Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Wiki MASTER THREAD: Actual FSD Beta downloads and experiences

This site may earn commission on affiliate links.
Really impressed with how the Beta handles different kinds of speed bumps. The wider “humps” that aren’t as jarring it slows down to around 18-19 mph, and for the narrower but more abrupt ones it slows down to below 10 mph. Impressive. Of note is that these are all well painted, so I don’t know if it is vision or map data. I assume it’s at least partially vision because the more abrupt narrower bumps were in a parking lot that isn’t mapped on the Tesla Nav screen.
...
When I tried FSD Beta out in my neighborhood shortly after sunrise on Sunday, I noticed that it slowed to around 19 mph for most of the speed humps, but it tried to barrel over one without slowing from 25 mph at all. (I hit the brake right before, to cushion the jarring somewhat.)

I think its vision detecting them, because most have three big white triangles on their leading edge on each side of the road, like this:

--------------------
🔽🔽🔽----------
-----------🔼🔼🔼
--------------------


I haven't had a had a chance to try going over an old-style speed bump yet.
 
I've noticed several times, when stopped at lights, that the FSD Beta visualization is much more likely to show multiple cars in front of you, than the production FSD visualization did. (I have a vision-only car.)

Here's a picture I snapped while I waited at a long red light just this evening. There are 2 cars shown consistent on my screen--the front one did not blink in and out.

vision_only_2_cars_in_front.jpg
 
I've noticed that from it's vantage point at the top of the windshield, the forward looking camera can see an suv or other tall vehicle over the top of a sedan and will display it. It usually doesn't see another sedan. Although in the above image, FSD also seems to think it can see the brake lights of the second vehicle. I guess they could be high on the hatch.
 
Last edited:
  • Like
Reactions: nvx1977
Well, my car thinks it is 25 mph. So something is wrong.
The present version of FSD beta (10.3.1) will not change to anything lower than 25 mph whether there's a speed limit sign or an OSM tagged node. At least that been my experience. And if an OSM node has neither a 25+ mph sign nor a speed limit tag, the FSD beta speed limit defaults to 25 mph. This is a change from 10.2 which had no default speed limit for tagless, signless nodes. It would continue to obey the speed limit from the previous node.
 
The present version of FSD beta (10.3.1) will not change to anything lower than 25 mph whether there's a speed limit sign or an OSM tagged node. At least that been my experience. And if an OSM node has neither a 25+ mph sign nor a speed limit tag, the FSD beta speed limit defaults to 25 mph. This is a change from 10.2 which had no default speed limit for tagless, signless nodes. It would continue to obey the speed limit from the previous node.
Will be interesting to see what 10.4.x brings.
 
Will be interesting to see what 10.4.x brings.
I hope so, even if all they do is to drop the speed limit for residential and private roads to something lower than 25 mph. A better option IMO would be to have separate speed limit offsets for residential vs. highway. I'm using the terms for road types consistent with OSM terminology.
 
  • Like
Reactions: jebinc
No luck with speed humps in my area, but they are not well marked. I thought you needed to program a destination to test FSD, but today, I decided to start driving and turn AP on. It was like letting the car go wherever it wants. I drove into a residential area with a bunch of roundabouts, and it did pretty well, and eventually drove me to the main road. I was the most positive experience I've had, and when I programmed it to return home, it seemed to do so with "more confidence." One other thing.. I cleaned my windshield around the front camera, streak free before I left, and during the drive, with windshield wipers went on once with a message "cleaning windshield in front of camera."
 
Really impressed with how the Beta handles different kinds of speed bumps. The wider “humps” that aren’t as jarring it slows down to around 18-19 mph, and for the narrower but more abrupt ones it slows down to below 10 mph. Impressive. Of note is that these are all well painted, so I don’t know if it is vision or map data. I assume it’s at least partially vision because the more abrupt narrower bumps were in a parking lot that isn’t mapped on the Tesla Nav screen.
I too am pleasantly surprised the speed humps in my neighborhood are all recognized. They’re all generally marked with chevrons, and the car slows to 18-19 mph which feels just about right.
 
The present version of FSD beta (10.3.1) will not change to anything lower than 25 mph whether there's a speed limit sign or an OSM tagged node. At least that been my experience. And if an OSM node has neither a 25+ mph sign nor a speed limit tag, the FSD beta speed limit defaults to 25 mph. This is a change from 10.2 which had no default speed limit for tagless, signless nodes. It would continue to obey the speed limit from the previous node.
I’ve for sure had mine drop to 15 mph. Because my neighborhood is a mix of narrow and wide streets I set my speed offset to +0 mph. I’ve been caught off-guard a few times when my car would crawl through some street before I realized it updated the speed to 15 mph.
 
I've noticed several times, when stopped at lights, that the FSD Beta visualization is much more likely to show multiple cars in front of you, than the production FSD visualization did. (I have a vision-only car.)

I have a radar car, but the beta email implies that all beta cars are, ahem, upgraded to vision-only: "Your vehicle is running on Tesla Vision!"

Meanwhile, the text about "rolling stops" on the screen description scares the heck out of me since they are illegal in CA. What does that really mean?
 
According to this article there are about 12,000 Beta testers as of 10.3.
That's an interesting data point. Do people think that number includes employee FSD Beta testers? I believe back in March, there were ~800 employee and early non-employee FSD Beta testers with a doubling through early FSD Beta 10.0/10.1. So my previous estimate of roughly 10k new FSD Beta from FSD Button seems to match up.

If my estimation approach is indeed reasonable, then the rollout last night probably enrolled over 600 new FSD Beta participants with 99+ scores.