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

Upset with lack of features on Hardware 2.0

This site may earn commission on affiliate links.
150x the power of a Skylake (14nm) Quad Core i7 at 2.9GHz/3.8GHz and Radeon Pro 460?

Wow! Which Tegra chip are you referring to?

AP 2.0 is using the nvidia px2 full autonomy hardware. A single px2 board consists of 6 Titan X gpu's and 2 tegra processors.

The nvidia images of the px2 full autonomy suite shows two of the px2 boards, so tesla AP2 could actually be 12 titan X's and 4 tegras, im still trying to confirm that though.

The processing capability is far beyond what you can imagine.

Let me explain, the average person hasnt really had a computer in their hands that is 150 times more powerful than one of the top of the line computers on the market.

When you think of the fastest desktop or laptop you probably think of some really nice video games, maybe intense video processing, virtual reality or just writing text documents.

The computers driving AP 2.0 are 150 times more powerful than a Mac book pro.

I know you're probably thinking but Elon and the order page says it's only 40 times more powerful.

well go back and look again it actually says 40 times more powerful than before (referring to AP 1.0) which is actually pretty good news for AP 1.0 cars because that means there is still a lot of life left in AP1.0

Anyways I digress my point being, the computer running the full autonomy is capable of more than the average person can fathom at this time.

Tesla has already demonstrated that they are testing that full autonomy.

So before any of you guys assume that this round of auto pilot will follow step with the last round consider this. #1 they've already released AP1.0 gotten legislative approval for that and this new suite is significantly more powerful than the last, things are not the same this time around.

I'm confident we can expect great things from AP 2 and we can expect them to be delivered faster than they did with AP 1.

And if you're wondering how I know all of this..... I'm addicted to YouTube, I've watched every video on px2 and read every article I could find.
 
Shouldn’t be so hard on Xmen’s expecting basic ADAS features to work with a loaded MX with Hardware 2.0 for $100k+. Back in Sep, I had on order a P100DL, joined this forum to learn more about Tesla, and saw a debate going on that “AP 2.0 was imminent." At the same time, I called my Delivery Specialist, then Tesla in Fremont, and one person said, “I’m at the factory and I know nothing about it…all this is just forum speculation and Tesla has nothing to do with the TMC forum.” I nonetheless decided to bale and forfeited my $2,500 down, explaining to the Delivery Specialist that there is too much being said about AP 2.0, so I’m going to take the Forum’s word for it. In November after Elon announced HW 2.0, the DS promptly called me back and said “you were right; let’s get your order re-activated.” At this point I had learned my $2,500 lesson that everything coming out of Tesla has to be checked out, so I googled around and saw that no ADAS features would be enabled, but “you luckies are going to be on the next-gen hardware platform.” By any standard, that’s just plain weird auto-company behavior (compare to Merz, for example; they just don't play these games), so I didn’t click on “dislike” or “funny” when I read Xmen’s post. Btw, I told the local DS here in Houston, “yeah that’s great being on the next-gen hardware, but it’s buck-naked at this point, so I’m going to wait until February or March when they finally finish up the software for it.”
 
Last edited:
The missing features are no different than what you get when you order a car without autopilot.

That is not really correct.

I mean, sure, if you order and take delivery of a no-AutoPilot AP2 car, its limitations will be the same as an Enhanced/Full AutoPilot AP2 car.

But compared to non-AutoPilot AP1 car or even a Classic pre-AutoPilot Model S, AP2 cars have more limitations.

For example, AP2 cars have no automatically turning on/off headlights or rain sensing for the wipes. This is because AP2 cameras handles the light/rain sensing that AP1 and pre-AP used to do via a separate sensor. Some say not even the auto-dimming rearview mirror works (though I personally wonder about that last part as the mirror sensors should be separate from the AP system still on AP2 but who knows, maybe they improved it to use the rear camera).

This is of course temporary and is expected to be fixed during December for all or most part.
 
At this point I had learned my $2,500 lesson that everything coming out of Tesla has to be checked out, so I googled around and saw that no ADAS features would be enabled, but “you luckies are going to be on the next-gen hardware platform.” By any standard, that’s just plain weird auto-company behavior (compare to Merz, for example; they just don't play these games)

This is a bit of double-edged sword... You are of course right that no ADAS features are enabled right now, but it goes beyond what I guess people refer to as ADAS nowadays: there is not even automatically turning on/off headlights (and I'm not just talking about automated high-beam, though that too is disabled of course, all of it is disabled) nor are there automated windscreen wipers until AP2 gets enabled. So, indeed, the situation is quite archaic at the moment.

You are also right that the established premium manufacturers would not do it like this in general. What they would do, in general, is ship "AP1" features until "AP2" features are available and then switch to shipping those. There are some extreme examples to the contrary: after the Fukushima disaster a sole supplier of rear-view cameras for some of these companies had to shutter doors (for a while at least). Some German manufactures in fact did, after this, ship cars without rear-view cameras that were installed after delivery, once supply was available. There are also some milder examples of upgrading multimedia system softare after the fact... But yeah, not for ADAS, not really unless to apply a recall-like safety update.

However, and this is the part that IMO makes this a double-edged sword:

Mercedes or the like would never, at least until their business model on this changes, ship feature-upgrades to an existing ADAS system either. They will simply ship their current Level 2 offering and once Level 3 is available, ship that. Those with Level 2 have no means of upgrading, not even within the scope of their hardware. When Level 4 ships, Level 3 is left out doing simply what it was doing before that.

That's fine when the feature is so mature that it has maxed out its capabilities. But this is not really true for the Germans either, they do small improvements all the time, while leaving older owners out. One small example is the Audi blind spot detection. Back in, I guess around model year 2010 or 2011, Audi upgraded this BSD functionality to work from 30 km/h upwards. Previously it worked from 60 km/h upwards. (Euro models.) Obviously a major improvement in practical terms because it made the system usable in city traffic and one that, really, was most likely just about improved software and validation. Both had radars in the rear corners, I doubt any hardware changed. But if you had the older model year, there was no update for you...

In the end, this means there are entire fleets of German premium cars out there that actually have far superior ADAS related sensory coverage than even an AP2 Tesla will have, limited to only a small subset of what they could do. They have more radars, more cameras - and soon they will have ample CPU power too. And yet, when what 2018 Audi A8 with Level 3 capability ships next year (with even more sensors), it will very likely (unless Audi changes their model to Tesla's) be limited to driving at Level 3 in slow highway traffic for its lifetime simply because that is what it was validated to do at the time of manufacturing.

So all this hovers back to the question, if Tesla's software upgrade model is - I would assume most agree - superior for the customer in the longer run, how early should they apply it - in this case, assuming the software won't be ready until December 2016. If the AP2 hardware was ready sooner, should Tesla have shipped it in, say, March/April for the Model S facelift? Or even last year with Model X? Or not until AP1 features can be matched? It depends on the customer, of course, I would wager a lot people would be happy taking the hardware as soon as possible - but at the same time there are others that would definitely not be OK with that.

In the end, it seems Tesla shipped AP1 hardware and AP2 hardware once their HW development and supply was ready. AP2 hardware seems to have coincided with the volume availability of Drive PX2. This did in fact cause an unfortunate period of limited functionality as they wait for the software to mature and pass validations.
 
This is a bit of double-edged sword... You are of course right that no ADAS features are enabled right now, but it goes beyond what I guess people refer to as ADAS nowadays: there is not even automatically turning on/off headlights (and I'm not just talking about automated high-beam, though that too is disabled of course, all of it is disabled) nor are there automated windscreen wipers until AP2 gets enabled. So, indeed, the situation is quite archaic at the moment.

You are also right that the established premium manufacturers would not do it like this in general. What they would do, in general, is ship "AP1" features until "AP2" features are available and then switch to shipping those. There are some extreme examples to the contrary: after the Fukushima disaster a sole supplier of rear-view cameras for some of these companies had to shutter doors (for a while at least). Some German manufactures in fact did, after this, ship cars without rear-view cameras that were installed after delivery, once supply was available. There are also some milder examples of upgrading multimedia system softare after the fact... But yeah, not for ADAS, not really unless to apply a recall-like safety update.

However, and this is the part that IMO makes this a double-edged sword:

Mercedes or the like would never, at least until their business model on this changes, ship feature-upgrades to an existing ADAS system either. They will simply ship their current Level 2 offering and once Level 3 is available, ship that. Those with Level 2 have no means of upgrading, not even within the scope of their hardware. When Level 4 ships, Level 3 is left out doing simply what it was doing before that.

That's fine when the feature is so mature that it has maxed out its capabilities. But this is not really true for the Germans either, they do small improvements all the time, while leaving older owners out. One small example is the Audi blind spot detection. Back in, I guess around model year 2010 or 2011, Audi upgraded this BSD functionality to work from 30 km/h upwards. Previously it worked from 60 km/h upwards. (Euro models.) Obviously a major improvement in practical terms because it made the system usable in city traffic and one that, really, was most likely just about improved software and validation. Both had radars in the rear corners, I doubt any hardware changed. But if you had the older model year, there was no update for you...

In the end, this means there are entire fleets of German premium cars out there that actually have far superior ADAS related sensory coverage than even an AP2 Tesla will have, limited to only a small subset of what they could do. They have more radars, more cameras - and soon they will have ample CPU power too. And yet, when what 2018 Audi A8 with Level 3 capability ships next year (with even more sensors), it will very likely (unless Audi changes their model to Tesla's) be limited to driving at Level 3 in slow highway traffic for its lifetime simply because that is what it was validated to do at the time of manufacturing.

So all this hovers back to the question, if Tesla's software upgrade model is - I would assume most agree - superior for the customer in the longer run, how early should they apply it - in this case, assuming the software won't be ready until December 2016. If the AP2 hardware was ready sooner, should Tesla have shipped it in, say, March/April for the Model S facelift? Or even last year with Model X? Or not until AP1 features can be matched? It depends on the customer, of course, I would wager a lot people would be happy taking the hardware as soon as possible - but at the same time there are others that would definitely not be OK with that.

In the end, it seems Tesla shipped AP1 hardware and AP2 hardware once their HW development and supply was ready. AP2 hardware seems to have coincided with the volume availability of Drive PX2. This did in fact cause an unfortunate period of limited functionality as they wait for the software to mature and pass validations.

While I don't disagree with what you said, you're assuming that Tesla moved to the new hardware voluntarily at a point in time they chose in advance. I don't believe this is the case.

We saw a bunch of first hand reports this past summer of a serious falling out between Tesla and Mobileye, and it was stated outright that one of the consequences of that was that the contract between the two would not be renewed.

I think we have ample evidence to conclude that Tesla was planning another generation with Mobileye - look at the blanked out space and disconnected, installed wiring on every X, and the schematic for a 2016 S with triple camera we saw this summer.

Once that spilt started happening, Tesla suddenly had a deadline - if they didn't ship their own hardware, they'd have to deliver cars without any Autopilot. So they rushed to get the hardware out the door, and didn't have time to develop the software, especially the neural networks.

It's remarkable that they managed to accomplish that at all - Mobileye was clearly convinced Tesla couldn't (based on the demands that Mobileye made) - and in a few weeks AP2 cars will be back to Tesla's preferred standards of integration/automation via firmware update.
 
Last edited:
  • Informative
  • Like
Reactions: jgs and OBX John
While I don't disagree with what you said, you're assuming that Tesla moved to the new hardware voluntarily at a point in time they chose in advance. I don't believe this is the case.

I didn't mean to make any statements regarding the timing of Tesla's reasoning (beyond availability of Drive PX2 anyway). What you say is certainly possible.

That said, my takeaway from some messages we heard through the grapevine was that Tesla had until the end of the year with Mobileye. It is possible Tesla could have waited a month or two more than they did, which would have allowed maturing the software.

But I would agree we can not know. The reason you state is possible. But equally, wanting to launch the new hardware as soon as possible to avoid any prolonged Osborne effect is plausible too (as the Mobileye spat had become public and may have affected some sales).

That said, there is another angle to look at this from: gathering of real-life data. How much, if any, Tesla's desire to use the early AP2 fleet for fleet-learning plays a role? Does it significantly help Tesla's validation to launch the hardware (with assumed data gathering) before the software? If so, it is possible this always would have been the ramp-up style, no matter other timing questions. That said, the time between shipping of AP2 hardware and estimated shipping of AP2 software is only perhaps a month, so maybe this isn't the reason...

Perhaps it is a mixture of reasons, some decisions which could have been made differently too, but in the end was the compromise they settled on.

p.s. As for speculation about a Mobileye based successor to AP1 I definitely agree and posted my analysis/speculation here Model X mule(s) show signs of nVidia Tegra X1 Drive PX platform - no rear mirror! and here Is Model X (or was?) getting an "AP 1.5" before full-autonomous "AP 2.0" suite upgrade? earlier this year.
 
Last edited:
Tesla has done this same thing before with AP1.0. They released the hardware a year before providing AP1.0 features, presumably, as @AnxietyRanger said, to collect data for AI training. While the Mobileye spat may have played a small part, I suspect Tesla was already moving in this direction long before Mobileye became aware of it.

Remember, even if you don't pay for EAP, all Tesla's with AP2.0 hardware are collecting data for AI training, so it would seem that in this particular situation, its most efficient to release hardware before software.
 
About 30,000 a year don't manage to stay alive. We've both just been lucky. So far...
I can't believe how much more focused on the car (not road) I am when AP/TACC is activated. When the car in front is stopped I am constantly thinking is it going to stop itself in time? Few times it did not and I had to override the brake. I am not risking my whole family life to test it out. Even if it stopped in time it would have been a very hard brake.

Then on the freeway, same thing, is the car going to keep in the lane at the next curve? Few times it did not and I had to override or it will swirl into adjacent cars. Then there's the truck thing.

The most annoying thing is when you are at an intersection, the car in front is going to enter the left turn lane. The car slows down significantly until the car is completely out of the way. By then the go-straight light already turned yellow/red. I lost count of how many times this has happened to me. So I normally accelerate myself now.

All these things I used to drive by instinct, now have to be "manually" processed by my brain again, taking the car behavior into another variable. I think that's why the fatality rate is lower when AP is on. I know some might not like what I say about AP but this was the case for me and is my honest opinion.
 
Is that what I should do?
I see that was presumptuous of me. My point was that reading through your earlier post it seemed like using AP1 irritated you ("the most annoying this is...") and even scared you ("not risking my whole family life"). If I had something that irritated and scared me as much as your post sounded to me, I would probably not use it unless I was forced to. But maybe I misinterpreted your level of hatred and/or fear of AP1... and in any case, you do you.

As for whether they can do better with AP1, I guess we'll find out. Would be nice!
 
There are tons of stuff that annoy me and I have came to work around them. If every time I get annoyed I shouldn't use it, the car would be staying in the garage forever. I did mention that because of AP I have came to be more aware and focused, and that I'd accelerate manually to overcome the most annoying thing.

When the first 8.0 rolled out the USB player was a risk to use when you are driving. They finally added alpha search on the right to remedy it. It's a car that drives at high speed on the road. Of course every minor thing could pose a significant risk.
 
  • Like
Reactions: jgs
About 30,000 a year don't manage to stay alive. We've both just been lucky. So far...

I can't believe how much more focused on the car (not road) I am when AP/TACC is activated. When the car in front is stopped I am constantly thinking is it going to stop itself in time? Few times it did not and I had to override the brake. I am not risking my whole family life to test it out. Even if it stopped in time it would have been a very hard brake.

Then on the freeway, same thing, is the car going to keep in the lane at the next curve? Few times it did not and I had to override or it will swirl into adjacent cars. Then there's the truck thing.

The most annoying thing is when you are at an intersection, the car in front is going to enter the left turn lane. The car slows down significantly until the car is completely out of the way. By then the go-straight light already turned yellow/red. I lost count of how many times this has happened to me. So I normally accelerate myself now.

All these things I used to drive by instinct, now have to be "manually" processed by my brain again, taking the car behavior into another variable. I think that's why the fatality rate is lower when AP is on. I know some might not like what I say about AP but this was the case for me and is my honest opinion.

Think about how terrible the average driver is. 50% of the population is worse than that!

We can't prevent everything but I think what really improves the survival odds beyond defensive driving is just being able to anticipate
what other drivers might do, including potentially risky actions. My Tesla is in production so I don't have AP experience to comment if it has any predictive logic or if it can only _react_ through the sensor technology.

My thoughts on what the AP use cases for me would be:

1 - Reduce fatigue on bumper to bumper highway traffic in SoCal.
2 - Reduce fatigue and speeding tickets by letting AP handle the highway routes to Vegas, Phoenix

Everywhere else I plan to drive because I just enjoy the physical aspect of driving.

AP doesn't also know to automatically launch a car from a red when an idiot in a civic hatchback with coffee can exhaust and shopping cart spoiler rolls up next to you blasting music. ;)