Pretty much certain that Tesla are developing in-house AP hardware so, whilst it might be a way off yet, do you think AP3.0 will be Tesla processing or are we likely to see more significant iterations of NVidia first?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Except so far that’s meaningless. Other than some redundant wiring and barely more processing power (nothing that runs neural nets as far as we can tell) there’s not much difference between the two. And there’s no functionality difference either.
Hey, there's neural net wipers! They're functionally a little bit closer to reaching parity with AP1....
I like the technical aspect of doing it they way they've been going, but the delivery time isn't great. If wipers took this long you have to wonder how long it will take to reach some semblance of self drive.
I like the technical aspect of doing it they way they've been going, but the delivery time isn't great. If wipers took this long you have to wonder how long it will take to reach some semblance of self drive.
I can't see any reason for implementing AP3. They created AP2 which Tesla means is enough for full autonomy. Then they upgraded it to AP2.5 that just added additional redundancy, computing power and a new radar (all can be retrofitted for free to AP2 cars that originally bought the FSD package if deemed necessary to achieve FSD).
Creating AP3 will just add another surface for extensive testing and validation. And tbh Tesla has enough with improving and validating AP1 and AP2. Adding AP3 just makes no sense before AP2 has the features promised.
I guess that when FSD on AP2 becomes released and it works well in most conditions, they will release AP3 with the shortcomings experienced with AP2/2.5.
AP3 will probably be more capable of driving under worse conditions, e.g. sensors cleaning themselves. Maybe add a sensor or two to deal with rare corner cases. And be fitted with a standard Vehicle-2-Vehicle communication protocol which should happen soon.
Creating AP3 will just add another surface for extensive testing and validation. And tbh Tesla has enough with improving and validating AP1 and AP2. Adding AP3 just makes no sense before AP2 has the features promised.
I guess that when FSD on AP2 becomes released and it works well in most conditions, they will release AP3 with the shortcomings experienced with AP2/2.5.
AP3 will probably be more capable of driving under worse conditions, e.g. sensors cleaning themselves. Maybe add a sensor or two to deal with rare corner cases. And be fitted with a standard Vehicle-2-Vehicle communication protocol which should happen soon.
Will there be future iterations of the AP sensors and processor? YES
AP is still early in the technology life cycle - so we should expect to see periodic hardware changes over time.
Like any new technology - AP will likely get better sensors and a faster processor - probably at about the same or slightly increased cost for the next few years.
And once EAP/FSD matures - then we'll see a shift towards less new functionality and more cost reduction.
This is what has happened to almost every technology (except perhaps smartphones - which seem to get more expensive each year).
Will we see a faster processor released for AP? Very likely - since Tesla doesn't yet have FSD working, they need to anticipate the possibility FSD will require more processing power than AP 2.x has today.
But, if Tesla is doing this right (and hopefully they are), the software they are developing for FSD will be able to run on a range of processors - from multiple sources - so that if they decide to shift to different hardware, the software will port over to the new processor without any major changes.
And, it's possible that if the AP2.0 processor isn't powerful enough for FSD - that Tesla will have an upgrade path to do a processor swap.
It's too early to tell how this is all going to work out.
But at least with AP 2.x Tesla cars, we have a chance to see EAP/FSD in our cars - something that someone buying an ICE over the last 15 months will not get...
AP is still early in the technology life cycle - so we should expect to see periodic hardware changes over time.
Like any new technology - AP will likely get better sensors and a faster processor - probably at about the same or slightly increased cost for the next few years.
And once EAP/FSD matures - then we'll see a shift towards less new functionality and more cost reduction.
This is what has happened to almost every technology (except perhaps smartphones - which seem to get more expensive each year).
Will we see a faster processor released for AP? Very likely - since Tesla doesn't yet have FSD working, they need to anticipate the possibility FSD will require more processing power than AP 2.x has today.
But, if Tesla is doing this right (and hopefully they are), the software they are developing for FSD will be able to run on a range of processors - from multiple sources - so that if they decide to shift to different hardware, the software will port over to the new processor without any major changes.
And, it's possible that if the AP2.0 processor isn't powerful enough for FSD - that Tesla will have an upgrade path to do a processor swap.
It's too early to tell how this is all going to work out.
But at least with AP 2.x Tesla cars, we have a chance to see EAP/FSD in our cars - something that someone buying an ICE over the last 15 months will not get...
Wow did not know that.
Can not believe Tesla would risk canabalizing even one s or x sale right now for a $6 cost camera.
Interesting choice. Even if there is a big revamp is around the corner.
Plus adding another dash in the lying column that s and x will always be tech leaders
lunitiks
Cool James & Black Teacher
What’s more interesting than who will supply the chip, is IMO what new sensors and/or tactile driver feedback will be added (if any)
Yes, I agree if Tesla is doing this right, multiple platforms and architectures should be fine in terms of code. But still failure to integrate test each combination well enough will cause issues, even if the theory is good and the programmer is confident. My worries is that Tesla is already accumulating a lot of different hardware since 2012. They should really think through their coming iterations carefully, they really don't need more test surface now before things they have sold start working.
It's not that long ago Tesla broke HomeLink on AP2.5 cars in a release. Those systems should be unrelated, and the mistake was probably really stupid (or unstructured). But the main cause for releasing that is failure to do proper real life integration testing on that combo. This won't be the last break we see.
I think they're trying to go with low cost AP computers until they get their own chip working. Then it seems likely they'd have to replace the computers in all FSD cars with their own board. The AP2 Nvidia board seemed pretty lightweight for FSD. They save money for now, and unify the FSD hardware.
I agree. They will probably retrofit their new chip on all cars that bought FSD.
It will be exciting to see how much they will increase the price for retrofitting those that did not buy FSD with the car, when FSD starts to work and becomes popular. With a high profit on retrofit, it will keep the cashflow up and pay for the free retrofits.
EAP cars probably won't be retrofitted. No need for lots of power to keep between 2 lines.
