Dont care about those red lines...
Typically you will have set the "Preview Render Files" to something NOT so ideal for Mastering (Makes for faster Previews)
So when you render out you wil definitely NOT want PPRO to use those (if inferior) previews to be part of the final results..
Example... You may wanna make your final render into ProRes 422.. But you have set yout preview files to be made in H264 or iframes.
Those are highly compressed already and you dont wanna recompress those... Delete ALL preview files prior to rendering your master that way you make sure that
premiere uses the original source media ! Can take a little longer but thats the sure way !
After that, I would open the Export Media Dialog, do the settings and hit "Queue" that will open AME and while that is rendering you can continue to work inside of PPRO !
If you can view your sequence to make sure its what you want without rendering, then you dont need to render have a choice of 2 ways to export.. You can export directly from PrP (which always uses AME) and if you have the Nvidia CUDA card it will speed up certain items. You can also add it to the AME render que, which will free up PrP to do other things. But it doesnt take advantage of CUDA processing from the Nvidia card at all.
Thank you very much for your helpful answers.
Please double-check this:
1. In case I have to render parts of the sequence for preview, I do this by choosing "Render Entire Work Area" (from Sequence Menue).
2. If I'm done with editing and ready for final output, I have to get rid of these "Preview Renders". I choose "Delete Render Files" (from Sequence Menue).
3. This action deletes ALL audio and video renders linked with this sequence.
4. Even if my sequens contains tons of fx, CC, ... I do export the UNRENDERED sequence; Export >Media (from File Menue).
5. I leave "Use Previews" unchecked and choose "Use Maximum Render Quality"
You can also add it to the AME render que...But it doesnt take advantage of CUDA processing from the Nvidia card at all.
It actually does, for most things. Seems only some scaling operations are unavailable for acceleration when going to the queue (probably something to do with interlacing), but all other effects are fully accelerated using the queue.