

[zerver]

I'm also much interested to know because I have two of these cards, but didn't have time to try them yet.

[imbaczek]

bump water with shorewaves works fine, as of 10.2.

what still doesn't work is texture alpha on on feature shadows - it works as if the alpha channel had 0% transparency during shadow drawing.

i can't think of anything else that comes to mind immediately. 10.2 drivers are a great step forward for AMD.

[Argh]

what still doesn't work is texture alpha on on feature shadows - it works as if the alpha channel had 0% transparency during shadow drawing.

IOW, it works like Units do... still... <rage>

[Tribulex]

what still doesn't work is texture alpha on on feature shadows - it works as if the alpha channel had 0% transparency during shadow drawing.

IOW, it works like Units do... still... <rage>

yeah this is shite

[zerver]

This texture alpha thing, the question is if the drivers really are bad or if it just needs some AMD specific tweaking. If anyone finds some open source OpenGL demo or something that has working shadow alpha, we should definitely look at how it is coded.

Most annoying is the lack of tree shadows.

[Argh]

Spring's not sending tex2 to the unit shadowmap rendering pass for Units, whereas it does for Features.

If somebody would fix that, I'll translate the ARB to GLSL and fix the fragment shader. The fix is actually pretty straightforward. It won't correct the problems with "floating" and other distortion, but it'd correct the alpha problem.

[Argh]

Code: Select all

//!!ARBvp1.0
//# vertex program for generating the shadow buffer

unifrom vec4 p16; //See lines, 259-264, http://github.com/spring/spring/blob/master/rts/Rendering/ShadowHandler.cpp
uniform vec4 p17; //this cooresponds to p17 in Spring source
uniform vec4 p18; //ditto

varying vec4 outPos;//output of function, for fragment shader

vec4 pos = gl_Vertex;

//ATTRIB pos = vertex.position;
//PARAM mat[4] = { state.matrix.projection };  //These should be sent as uniforms
//PARAM mat2[4] = { state.matrix.modelview };  //Ditto
//OUTPUT opos = result.position;  //Output, or gl_Position
//TEMP temp,temp2;

vec4 temp;
vec4 temp2;
mat4 projection = gl_ProjectionMatrix;
mat4 modelview = gl_ModelViewMatrix;

//DP4 temp.x, pos, mat2[0];
temp.x = dot(pos,modelview[0]);
//DP4 temp.y, pos, mat2[1];
temp.y = dot(pos,modelview[1]);

//ABS temp2,temp;
temp2 = abs(temp);

//ADD temp2,temp2,program.env[17];
temp2 = temp2 + p17;

//RSQ temp2.x, temp2.x;
temp2.x = 1.0 / sqrt(temp2.x);
//RSQ temp2.y, temp2.y;
temp2.y = 1.0 / sqrt(temp2.y);
//ADD temp2,temp2,program.env[18];
temp2 = temp2 + p18;
//MAD temp, temp, temp2,program.env[16];
temp = (temp * temp2) + p16;
//DP4 temp.z, pos, mat2[2];
temp.z = dot(pos,modelview[2]);
//DP4 temp.w, pos, mat2[3];
temp.w = dot(pos,modelview[3]);

//DP4 opos.x, temp, mat[0];
outPos.x = dot(temp,projection[0]);
//DP4 opos.y, temp, mat[1];
outPos.y = dot(temp,projection[1]);
//DP4 opos.z, temp, mat[2];
outPos.z = dot(temp,projection[2]);
//DP4 opos.w, temp, mat[3];
outPos.w = dot(temp, projection[3]);

//#need texture coord for alpha testing (to shadow masked texels correctly)
//MOV result.texcoord[0], vertex.texcoord[0];
gl_TexCoord[0] = outPos;

//END

I think that's right. However, we'd need to rewrite the shadowmap to use a fragment program and output the results to a texture in a FBO.

[EDIT]OIC; it's all getting merged into one shadow-drawing pass, using the same shader. Well, that's certainly heading in the right direction.[/EDIT]

[slogic]

Trees:


Reflective water:


Reflective +refractive water:

[slogic]

Bump-mapped water:

[Tribulex]

im pretty sure that trees shot shows how the shadows are bugged.

[Caydr]

I'm dumb, isn't dynamic water a better option than bump-mapped in any case?

[Gota]

I'm dumb, isn't dynamic water a better option than bump-mapped in any case?

No