OpenGL - FBO black screen when rendering to texture on NVIDIA card

Question

I've followed some tutorials on how to setup and render an FBO. At first it worked and the scene rendered fine, but it turns out the program was using the integrated GPU. (I'm using my laptop)

Then, out of curiosity, I ran it with the "higher-performance" GPU (an Nvidia GeForce GT540M) and the screen was all black. At this point I tried saving the FBO's color texture into a file and the scene was actually being drawn there.

But eventually I found a solution. Previously I would only clear the FBO (color and depth buffers), but now I clear both the FBO and the default framebuffer and the scene is rendered again.

So the question is, is it bad that I have to call glClear twice? Do I really have to call glClear twice? Why would the one clear work on the integrated card?

I can show some code if it helps.

Framebuffer initialization

bool FrameBufferObject::initializeFBO( int width, int height ) {
    glActiveTexture( GL_TEXTURE0 );

    if ( !_colorTexture.createEmptyTexture( width, height ) ) {
        return false;
    }
    _textureId = _colorTexture.getTextureId();

    if ( !_depthTexture.createDepthTexture( width, height ) ) {
        return false;
    }
    _depthTextureId = _depthTexture.getTextureId();

    glGenFramebuffers( 1, &_frameBufferID );
    glBindFramebuffer( GL_FRAMEBUFFER, _frameBufferID );
    glFramebufferTexture2D( GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, _textureId, 0 );
    glFramebufferTexture2D( GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, _depthTextureId, 0 );

    if ( glCheckFramebufferStatus( GL_FRAMEBUFFER ) != GL_FRAMEBUFFER_COMPLETE ) {
        return false;
    }

    glBindFramebuffer( GL_FRAMEBUFFER, 0 );

    _width = width;
    _height = height;

    _isInitialized = true;

    return true;
}

Color Texture

bool Texture::createEmptyTexture( int width, int height ) {
    if ( isLoaded() ) {
        closeTexture();
    }

    GLuint textureId = 0;

    glGenTextures( 1, &textureId );

    if ( getOpenGLError( "Unable to generate TextureID." ) ) {
        return false;
    }

    glBindTexture( GL_TEXTURE_2D, textureId );
    glTexImage2D( GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE );
    glBindTexture( GL_TEXTURE_2D, 0 );

    if ( getOpenGLError( "Error creating empty texture." ) ) {
        glDeleteTextures( 1, &textureId );
        return false;
    }

    _isLoaded = true;
    _textureWidth = _imageWidth = width;
    _textureHeight = _imageHeight = height;
    _textureId = textureId;

    return true;
}

The Depth Texture is the same, except it uses the GL_DEPTH_COMPONENT format,

glTexImage2D( GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT, width, height, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_BYTE, NULL );

Post Processing Shader

Vertex shader

#version 330 core

in vec2 inVertex;
in vec2 inTexture;

out vec2 texCoords;

void main() {
    texCoords = inTexture;
    gl_Position = vec4( inVertex.x, inVertex.y, 0, 1 );
}

Fragment shader

#version 330 core

in vec2 texCoords;

uniform sampler2D texture0;

layout(location = 0) out vec4 outColor;

void main() {
    vec4 color = texture( texture0, texCoords );
    outColor = color;
}

The rendering code looks like this,

glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );

fbo.bindFBO();
glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );

// .. render scene ..

fbo.unbindFBO();

// This is a namespace
PostProcShader::shader.useShader();
PostProcShader::render( fbo.getColorTexture() );
PostProcShader::shader.disableShader();

SDL_GL_SwapWindow( window );

Where the post processing shader simply renders the texture on a screen-sized quad. The scene uses 3 shaders: one for 3D objects, the second for the skybox and last one for fonts.

I'm using C++, SDL2 and (of course) OpenGL/Glew in VS 2015.

Edit:

Depth Test initialization

glEnable( GL_DEPTH_TEST );
glDepthMask( GL_TRUE );
glDepthFunc( GL_LEQUAL );
glDepthRange( 0.0f, 1.0f );
glClearDepth( 1.0f );

When you draw the FBO to the screen do you have depth testing enabled? What's your depth function? — user253751, Sep 27 '15 at 01:13
Yes, then should I disable depth testing when drawing the FBO to the screen? (I edited the question) — aslg, Sep 27 '15 at 01:16
Well, yes, if you don't want depth testing, you should disable depth testing. Also, if you never use depth testing for the default framebuffer, you can create it without a depth buffer. Another option is to use `glBlitFramebuffer()` to copy the FBO content to the default framebuffer. Then you won't need nearly as much state setup, no extra shaders, etc. — Reto Koradi, Sep 27 '15 at 02:05
@aslg You might as well try it. I don't know if it will fix your problem or not. — user253751, Sep 27 '15 at 04:42
`glBlitFramebuffer` looks promising, I'll give it a try. @immibis disabling Depth test works as well. — aslg, Sep 27 '15 at 09:29

score -1 · Answer 1 · answered Sep 05 '19 at 11:53

I got a similar problem when I went from Intel HUD to Nvidia. However my solution was simple. I had to ensure I called glClear(GL_COLOR_BUFFER_BIT|GL_DEPTH_BUFFER_BIT) before rendering to the new framebuffer (on Nvidia) otherwise I would see nothing rendered. I suspect that on Nvidia the depth buffer is not set to the same defaults as on Intel so clearing it sets it to the opengl default clear value.

OpenGL - FBO black screen when rendering to texture on NVIDIA card

1 Answers1