HW2.5 capabilities

[Sterph]

I always assumed that Tesla was using NNs outside the car to generate/develop the updates, which are then sent out and used more-or-less like any regular none-NN update. This would explain the erratic NN-style behavior we see in current AP2.0 cars, yet why the car itself isn't using NN processing to drive on Autopilot.

Is there any particular reason this is definitely not the case, or have we already disproved this?

 

[djcasava]

For what it's worth, I ordered a MS75D on Aug 20 in NJ with a confirmation on Aug 23. It originally had a delivery date of late September but after the confirmation date, the delivery date changed to October/early November. I checked the options code and it shows that the Tesla will have AP 2.0 and not AP 2.5.

My OA told me that if I wanted the car by September in NJ, I had to order the car by Aug 20th.

1. Did you think the OA know that all vehicles ordered prior to Aug 20th WON'T get AP 2.5?
2. Or simply pushing to get a vehicle delivered by end of 3rd quarter to report financial numbers?

It would really stink if I missed out on AP 2.5 by one day.

 

[speedy]

Check your options again. When mine was confirmed Aug 10th, it was APH2. I checked it again recently, and it is now APH3. Yours should be the same. As to whether APH3 is AP2.5 or something newer, I don't know...

 

[ab26]

Check your options again. When mine was confirmed Aug 10th, it was APH2. I checked it again recently, and it is now APH3. Yours should be the same. As to whether APH3 is AP2.5 or something newer, I don't know...

Are you checking your options using the view design URL? A friend of mine confirmed his model x on 17th August and the code still shows APH2.

 

[James*]

For what it's worth, I ordered a MS75D on Aug 20 in NJ with a confirmation on Aug 23. It originally had a delivery date of late September but after the confirmation date, the delivery date changed to October/early November. I checked the options code and it shows that the Tesla will have AP 2.0 and not AP 2.5.

My OA told me that if I wanted the car by September in NJ, I had to order the car by Aug 20th.

1. Did you think the OA know that all vehicles ordered prior to Aug 20th WON'T get AP 2.5?
2. Or simply pushing to get a vehicle delivered by end of 3rd quarter to report financial numbers?

It would really stink if I missed out on AP 2.5 by one day.

Mine used to say APH2 and when it went into production it updated to APH3.

 

[ab26]

Check your options again. When mine was confirmed Aug 10th, it was APH2. I checked it again recently, and it is now APH3. Yours should be the same. As to whether APH3 is AP2.5 or something newer, I don't know...

Can you confirm if you car went into production already? If yes, it may indicate that the code gets updated only after it goes into production. (similar to what James is saying).

 

[speedy]

Can you confirm if you car went into production already? If yes, it may indicate that the code gets updated only after it goes into production. (similar to what James is saying).

No, my car is due to enter production in early September (soon!). I know for a fact that the original options (in the 'View Spec" URL) changed as I saved those options when I first placed the order. Then after reading about the AP2.5 changes, I checked the options again on the 26th, and sure enough, it says APH3 now.

This is on a Model S, not sure if they made the same changes on the Model X

 

[Mjølner]

Mine changed 3 days after entering production.

 

[Bebop]

Check your options again. When mine was confirmed Aug 10th, it was APH2. I checked it again recently, and it is now APH3. Yours should be the same. As to whether APH3 is AP2.5 or something newer, I don't know...

Mine was built this past Thursday and it shows APH2. It was a new inventory car straight from the production line.

 

[Garshom]

I just took delivery of my Tesla (which was a dealer demo built at the end of May 2017) . Care was located in Atlanta and I live in South Florida so I flew up to Atlanta and took delivery at the Marietta store. The drive home could not have been better. We used Autopilot Autosteer the entire way including through horrific rush hour traffic in Atlanta etc. There was even one point where AutoSteer punished me for not having my hands on the steering wheel after being reminded so it shut itself off for that leg of the drive( We could have re engaged it after pulling over and puttiing it in park but didn't want to waste the time). The amount of driver's fatigue it saves was truly amazing and I could not imagine a road trip without it.

With regard to HW 2.5/ 3.0 the only thing I received in my "MyTesla" page after taking delivery was a $4,000.00 option to make the car autonomous. I could pay for the option now and see what it involves but I am willing to bet that for that price it includes whatever hardware and firmware upgrades might be necessary to make the car autonomous. I was explained that the current software does not utilize any of the side cameras or the B Pillar carmeras etc. for AutoPilot even though, its my understanding , that they are used to gather data. So even though my car has the additional hardware , It is not functional . I am willing to bet that the $4,000 must include any upgraded hardware that the government is requiring to make the car truly autonomous. So as long as you have the hardware 2.0 in terms of cameras , then any redundant wiriring or upgraded radar must be included . It makes no sense for them to offer this to me if my car was incapable of acheiving this.

 

[LastGas]

...

It would really stink if I missed out on AP 2.5 by one day.

I feel for ya. I missed rear-seat cup holders by 3 days. (And AP2 by 6 weeks)

 

[chillaban]

I feel for ya. I missed rear-seat cup holders by 3 days. (And AP2 by 6 weeks)

I missed cupholders by like a month but it sure sounds like you more of dodged a bullet, since we're almost a year into HW2 and in my opinion AP2 is still trailing AP1 in capabilities.

I bought AP1 and a few months back I've traded for an AP2 car (for range reasons, not AP reasons), and if it makes you feel any better, I don't think you're missing much from an AP standpoint, and probably won't be for at least several more months.

 

[ElectricTundra]

I'm coming up on 3 yrs with my P85D HW1. Many new bits have come out since that would be nice to have but if I traded today then more will come out within days, weeks, or months. I've gotten a lot of enjoyment out of driving it and a lot of benefit from AP1. Glad I didn't wait to buy it and quite content to enjoy it for another year or two.

If I didn't already have an S then I'd not wait to buy one now. There is no time that there won't be newer stuff in the future and missing out on the benefits of driving it today and for the next few years wouldn't be worth the wait for me.

That said, a new S for me will probably be when FSD is actually available. A new X for my wife will be whenever she decides its time and that will likely have zero to do with features or technology and more to do with having driven one car for 5 or 6 years so it's time.

 

[jimmy_d]

Ok, but until then we must speculate.

Anyone: Would liquid cooling make any sense without it being coupled to the cars existing cooling system? (Like e.g. to the battery coolant loop?) Could ECU2.5 possibly include a discrete liquid cooling system or would that be meaningless wrt effectivness?

(Trying to find traces of liquid cooling in the service documents and asses the likelyhood of ECU2.5 retrofit...)

Yes, there are lots of reasons you might want to do that. Having stand alone liquid cooling in an isolated subsystem lets you efficiently move significant amounts of heat away from the place it's generated (on a single IC perhaps) to a location that allows for easy dissipation into some preferred environment. Sometimes that's a couple of inches just to escape an EMI enclosure or to get away from other temperature sensitive parts. I can also think of good reasons to not want to connect to the vehicle's liquid cooling transfer medium directly: cost, complexity, reliability, modularity can all suffer. Of course, dumping heat directly into the vehicle's head transfer fluid is thermodynamically efficient, but for the amount of heat generated by this system it's probably not a big deal - it's on the order of the heat generated by a single passenger and a small fraction of the heat load the car experiences from just being outside in daylight. Dumping heat directly into the cabin environment or into the ventilation system after using an isolated liquid cooling system to move the heat outside of the electronics enclosure seems like a reasonable choice for a compact, high reliability system with the order of 100W/liter thermal density.

 

[number12]

For what it's worth, I ordered a MS75D on Aug 20 in NJ with a confirmation on Aug 23. It originally had a delivery date of late September but after the confirmation date, the delivery date changed to October/early November. I checked the options code and it shows that the Tesla will have AP 2.0 and not AP 2.5.

My OA told me that if I wanted the car by September in NJ, I had to order the car by Aug 20th.

1. Did you think the OA know that all vehicles ordered prior to Aug 20th WON'T get AP 2.5?
2. Or simply pushing to get a vehicle delivered by end of 3rd quarter to report financial numbers?

It would really stink if I missed out on AP 2.5 by one day.

That's the exact story that happened to me with ap1 AP2. Except they just pushed my delivery earlier because they could. TSLA lives quarter to quarter.

 

[AnxietyRanger]

Yes, there are lots of reasons you might want to do that. Having stand alone liquid cooling in an isolated subsystem lets you efficiently move significant amounts of heat away from the place it's generated (on a single IC perhaps) to a location that allows for easy dissipation into some preferred environment. Sometimes that's a couple of inches just to escape an EMI enclosure or to get away from other temperature sensitive parts. I can also think of good reasons to not want to connect to the vehicle's liquid cooling transfer medium directly: cost, complexity, reliability, modularity can all suffer. Of course, dumping heat directly into the vehicle's head transfer fluid is thermodynamically efficient, but for the amount of heat generated by this system it's probably not a big deal - it's on the order of the heat generated by a single passenger and a small fraction of the heat load the car experiences from just being outside in daylight. Dumping heat directly into the cabin environment or into the ventilation system after using an isolated liquid cooling system to move the heat outside of the electronics enclosure seems like a reasonable choice for a compact, high reliability system with the order of 100W/liter thermal density.

Great rationale, thanks!

 

[Bebop]

It would really stink if I missed out on AP 2.5 by one day.

It doesn't matter. Even IF 2.5 was needed for FSD, Elon said it would be a absolutely 100% free upgrade for all current AP2.0 cars. So no need to worry about it. Just try to get your car as soon as possible so you can finally drive it!

 

[verygreen]

I wish we can separate out all the discussion about "will my car get HW2.5" and "should I wait for HW2.5" into some separate thread. While it does not bother me now, whoever might be reading this say a few months from now would certainly not appreciate all the sidelined discussions.

 

[jimmy_d]

Just discovered this thread, lots of interesting speculation going on but I can't help feeling that there's a lot of tea-leaf reading too. It reminds me of being on the other side when I was developing a product and reading comments on public message boards speculating on the development. Virtually 100% of the stuff on the boards was just so wrong that it was depressing to read. Reading this thread was just reminding of that experience.

AT that company the technical stuff we did was complicated, the company was complicated, the market was complicated. Our vendors were constantly changing stuff around on us and telling us just as the parts showed up on the loading dock. Marketing, sales, management, engineering, operations, and the factory all had somewhat different takes on where we were and where we were going. It was hard to figure out exactly what customers were asking for, whether they understood the implications of their requests, and whether they were going to ask for something incompatible next month. On top of this we were in a fast moving market where the situation on the ground was constantly changing so we did a lot of things for temporary or speculative reasons, or to protect ourselves from some subtle but potentially important risk. If you asked any two people in the company for an opinion on a technical detail you'd get two different answers. It wasn't chaos, it was just complicated. It wasn't a dysfunctional process, it was just so elaborate that you saw it differently depending on where you stood.

Despite the internal reality of the company you'd see these really simplistic opinions take hold on the public message boards which I'm pretty sure every single insider would have laughed at. Outsiders would be completely convinced of things that any insider would have dismissed as absurd.

Anyway, I don't mean to criticize the discussion here. I enjoy it. I'm not a Tesla insider so I can't do any better than anyone else at understanding what's going on with AP2. But I love my AP1 car, I just bought an AP2 car, and I'm really excited about where all this stuff is going so I really appreciate having a place to kick it around with other folks.

But for people who are getting worried about whether they are going to be abandoned by Tesla on the basis of a wiring chart or some log file entries, or maybe by some conspiracy minded poster: I humbly suggest that you take everything derived from what you read here with a giant grain of salt. We're probably somewhere between 95% and 99% wrong on the stuff we speculate about.

 

[jimmy_d]

Its clear that you do not know how a business operates. Better tech is never cheaper. Just let the disagree fly, it won't change reality.

I think your observation has merit in slow moving domains, but not in domains where the tech is rapidly advancing. I spent decades selecting and procuring parts for high volume high tech products that my company manufactured. Every single year the specs on the parts got better and the prices went down. And every time our order volumes increased we would renegotiate with a vendor for better terms and pricing. From what I can see cameras, automotive radar, and all the electronics that go into the compute infrastructure for autopilot are prime candidates for the kind of steep learning curve that leads to better tech at lower prices with meaningful amounts of improvement occurring annually or faster.