Guys, it may not have been covered all together in this particular thread, but
there are several things that (at least I continue to believe) are fact related to USB scan times through 2.42.40:
- More depth to the directory structure will add scan time.
- Remember though, with 8.0 2.36.108, if it was too shallow, it would cause Media Player to (I suspect run out of resources) and never complete scanning for some people. My same stick of 6100 tracks worked in 7.1, but would never complete scanning even after 2 hours with the initial drop of 8.0. When I moved from a 3 to a 2-deep directory structure, the same tracks completed scanning and were playable. Others had no problem with libraries much larger.
- As I continue to recommend, if you have any sizable library, at least a 2-deep directory structure is best, and don't add more than you really need as it impacts scan time and memory consumption.
- The physical size of the files and total data just don't make that big of a difference.
- I continue to suspect that as Tesla evolves their scanning algorithms, this is changing, but the length of directory names, filenames, and tag content itself (including how much is or isn't duplicated) influences scan time. It's why none of our numbers match from person-to-person despite everyone asking questions and posting numbers that make at least me a bit crazy. As to how these things influence scan time, only Tesla programmers could hesitate a guess. X-programmers like myself can think of a zillion ways that different approaches would influence time up or down, and without any sort of detailed release notes to this fine level of detail, it's not worth our guessing here as our world can change with each new firmware drop.
To prove my points to non-believers, I just completed several tests. My MS was sitting in my garage OFF, with only the driver door open, no music playing, and I wasn't fiddling with the IC or CID. Before each test, I performed a full CID reboot (foot on brake version), then waited at least 5 minutes for it to settle down, Nav maps being fully displayed and my WiFi connection having been re-established. Thats about as close to the same environment for each test as I could get.
TEST 1) 6829 Songs, composed of 430 Artists, 1196 Albums, 13 Genre and 1 Folder - 161GB in size on a PNY Elite 256GB Stick
- A 3-level directory structure: YYYY-MM-DD/ALBUM-ALBUMARTIST/tracks
- It took 21 minutes to scan
- This becomes my baseline test case
TEST 2) Same 6829 Songs, composed of 430 Artists, 1196 Albums, 13 Genre and
1216 Folders - 161GB in size on a PNY Elite 256GB Stick
- A 2-level directory structure: ALBUM-ALBUMARTIST/tracks
- It took 17 minutes to scan
- I believe this is a good proof point towards my premise point #1. FWIW, this is what I have migrated back to in my MS.
TEST 3) Same 6828 Songs plus 1*, composed of 430 Artists, 1196 Albums, 13 Genre and 1216 Folders -
201GB in size on a PNY Elite 256GB Stick.
*What I did here was manually reconstruct one track to fill up a whole lotta space with a bunch of nothing, so I still had 6829 tracks, it's just the aggregate contents of all files on the stick was ~25% larger.
- The same 2-level directory structure: ALBUM-ALBUMARTIST/tracks
- It took the same 17 minutes to scan
- I believe this is an adequate proof point towards my premise point #2.
TEST 4) Same 6829 Songs, composed of
1 Artists, 1 Albums, 1 Genre and 1216 Folders - 161GB in size on a PNY Elite 256GB Stick
What I did was make a mass change of the TRACKARTIST, ALBUM, and GENRE tags in each of the 6829 tracks to be the same for my whole library. Completely unreasonable use-wise, but a great test case (yes, I used to make test data in the good 'ol days for my programs and systems).
- The same 2-level directory structure: ALBUM-ALBUMARTIST/tracks
- It took a whopping 39 minutes to scan (compare against test case #2). That's extraordinary isn't it?
- I believe this is a great proof point towards my premise point #3.
THE NET IMHO:
- If we want to do high level comparisons of USB scan times, we need to specify both number of tracks and depth of the directory structure. Even with that, as I've suggested for a very long time, there is HUGE variability in how each of us names and tags each and every track that prevents us from providing any useful comparative scan data that we can draw any real conclusions from. I'm over it, and IMHO you should be too.
- I just know scan times are far too long, and could use an extraordinary amount of tuning AFTER other bugs are addressed.