diff --git a/gfx/gl.c b/gfx/gl.c
index 74621b52ad..718ea64077 100644
--- a/gfx/gl.c
+++ b/gfx/gl.c
@@ -1542,7 +1542,7 @@ static bool gl_read_viewport(void *data, uint8_t *buffer)
    RARCH_PERFORMANCE_INIT(read_viewport);
    RARCH_PERFORMANCE_START(read_viewport);
 
-   glPixelStorei(GL_PACK_ALIGNMENT, 1);
+   glPixelStorei(GL_PACK_ALIGNMENT, get_alignment(gl->vp.width * 3));
 
 #ifdef HAVE_OPENGLES
    glReadPixels(gl->vp.x, gl->vp.y,
diff --git a/retroarch.c b/retroarch.c
index f75ecdf5e1..fbc6215dfd 100644
--- a/retroarch.c
+++ b/retroarch.c
@@ -226,9 +226,11 @@ static void recording_dump_frame(const void *data, unsigned width, unsigned heig
          return;
       }
 
-      // Big bottleneck. Also adds one frame "delay" to video output as we haven't rendered the current frame yet.
-      // It will probably improve performance a lot though, who knows.
-      video_read_viewport_func(g_extern.record_gpu_buffer);
+      // Big bottleneck.
+      // Since we might need to do read-backs asynchronously, it might take 3-4 times
+      // before this returns true ...
+      if (!video_read_viewport_func(g_extern.record_gpu_buffer))
+         return;
 
       ffemu_data.pitch  = g_extern.record_gpu_width * 3;
       ffemu_data.width  = g_extern.record_gpu_width;