In programs I write using strictly forward compatible, core profiles (and typically 4.2), I have noticed that per-vertex attributes specified using glVertexAttrib* seem to be ignored on devices with an AMD card, while they work exactly as expected on NVIDIA devices. On both, I disable the corresponding per-vertex attribute array and then specify the fixed attribute I want with glVertexAttrib*. But on AMD, it seems to just use the last attribute specified with a real per-vertex attribute array and ignores the one set with glvertexAttrib. Is this a known problem? Is there a workaround or planned fix?