yeah it does, but I have never had issues this bad until 10.5. Like I said, it makes it a pretty miserable drive. Used to be very rare, now its every mile no matter the road or conditions.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla FSD Beta Release 10.5 2021.36.8.8 - 10.5.1?
- Thread starter captainreynold
- Start date
EVNow
Well-Known Member
That looks unique to you. Most people think 10.5 is a slight improvement. You can try recalibrating cameras.
aircooled
Member
I have the same problem with the slow down breaking for no apparent reason in the neighborhood. I received 10.5 a few days ago after having 10.4 for only 2 or 3 days. I am not seeing much if any improvement on my 2 short drives with 10.5 through the neighborhood. It still has the problem of slowing down rapidly every so often at side streets for no apparent reason. Probably as often as every third or fourth street. It's really bad. It takes turns wide and has some jerky steering wheel action when stopped and creeping forward to make a turn. 10.5 seems a lot like 10.4 so far for me and that is not a good thing. It so similar it makes we wonder if I'm really running 10.5. The download went without a hitch and the screen indicates 2021.36.8.8 but it took a few days for 10.4 to show up the toggle to turn on FSD once I loaded it also.
Atari2600
Active Member
Drove around 160 miles on a 60 MPH 4 lane divided highway that is not limited access so FSD 10.5 was in effect the whole way. Since I forgot my mobile charger which I recalled about 10 miles into the trip and it was cold and raining I drove the speed limit the entire way down but the car always wanted to get into and stay in the left lane which was inappropriate because I was by far the slowest car on the road that day. I ended up using TACC most of the way. It was annoying because I would signal to get into the right lane and 1/4 mile later it would signal and go into the left lane again. Totally unacceptable.
On southbound state route 104 near McDermott Ohio there was a railroad crossing that had markings on the road and 20 MPH yellow sign and the car tried to take it at the speed limit 55 MPH. I took over and drove over it around 50 MPH which would have resulted in a very bad safety score but was not unsafe.
At the intersection of 104 and and 73 traveling southbound on 104 I had to take over because for some reason it did not want to stop for that stop sign. From the picture you can see there is buffer space to allow you to run out past the stop about 5 car lengths before bad things happen.
On southbound state route 104 near McDermott Ohio there was a railroad crossing that had markings on the road and 20 MPH yellow sign and the car tried to take it at the speed limit 55 MPH. I took over and drove over it around 50 MPH which would have resulted in a very bad safety score but was not unsafe.
At the intersection of 104 and and 73 traveling southbound on 104 I had to take over because for some reason it did not want to stop for that stop sign. From the picture you can see there is buffer space to allow you to run out past the stop about 5 car lengths before bad things happen.
You are expecting to use FSD Beta as a driving aid while your family is in the car?!
FSD is not ready for release (and may never be).
FSD is in Beta, we have volunteered to identify edge cases, period.
No expectation of performance is relevant at this time.
TravelFree
Active Member
I need more time with 10.5 but after about 100 miles, I would grade it the same overall score as 10.4. Still too dangerous to call it an improvement. I had high hopes for the merging lane change feature as stated in the release notes, but unfortunately it didn't always work as claimed.
Reminder- my main use for FSD will be for better FSD on the highway than when I first bought it. Don't need or want to depend on it for city driving even after it is open to the public. Maybe some day.
First version of FSD available to the initial batch of 100 scores was 10.2, not 10.1. I am also part of that cohort (received FSD beta on Oct 11).
Phantom braking frequency (hard or soft) has been consistent throughout all the point releases so far, in my experience. Ever since 10.2, if I had passengers with me, I would run FSD but manually use the accelerator to prevent any phantom braking. This resulted in a FCW where it would have applied the brakes. I felt that between hearing the FCW vs feeling the braking, the FCW was less jarring for my passengers. Still doing this for 10.5. Frequency of phantom braking / FCW doesn't seem any different.
yeah, my bad, I had forgot which version was first.
Already, with passengers, I don't use FSD door to door, but I still like to use it on straightaways(basically lane assist with cruise control). But with the phantom braking, its just not feasible right now.
I guess I could make another driving setting to have one for drives with other people and have FSD turned off and maybe it would work better. Haven't tried that yet.
sperkin
Active Member
FSD beta is actually worst now on the highway. It kept phantom breaking at every break in the highway. At one point it it slammed on the brake from 80 to 55. No one was behind me at the time so I was testing to see how slow it would go. It also gives forward collision warning on sharp bends on 1 lane highways when oncoming cars are coming around the bend. The auto high beam also caused some road rage. It turn On and Off as cars are near the radius that triggers it so it's flash at them until you get closer to their rear end.
Lat year I took the same trip and it was perfect in good old Autopilot.
If you are comparing a single trip this year to a single trip last year (say, Thanksgiving-specific travel), then you're likely seeing the difference between vision-only and the old radar-based AP, not something specific to FSD. The highway logic is current production code that doesn't use HW3. The auto high beam is something specific to FSD builds, as it forces it on you when you engage FSD.
Fortunately we can disable auto high beams. Just push left stalk away from you, and it's off. But you need to do this every time you engage FSD.
sperkin
Active Member
TravelFree
Active Member
OK, after 3 hours of test drives now I'm firmly in the camp that v10.5 sucks!
The only time I give it a passing ( barely) grade is in a straight road with stopping at red lights. However, I don't like how it wants to slam on the brakes at the last 10 seconds approaching a red light. The only time it approached a red light at humanely normal deceleration was when it had another car slowing down in front. Where there were hard left and right turns at 30mph it failed to slow and actually crossed over to the left on a blind turn. Failed to follow the navigation route taking me 4 miles out of the NOA route. It scared the crap out of other drivers when it would go to the center of an unmarked road. Failed to make right on red causing the car behind me to lay on the horn until I took over.
I'll continue to use it on long straight rural roads, but be cautious if the road has a hard turn. Still pretty good on the interstate. Automatic passing of slow cars is better than non FSD Beta.
The only time I give it a passing ( barely) grade is in a straight road with stopping at red lights. However, I don't like how it wants to slam on the brakes at the last 10 seconds approaching a red light. The only time it approached a red light at humanely normal deceleration was when it had another car slowing down in front. Where there were hard left and right turns at 30mph it failed to slow and actually crossed over to the left on a blind turn. Failed to follow the navigation route taking me 4 miles out of the NOA route. It scared the crap out of other drivers when it would go to the center of an unmarked road. Failed to make right on red causing the car behind me to lay on the horn until I took over.
I'll continue to use it on long straight rural roads, but be cautious if the road has a hard turn. Still pretty good on the interstate. Automatic passing of slow cars is better than non FSD Beta.
What was your speed before it started decelerating?
What I've noticed is that if it's a fast road, say 50mph, the car will feel like it should have slowed earlier. The car simply doesn't see the light until it's somewhat close, and at higher speeds, this feels way too late. If the road is only 40mph or under, it feels like proper deceleration. This is the same reason why the car feels like it starts to decel way too late at highway speeds. By the time it sees the first slow/stopped car, a human would have started slowing way before then. Just look at the visualization next time. When you want to brake, the car doesn't visualize the light or lead car on a highway yet.
TravelFree
Active Member
The two roads where I test Red Light controls on my test route is 45mph and 35mph. I have 0 offset so that is my FSD speed. BTW- I also am set for "Chill" mode. The car displays the light at the top of the display so at that point it should begin to act on it if yellow. If I see a Yellow in the display, I feel it should begin to slow down so if it turns red say 50 ft from the stop line. It has enough room to stop. I never see it slow when the light turns green to yellow. Around here we have Yellow that lasts for 3 seconds all the way to 15 seconds. There isn't any set time. We have some roads here with traffic lights that are 25mph but none of them are on my test route.
I also saw one YT video that confirms my experience with attempt to ruin red lights for the same reason.
Overall, the worst performance I see with stop sign failure, and left and right turns from a stop sign. Going straight through from a stop sign seems pretty smooth once it gets going. How it executes is very inconsistent.
I run a video cam to record and review my experiences and one thing I observed is if the speed limit is 35 in a residential area with hard left and right bends, FSD shouldn't be obligated to quickly come up to that speed as in floor it only to slam on the brakes when it encounters the sharp turns I usually take at 10-15mph.
If it is, they seem to have paused it over the Thanksgiving holiday. I've been religiously watching Teslascope, and it seems like the amount of pending + installing + waiting for wifi + installed has been pretty fixed since Wednesday night.
If I had to guess, they aren't "blessing" new FSD beta track cars outside of business hours since they're starting to encounter issues as they push to a wider variety of vehicle types (remember Tesla does weird stuff where they tweak parts used constantly):
The people doing operations would probably be pissed if they had to deal with some issue only affecting like 5 cars made in early 2019 over their holiday.
TravelFree
Active Member
According to TeslaFi logging, it is but just a couple per day as 2021.36.8.8. Most of those went out on 11/23 to over 1200 members. My wife's Model Y hasn't received the invite yet either and she jumped from 98 to 99 back on 11/5. But both cars are under my name in the Tesla account as primary so we figure Tesla only authorized one car per account. Haven't heard anyone else having two Teslas get the FSD-beta under their account.
Anyway, I wouldn't worry about it as the beta is really flaky, unpredictable, and more of a PIA to use anyway. I only turn it on when doing specific testing for the program. If you were thinking it will make driving easier in the city, it doesn't...yet. It actually makes city driving much more dangerous, causes other drivers to believe you are drunk or stupid. Causes road rage. If Tesla didn't give me a way to drive with it disabled in the profile, I would drop out. The interesting thing is, it is so amazing to see the car drive itself in traffic which is fascinating and then it will do something really scary like slam on the brakes for no reason or steer into a head on collision with another car with no warning. When I use it for 30 minutes, I am exhausted driving, aborting, and reporting.
Knightshade
Well-Known Member
It's been "temporary" for 6 months now though... May release with 75 limit, bumped to 80 mph in June... hasn't improved since.
I've seen no evidence it has anything to do with HW3, and more likely has to do with the main forward camera having less range than radar does.... (150m vs 160m in this case).
Continental Automotive
The ARS410 realized a broad field of view by two independent scans in conjunction with the high range functions.
www.continental-automotive.com
170 meters of range on the radar in AP2.5 cars.
Vs 150 meters of range on the front main cam.
That's about 558 feet for radar, vs about 492 feet for vision...a difference of about 66 feet.
Braking from 90 mph vs 80 mph takes roughly 88 feet longer (given all else being equal), so right ballpark at least to explain the difference.
Why have you not included the narrow view forward camera? It has the same field of view as the radar but has a range of 250 meters. Is there some evidence that Tesla isn't using it?It's been "temporary" for 6 months now though... May release with 75 limit, bumped to 80 mph in June... hasn't improved since.
I've seen no evidence it has anything to do with HW3, and more likely has to do with the main forward camera having less range than radar does.... (150m vs 160m in this case).
Continental Automotive
The ARS410 realized a broad field of view by two independent scans in conjunction with the high range functions.
170 meters of range on the radar in AP2.5 cars.
Vs 150 meters of range on the front main cam.
That's about 558 feet for radar, vs about 492 feet for vision...a difference of about 66 feet.
Braking from 90 mph vs 80 mph takes roughly 88 feet longer (given all else being equal), so right ballpark at least to explain the difference.
Knightshade
Well-Known Member
Comment from @verygreen suggesting the vision-only stuff is training on the main and fisheye cams (no mention of the narrow forward cam)
Similar threads
