Elon, Where is the FSD features you promised?

[whitex]

Did you read the quote? They aren't going to do any retrofits at all unless it is determined that HW 2.0 can't achieve FSD.

How and when is that determination going to be made? Is there a deadline, like 2019, 2020, 2025? Or is it like FSD today, some day in the future. Even if FSD is delivered on some new hardware, Tesla can claim it's still coming for AP2.0. Once the warranty on AP2.0 runs out on all cars, they can just claim they are all broken and require a fix which consists of a new AP hardware and sensors, of course after warranty at the cost of the owner.

 

[J1mbo]

This was the statement that interested me:

This hardware set has some added computing and wiring redundancy, which very slightly improves reliability, but it does not have an additional Pascal GPU.

If they have added compute redundancy without another GPU then have they added another Parker? Would make sense if they they were having a problem with the volume of I/O.

Personally not too bothered about 2.5. The real step change wil be when they move to Xavier.

 

[lunitiks]

What's "wiring redundancy"...?

 

[AnxietyRanger]

What's "wiring redundancy"...?

Intentional vagueness?

 

[NerdUno]

Maybe, but for your purposes, this means going to an SC to ask is worthless because they don't have anything set up to do retrofits. They will only do so after they make the determination FSD can't be done on HW 2.0.

Cheer up. We could all be dead by then.

Anyway it's worth pointing out that there is no "EAP" - as in Enhanced autopilot - at this point. It has no new features and does no better than my AP1 mobileye system.

Now THAT'S what I call UNDERSTATEMENT.

 

[BlueRocket]

Wouldn't it be ironic if I were in fact EM? Of course, I'm not, but that would be a funny twist.

That would explain a lot... Nothing getting done on AP and FSD because you're too busy posting on TMC

 

[lunitiks]

Now THAT'S what I call UNDERSTATEMENT.

Yes. That's why I think it's actually short for Emergency Action Plan.

Or Early Access Program / Experimental Approval Process.

All the while Musk led us to believe it was Easy As a Pie

FSD? Well that's either Future Software Development, Forget Self-Driving and/or Future Settling Decision. Dunno what's worse. Let's just hope it doesn't stand for Full Scale Dysfunction

 

[bob_p]

What Tesla has just done is create the perception that current Model S/X EAP/FSD is inferior to the EAP/FSD on the new Model 3, which costs 1/2 the price of a Model S.

The Tesla developers may believe they have sufficient processing power to implement FSD on the AP2 processor, but that belief is based on early prototype and development testing - and there is a strong possibility that more horsepower could be needed to meet the FSD goals.

Until they have FSD operating - they really can't be sure how much processing power will be needed.

Tesla should address this issue for customers with AP2 and FSD already purchased. We took a risk on Tesla by pre-purchasing a feature that may not be available for years (and possibly never).

 

[daktari]

It's a matter of cost/risk. Tesla doesn't use all processing power now for EAP. Actually they probably don't know how much processing power they will need in the end for 100% FSD.

If they think AP2.0 hardware might be enough, but they know AP2.5 is much more likely to be enough. And AP2.5 is a few hundred $ more expensive. Then it's an easy question and cheap insurance.

It's cheaper to replace maybe 20k cars later, than eventually having to replace all 500k cars (with M3 ramp-up numbers and all).

I don't know how much more costly AP2.5 is, but if the cost is considerable, it will be a huge expense for the company with 500.000 Model 3 a year. I bet they either pay the same/lower for new HW or have a real good belief the upgrade is needed.

 

[BigD0g]

Guys guys guys, don't get hung up on 2.5, I'll bet large sums of money that 3.0 will be needed with an additional pascal cpu before this show is over. Tesla is not going to upgrade crap until they "actually" have it working and know what is required. They may even skip the dual px2 board and jump to xavier by the time they get this thing sorted out.

I would not get hung up on 2.5, this path is nowhere near built yet. I wouldn't be concerned at all, unless we start seeing new models with additional radars, at that point I declare 2.0 is screwed, however as long as they are simply updating / upgrading the APE box, we're all safe, well the ones that prepaid for FSD anyways.

 

[stopcrazypp]

This was the statement that interested me:



If they have added compute redundancy without another GPU then have they added another Parker? Would make sense if they they were having a problem with the volume of I/O.

Personally not too bothered about 2.5. The real step change wil be when they move to Xavier.

Guys guys guys, don't get hung up on 2.5, I'll bet large sums of money that 3.0 will be needed with an additional pascal cpu before this show is over. Tesla is not going to upgrade crap until they "actually" have it working and know what is required. They may even skip the dual px2 board and jump to xavier by the time they get this thing sorted out.

I would not get hung up on 2.5, this path is nowhere near built yet. I wouldn't be concerned at all, unless we start seeing new models with additional radars, at that point I declare 2.0 is screwed, however as long as they are simply updating / upgrading the APE box, we're all safe, well the ones that prepaid for FSD anyways.

Yeah, I really doubt Tesla won't move to Xavier when that is ready (boards ready for testing near end of this year, production probably sometime next year). It has the same power as the full PX2, but with much less energy consumption.

 

[S4WRXTTCS]

What's "wiring redundancy"...?

I'm a little confused by why you're confused.

In electronics failures with wiring/connectors is a significantly percentage of the overall failures.

So they need some redundancy. They likely found areas they could improve in and did so.

I still haven't seen anything in detail about what changed. I think people are assuming it's way more than what it actually is.

 

[whitex]

I'm a little confused by why you're confused.

In electronics failures with wiring/connectors is a significantly percentage of the overall failures.

So they need some redundancy. They likely found areas they could improve in and did so.

I still haven't seen anything in detail about what changed. I think people are assuming it's way more than what it actually is.

Wiring redundancy is great, but what about other redundancies? I would expect that sensors or computers die more frequently than wiring unless there are some mechanical fatigue issues. I wonder what mode of failure the "wiring redundancy" is addressing, unless of course all it's addressing is marketing/PR.

 

[electracity]

Redundancy in a driving system is not mostly about fail-over, which enhances up time It's about agreement between logically separate systems. In this scenario that may mean different sensors evaluated by separate processes on the same cpu.

An ideal level 5 car probably should have the capability to still move with a partial failure. The shoestring budget of HW2.5 makes this sort of redundancy unlikely.

I think HW 2.5 is probably just a less expensive more refined board. AP2 was a rush job and the board design may have included mobileye tech that was hurriedly removed.

 

[boonedocks]

...AP2 was a rush job and the board design may have included mobileye tech that was hurriedly removed.

I will take some of that Mobileye Tech that may have been included HW1 AP1 worked so well.

HW2 EAP is just pathetic in its current FW incarnation. Unusable with my family in the car (I feel it is actually a danger much less makes everybody think I am not paying attention with all of its swerving and hard braking)

 

[tftf]

Personally not too bothered about 2.5. The real step change wil be when they move to Xavier.

What will happen after Xavier? Nvidia is not asleep.

As we all know, Xavier will ship (at least samples) in late 2017.

There will be new major advances until 2020, Xavier "1.0" will be outdated by 2020 as well given the advancements from 2015 to 2017:

NVIDIA Teases Xavier, a High-Performance ARM SoC for Drive PX & AI

Why Tesla ever shipped HW with "FSD features" since late 2016 with FSD still years away is beyond me.

Tesla will very likely have to exchange and update the HW (at least), maybe even add sensors or cameras depending on upcoming regulations in large car markets around the world.

Good luck getting all of this (especially retro-fitting HW!) into cars shipping since late 2016.

Even if it's possible, it opens a can of worms due to the sheer numbers of cars and costs involved.

Call me a short and a pessimist, but these FSD promises smell like a catalyst for many angry/duped customers as well as many lawsuits....

 

[Bebop]

Everything is going to be okay. I don't think we should worry.

The nuclear option is essentially Tesla having to recall a bunch of AP2.0/FSD cars and retrofit them with the proper hardware for actual FSD capability. Even to this day on the new order page, these cars are being sold to you with plain text that says it is capable of FSD.

If we find out in fact that these cars are no longer capable of FSD, then they will be in such heap of crap and here come the class action lawsuits. Which I'm sure everyone here will be a part of. Including me. Because I was sold something that clearly wasn't what it actually was.

The last thing Tesla wants are loads of lawsuits. Let's wait and see what happens in the next major update (9.0) before we jump to conclusions. If nothing serious happens in 9.0. Then I would start to become a bit worried at that point.

 

[AmpedRealtor]

We are in autonomy's infancy right now. In fact, it's more like a fetus that hasn't even been born. Of course the technology will change and features that are impossible today will be possible in a year, or three. This space will be constantly evolving like all of technology. Tesla will move ahead with hardware versions 3.0, 4.0 and keep going. They will add radars, maybe lidars, and whatever else they deem necessary over time. Those who bought on the promise in 2016 will get some version of that promise, but don't expect Tesla to update your vehicles to the latest and greatest. They won't. They simply can't afford to do so.

 

[alcibiades]

We are in autonomy's infancy right now. In fact, it's more like a fetus that hasn't even been born. Of course the technology will change and features that are impossible today will be possible in a year, or three. This space will be constantly evolving like all of technology. Tesla will move ahead with hardware versions 3.0, 4.0 and keep going. They will add radars, maybe lidars, and whatever else they deem necessary over time. Those who bought on the promise in 2016 will get some version of that promise, but don't expect Tesla to update your vehicles to the latest and greatest. They won't. They simply can't afford to do so.

The issue isn't with Tesla making upgrades as tech advances to implement better the cars' abilities. There's of course no issue with that.

The problem is that Tesla's Musk Madoff mouth wrote a check in October 2016 that its cars can't cash, so stockholders someday will. And it is going to be very expensive.

 

[BigD0g]

Yah, still waiting to see that S curve for AI for self driving. 3 months maybe 6 months definitely. I guess we need to find the Elon definition for definitely.