New FSD details

[selfbp]

I just don’t understand how it all came to be this way. They kicked butt bringing out these awesome traits that made Tesla what it is today. And here recently it’s more relatable to this kind of scene

And I know they’re trying, but sometimes it’s really hard to accept that nobody in authority positions realized these issues before this all gets cranked out the grape vine...

 

[michidragon]

Oddly enough I'm *less* likely to buy FSD than I was before, now that I know Model 3 owners got a break due to the new AP delineations, and we don't.

Maybe for some reason they'd rather S/X owners not buy FSD.

 

[Tim-in-CA]

According to what folks are seeing in their accounts now, looks like if you have a Model 3 with EAP, you can upgrade to FSD for $3000; but if you have an S or an X with EAP, it still costs $5000 to upgrade to FSD. (Some Model 3 guys have already purchased it tonight, according to the reddit sub.)

M3 owners have it showing for $3k in their accounts, but all S/X owners still have it showing $5k.

So looks like the "add the missing features for $3000" thing that @Troy mentioned applies to Model 3 only. S/X EAP owners will still have to pay $5000 even though some of those features we have were rolled into FSD.

What the business justification for this is, I have no idea. Can't think of anything besides "We think S/X owners can afford it", or something.

This discrepancy, and the who-knows date on Sentry Mode for S/X sure does make me feel like S/X owners are being deprecated or pushed to the side.

That’s a huge bummer. I’d purchase FSD for 3K, 5K seems like gauging seeing as EAP already includes several of the “new” FSD features. 2018 build here.

 

[MIT_S60]

Oddly enough I'm *less* likely to buy FSD than I was before, now that I know Model 3 owners got a break due to the new AP delineations, and we don't.

Maybe for some reason they'd rather S/X owners not buy FSD.

On the bright side, Tesla saved us from spending $3k tonight by leaving the S/X upgrade price at $5k

 

[michidragon]

Maybe one of those fancy YouTube tesla bloggers can bring up a "What gives" about this... I know some of them are on this forum.

 

[BinaryField]

Is it possible that the $5000 FSD upgrade price is simply an oversight, and the number has not yet been updated? It wouldn't be the first time this type of error has happened.

 

[MIT_S60]

Is it possible that the $5000 FSD upgrade price is simply an oversight, and the number has not yet been updated? It wouldn't be the first time this type of error has happened.

Wouldn't it be more likely the $3k FSD upgrade price is a glitch than the $5k being an oversight?

 

[whitex]

Autopilot is now only $3000 but no longer includes Nav-On-Autopilot and the self-parking. If you want those options, you'd have to add on the FSD which is another $5000.

FSD option also says traffic light recognition coming later this year .

View attachment 381624

Did you notice that features to come later this year are going to come before the features listed above (which most people would interpret as available now, which is obviously not the case)? Also:
"Summon: your parked car will come find you anywhere in a parking lot. Really."
Sounds really good, right, though not part of features for later this year, so probably in a few years, but then there is more:
"features require active driver supervision and do not make the vehicle autonomous."
There is the catch, the car will find you anywhere in the parking lot, as long as you are in it or right next to it to supervise it. So yes, it will find you almost immediately every time - if it doesn't, then you're not using it as intended by being inside the car or within an arm's reach.

Being a long time customer, I learned how to read their marketing. Remember 691hp car (which customers were supposed to have known that it was just a number for the motors, but the car would be limited by the battery to 463hp).

 

[michidragon]

Wouldn't it be more likely the $3k FSD upgrade price is a glitch than the $5k being an oversight?

If it is, a bunch of folks just bought it, so I don't know how that'll turn out..

 

[ricebucket]

This EAP -> FSD change really sounds like a tacit acknowledgment that true FSD is a long ways off (i.e. fully level 5). So instead they moved some features from EAP into FSD just so that FSD buyers would actually have something to show for what they paid. (Yes, I know there's the traffic light stuff too, at a future date, but again, this is not even close to what people would consider true autonomy.)

 

[S4WRXTTCS]

Is it possible that the $5000 FSD upgrade price is simply an oversight, and the number has not yet been updated? It wouldn't be the first time this type of error has happened.

I completely agree.

Elon Musk himself tweeted that they were going to look into it over the next couple days. People need to give it a few days to settle out.

I expect to see that the upgrade from AP2.5/MCU2 on a Model S/X will be $3K like it is for the Model 3.

I don't know what Tesla will do about the people with AP2/MCU1, or AP2.5/MCU1.

Those could cost more as they have differences that the newer Model S/X, and the 3 don't have. Those configurations are probably why Tesla needs a few days to sort things out.

 

[michidragon]

I completely agree.

Elon Musk himself tweeted that they were going to look into it over the next couple days. People need to give it a few days to settle out.

I expect to see that the upgrade from AP2.5/MCU2 on a Model S/X will be $3K like it is for the Model 3.

I don't know what Tesla will do about the people with AP2/MCU1, or AP2.5/MCU1.

Those could cost more as they have differences that the newer Model S/X, and the 3 don't have. Those configurations are probably why Tesla needs a few days to sort things out.

So, I guess that makes sense. "Model 3" is invariably AP2.5/MCU2(3); whereas Model S/X is all over the place; so maybe you're right...

Of course, then there's the thought though, that because S/X varies so much they may just call it $5000 across the board, whether it's AP1/2/2.5 MCU1/2.....

 

[S4WRXTTCS]

This EAP -> FSD change really sounds like a tacit acknowledgment that true FSD is a long ways off (i.e. fully level 5). So instead they moved some features from EAP into FSD just so that FSD buyers would actually have something to show for what they paid. (Yes, I know there's the traffic light stuff too, at a future date, but again, this is not even close to what people would consider true autonomy.)

A few days ago a few of us were predicting that EAP would disappear. Or at least we wanted it to as in my case.

The reason for this is it was getting in the way of the iterative nature that needed to happen to progress towards FSD. The way to think of FSD is it's the destination. It doesn't really have "features" as you either have unsupervised driving or you don't. Sure you can limit the type (L3, L4, etc), and you can limit the speed (like the Audi A8 in Germany for L3).

The stuff that got moved was things that HW3 will have the most impact on. They needed room to grow.

What's funny about the timing is we never got to see what Smart Summons, Smart Parking, or what unconfirmed NoA was really like before this announcement. Before I even have a chance to enjoy those things I'll know that there is an available upgrade.

Now I hope those things work great. So great that Tesla is simply putting the final touches on it before only doing maintenance like releases for EAP. That way all those customers are satisfied, and they won't worry about what HW3/FSD customers get.

The FSD team needs to be free of legacy stuff. Having to support legacy stuff is like having an arm tied around your back. It's really hard to innovate like that.

By moving things around they have room to grow.

First there is sign detection, and then I imagine they'll add debris detection followed possibly by pothole detection.

I think this progression in both SW, and HW will continue over many cycles before we have anything that resembles the promised cross country FSD.

 

[iTech]

I completely agree.

Elon Musk himself tweeted that they were going to look into it over the next couple days. People need to give it a few days to settle out.

I expect to see that the upgrade from AP2.5/MCU2 on a Model S/X will be $3K like it is for the Model 3.

I don't know what Tesla will do about the people with AP2/MCU1, or AP2.5/MCU1.

Those could cost more as they have differences that the newer Model S/X, and the 3 don't have. Those configurations are probably why Tesla needs a few days to sort things out.

MCU has no bearing on AP. Swapping out the AP2/AP2.5 computer would be the same procedure.

 

[S4WRXTTCS]

MCU has no bearing on AP. Swapping out the AP2/AP2.5 computer would be the same procedure.

But, there are dependencies on bandwidth and being able to save. I don't know if MCU1 can handle saving a dash cam video from all the cameras while also doing the other functionality.

Plus there is Validation and Verification of the SW where you have to do it on both platforms. It's a very different platform in terms of x86 versus Arm. Plus there is likely so few AP2.5/MCU1 so why bother? Simply upgrade those people to AP3/MCU2.

There is likely a lot of things that a FSD car needs to do where we can't think of it as just the AP computer. It's everything in the car working together where they have different jobs to do.

In fact one of my reservations about FSD is the fact that the navigation right now isn't really all that good. Can the MCU1 hardware support a really good offline navigation system? I dunno.

Sure it's still much simpler than figuring out what to do with AP2.

Who knows what they're going to do about that one.

The cameras are different (the filter on the sensor is different)
The radar is different
It doesn't have the wiring redundancy that HW2.5 has.

I'm really, really curious to see what Tesla does with a AP2 owner who bought the FSD package. That person is going to get the FSD upgrade, but it remains to be seen how much Tesla will update.

I wouldn't be surprised if they had to update everything. If they don't there will be a lot of angry SW developers at Tesla for making them go through hell to support different sets of equipment that have non-trivial differences.

When something is against a rock and a hard place you never know which side is the one to give.

 

[Axael]

The way I understand it, it will work like this. If I find out new information, I will update this chart and post the updated version here.

What is your thought about the fact that EAP was advertised as being able to do auto lane changes without any driver input?

 

[emmz0r]

A few days ago a few of us were predicting that EAP would disappear. Or at least we wanted it to as in my case.

The reason for this is it was getting in the way of the iterative nature that needed to happen to progress towards FSD. The way to think of FSD is it's the destination. It doesn't really have "features" as you either have unsupervised driving or you don't. Sure you can limit the type (L3, L4, etc), and you can limit the speed (like the Audi A8 in Germany for L3).

The stuff that got moved was things that HW3 will have the most impact on. They needed room to grow.

What's funny about the timing is we never got to see what Smart Summons, Smart Parking, or what unconfirmed NoA was really like before this announcement. Before I even have a chance to enjoy those things I'll know that there is an available upgrade.

Now I hope those things work great. So great that Tesla is simply putting the final touches on it before only doing maintenance like releases for EAP. That way all those customers are satisfied, and they won't worry about what HW3/FSD customers get.

The FSD team needs to be free of legacy stuff. Having to support legacy stuff is like having an arm tied around your back. It's really hard to innovate like that.

By moving things around they have room to grow.

First there is sign detection, and then I imagine they'll add debris detection followed possibly by pothole detection.

I think this progression in both SW, and HW will continue over many cycles before we have anything that resembles the promised cross country FSD.

Room to grow? They still have to honor older hardware with NoA.
Developers have multiple branches and they merge stuff back in if there's a fix for everything etc.

 

[Troy]

What is your thought about the fact that EAP was advertised as being able to do auto lane changes without any driver input?

Yes, that's what Tesla advertises for Enhanced Autopilot. I just took the screenshot below on the Autopilot page here. It says "Automatically change lanes without requiring driver input" under the title "Enhanced Autopilot". They used the exact same phrase in October 2016 when Autopilot 2.0 was first introduced. See my message here from Oct 2016.

My guess is, this phrase refers to the fact that you don't need to do anything after initiating the lane change. I think you are connecting this phrase to the upcoming Auto Lane Change feature that won't require confirmation. However, Tesla has been using this phrase since Oct 2016 to describe what is available now. Therefore I will separate these two and comment on both:

1. What I think about Tesla using the phrase "Automatically change lanes without requiring driver input" under the title "Enhanced Autopilot" on the Autopilot page here since October 2016:

Exaggerating Autopilot capabilities with these kinds of misleading marketing statements is dangerous because people can get rear-ended. The important thing that they were supposed to emphasize is that the car was not using the rear cameras in 2016 or 2017 and drivers were supposed to look back to confirm that the target lane is clear. Even today, I don't think Auto Lane Change can avoid getting rear-ended if there is a car approaching at high speed from behind on the target lane. Based on my calculation, the rear camera's range needs to be 350 meters instead of the current 100 meters to make sure the lane is clear. However, currently, I don't think Autopilot uses even the existing 100 meters rear visibility to check the target lane. I don't think they use anything else but the ultrasonic sensors that can see 1 or 2 car lengths behind.

2. What I think about Tesla planning to release 'Navigate on Autopilot' with 'Auto Lane Change Without Confirmation' that was mentioned in a blog post here:

I don't think the car can do Auto Lane Change Without Confirmation if the target lane appears clear because it would require making sure that there is no bullet car approaching from behind. That would require 350 meters rear-visibility. However, Auto Lane Change Without Confirmation is possible if the target lane is already occupied but there is enough gap to change lanes. If there is already a car on the target lane behind you, you don't have to worry about a bullet car because if there was one, it would hit this other car behind you and not you when you change lanes. So, I would expect them to release it this way.

However, I think you are asking about my chart because it shows 'Auto Lane Change Without Confirmation' under FSD and not EAP even though Tesla's blog post says this feature will be available in future versions of Navigate on Autopilot which was part of Enhanced Autopilot at the time. I decided to do it this way because Elon has now separated Navigate on Autopilot into two parts: City and Highway. The city version is part of FSD while the highway version will be available to EAP users. To me, it looks like the city version of Navigate on Autopilot will need AP3.0 hardware. Therefore I guess 'Auto Lane Change Without Confirmation' will also require AP3.0. However, I could be wrong.

 

[diplomat33]

What is your thought about the fact that EAP was advertised as being able to do auto lane changes without any driver input?

Auto lane change without driver input is part of NOA which is still part of EAP for current owners. So Tesla has kept their promise on this.

 

[diplomat33]

FSD upgrade for EAP owners is only $3000 now. Yes! I am very happy!