As I know, OpenGL has highly supported NPOT textures since 3.0. Now I' m facing an bizarre phenomenon that NPOT texture cannot be allocated with precise corresponding data.
Words are plain. My testing code is as below.
Code :
	glGenTextures(1, &texVoxels);
	glBindTexture(GL_TEXTURE_3D, texVoxels);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
	glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE);
 
	int width = 513;
	int height = 513;
	int depth = 513;
	unsigned char *data = new unsigned char[width*height*depth];
	memset(data, 0, sizeof(unsigned char)*width*height*depth);
 
	glTexImage3D(GL_TEXTURE_3D, 0, GL_LUMINANCE8, width, height,depth, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE, data);
 
	delete []data;
	data = NULL;

As code shown above, I need to allocate an NPOT cubic texture of 2^N +1. But program will exit error on glTexImage3D(). The error, I think, is caused by "the size of data less than the texture needed". Because if I allocate data larger enough, the program will pass.
Could anybody please explain this thing for me? I think it is a hardware-related question. Anyway, how could I use NPOT texture and allocate data like this? Even though when the size of data is larger enough the program passes, I am really not sure about how data exactly corresponds to the texture location then.
PS. My graphic card is quadro fx 4800, which supports OpenGL 3.3.

Best regards.