I installed Engine 4.2 to try stems. I have 32gb, ryzen 7800x3d (8core, 16 threads), 7900xtx
and after uploading a few test songs into the stems database at the beginning, it went relatively smoothly.
My database consists of about 6,000 songs and now I dragged a larger playlist of about 500 songs as a test and it takes me over 2 hours to scan it.
I calculated it with a stopwatch and the average is 15-20 seconds per stems track. So, converting it into hours, it comes to 180-240 songs per hour, so scanning the entire database will take 25 to 33 hours.
Do you scan these songs just as slowly?
Second question, where to turn on stems on computer in engine 4.2 to listen to the scanned songs? I couldn’t find any option anywhere.
1 Like
i7 13700KF here with 128GB and a RTX3060. On that system, it averaged 4.4 tracks per minute after 2 hours of processing yesterday. That would be almost 14 seconds per track, most of them were 4 or 5 minutes long. It is a relatively heavy process. For comparison, with normal analyzing it processes around 4 tracks per second. I hope this info helps a bit for context.
Regarding your second question, unfortunately we can only hear the stems on the hardware.
2 Likes
I wonder how the stems analysis on rekordbox compares. I noticed that it absorbs about 40% CPU usage, but I hope that they will optimize this process in the next updates.
It’s a pity that you can only listen to stems on a controller. This severely limits your ability to create playlists on your computer…
i7 4770 here from 2013. 100% cpu usage.
Stem analyzed 14 tracks, average 4:45 length per track.
It took about 45 seconds per track.
So your 13700KF is about 3 times faster.
1 Like
This morning, after 26 hours of continuous work, the database of approximately 6,000 songs was finally rendered by stems.
Rendering stems need optimization to make better use of computer resources
