-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta 10.69
- Thread starter Buckminster
- Start date
Daniel in SD
(supervised)
Nice that they are spending so much time with Chuck. As a newer user of FSD and AP, I have had some real problems with both products as enumerated in my recent posts. I would like to see the PB matter resolved! If highway use was more reliable (And it certainly should be at this point), I would resubscribe. I am sure I will test 10.69 out of curiosity but when we received the current iteration a while back, it felt so much better than the previous version but then seemed to get worse over time. Not sure how that could be but the perception is it started off smoother and towards the end when we cancelled, it was so bad it was un-usable, at least for us.
Presumably this single instance is a concrete example for the team to focus on to solve the more general problem of unprotected left turns. It seems like there were multiple architectural changes necessary to get it closer to passing this test and hopefully makes the system more reliable and capable, e.g., understanding how cross traffic will behave from further away.
The "single stack" highway driving goal for FSD Beta 11 has been around for quite some time, and unprotected left turns could have taken away resources that could have brought more reliable highway driving sooner as controlled access highways don't have cross traffic. But then again, Navigate on Autopilot sometimes does work on highways that do have infrequent cross traffic that FSD Beta will need to support, and similarly entering/crossing these highways from a side street is something FSD Beta will need to handle and probably made more reliable with 10.69.
Highway usage of FSD Beta has been in shadow mode, so Tesla is actively making progress with single stack and indeed quite likely the reliability isn't exceeding the current highway-only stack. As you suggest, phantom braking is probably a main issue, and you can look for improvements that reduce false positives.
Was it really necessary to so quickly start a separate thread? We have now have a mess with two threads to watch. Would have been much better to ask the forum moderator to see if the initial 10.13 thread could have been renamed.
This wouldn't be the first time versions got skipped for the public FSD Beta audience. I believe 10.7 was an internal release before becoming 10.8 for the holiday update, so most testers went from 10.6 to 10.8.
Similarly, there had been previous wider releases that were partially rolled out to say 10% or 30% of the audience then launching the subsequent release. Maybe Tesla has enough internal testers to not need to stagger the wider release as much anymore or received too much noise from people asking why they haven't received the update yet?
Clearly 10.13 is safe enough to not roll back the software for those who do have it, but I guess it's preferable for Tesla to leave most on 10.12 until 10.69?
Similarly, there had been previous wider releases that were partially rolled out to say 10% or 30% of the audience then launching the subsequent release. Maybe Tesla has enough internal testers to not need to stagger the wider release as much anymore or received too much noise from people asking why they haven't received the update yet?
Clearly 10.13 is safe enough to not roll back the software for those who do have it, but I guess it's preferable for Tesla to leave most on 10.12 until 10.69?
At least we have new maps to play with. Look forward to seeing if that solves some of the bad lane choices in my area
My local freeway was under construction during the previous map release, so it’s been tagged as a 55mph speed limit on my Tesla since then. Been super annoying trying to use AP/NoA lol
What about the chance that quickly installing updates when offered is one of the criteria for being accepted into the beta?this.
I’m still on 2022.16.2 in the offest of off chances that .3 would bump me too high.
Mini-max is very clear here. Your potential pain in updating beyond the next release is way too great compared to the teeny benefits in 2022.20.
Why does anyone still believe the *sugar* he says on social media lol, the guy is just trolling
Looks like we've hit some throttled 4.5Mbps upload limit to Tesla's server but even with most of fleet not on the latest 10.13/10.69 version, they still send back 20GB of video snapshots most likely to autolabel for perception:
To be clear, we can upload around 20GB per vehicle each day, and recently the combined upload speed of the vehicles to Tesla servers is limited to 2.0GB/hr while our regular internet upload, e.g., video conferencing, is unchanged. Not clear if it's from Tesla's end or our ISP for that specific endpoint.
But in any case, there's hope that our continued driving with 10.12 is useful data for Tesla.
To be clear, we can upload around 20GB per vehicle each day, and recently the combined upload speed of the vehicles to Tesla servers is limited to 2.0GB/hr while our regular internet upload, e.g., video conferencing, is unchanged. Not clear if it's from Tesla's end or our ISP for that specific endpoint.
But in any case, there's hope that our continued driving with 10.12 is useful data for Tesla.
I notice about 7 GB/day upload when the car is unused, almost certainly sentry mode video. When the car is driven using FSD beta, there's from 12 to 20 GB of data is sent up, including the sentry mode video. 22 GB is the biggest day in the last month.
I went on vacation recently, driving 2100 miles, but see no huge upload after I got home. There was no WiFi connection during that time. I wonder if the car uploaded FSD video data on cellular while on the road? Maybe it trashes the data if there's no WiFi connection after some period of time.
KArnold
Active Member
Sentry data is uploaded to Tesla? Or to your own server? If the former, I wonder why Tesla would want that? I may have missed that tidbit.
So not only so they know my every move, they have a video of everyone who has gotten into my car? Hmmm.
I don't continuously access sentry mode video, so I can't say. But, I noticed that turning off sentry mode caused the upload rate to drop to almost zero in a test I did a while back. Perhaps the video is sent up to the server for buffering before going to the app. But why it would do so when the app is not accessing is curious.
Similar threads
- Article
