Darktable ignores embedded JPG for preview

The whole point of a raw editor is to process the raw data, not the embedded pre-processed JPEG data. Darktable only uses the embedded JPEG-data for the very initial preview, which you can see directly after importing as thumbnail in the lighttable (but that also takes some resources, although minor). After processing, including just opening the picture in the darkroom mode, the processed RAW data is used as thumbnail.

If darktable is slow for you, I recommend you get faster hardware, either CPUs or GPUs (latter is recommended). Just using the embedded JPEG-data for thumbnails won't make your experience with darktable or any other raw-data processing programm faster.

For more details on how darktable handles thumbnail creation, please refer to the manual section 2.2.3.1. Thumbnail creation:

[...]

When darktable imports an image for the first time, there are two possible sources from where to take a thumbnail. darktable can either try to extract an embedded thumbnail out of the input image – most raw files contain these kind of thumbnails generated by the camera – or process the image by itself using default settings. You can define how darktable gets its thumbnails in the GUI preferences dialog (see Section 8.1, “GUI options”).

[...]