-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
So when do I schedule the service appointment to beat everyone to the front of the line here? :D
- Thread starter MXWing
- Start date
No worries. If he says 6 months it will be at least a year.
AP3 hardware isn't an abstract idea like "Coast to coast FSD" or "Tesla Mobility Network".
From my understanding, they were already testing this during the time of the Q2 conference call. So we were looking at about a year from that time.
I speculate that installing this chip (when it does arrive) does not mean instant gratification. In other words, I doubt you'll see any difference as the end user until there is a software roll out that will take advantage of the chip's capabilities - whether that is FSD or otherwise.
Maybe I am wrong. For all I know the 360 view of virtual cars will be smooth as silk and no longer jerky the moment the car re-boots after chip install.
Maybe I am wrong. For all I know the 360 view of virtual cars will be smooth as silk and no longer jerky the moment the car re-boots after chip install.
It's going to depend on how the architecture of the chip is setup. Cars with AP3 hardware in production comes preloaded with correct software.
Drop in replacements imply software change at same time.
Software isn't the same if the chip architecture is different which sounds like the case here. Example is Tegra for MCU1 and x86 for MCU2.
Uncle Paul
Well-Known Member
Its going to cost you much more than that because you will have alimony and potentially child support as well.
MXFan
Member
wamochi
Member
i wonder if they will automatically notify owners that have already paid for FSD ... (or if it's up to us to ask)
Since no one has FSD features, it would make sense to start the consumer rollout of FSD features on AP3.
Since FSD features won't work on AP2 and AP2.5 you will know to go for the upgrade.
They can email everyone and also put it in the car notes when people upgrade their software.
I have a HW engineering background. I've done tapeouts & part bringup, board design, etc. This is much more involved than it seems. As an org, this is their first go-round w/ doing this and the first time (even if you hire all the best people who have done it before elsewhere) it always rough. A novel computational architecture makes this even worse as there aren't good compilers/debuggers/etc for it.
Anyone know what fab/process they're using or when they did the tapeout? A quick google doesn't find anything w/ a tapeout date. That's not a public event, but I guess I'd be a bit surprised if a co the size of Tesla could keep a think like that quiet. For a project of this type/size and it being their first go around I'd be truly surprised for them to hit field deployment less than a year after first tapeout.
Anyone know what fab/process they're using or when they did the tapeout? A quick google doesn't find anything w/ a tapeout date. That's not a public event, but I guess I'd be a bit surprised if a co the size of Tesla could keep a think like that quiet. For a project of this type/size and it being their first go around I'd be truly surprised for them to hit field deployment less than a year after first tapeout.
If I were a betting man I would put my money on the chip roll out coinciding with version 10 in 2 years, and in addition this is when we will see FSD features being split from non FSD features for those who paid. this will be my internal assumption and if it happens earlier I will be pleasantly surprised.
Super_Popular
Well-Known Member
That's awesome that t.he chip will be installed, still need to what for Federal approval and that is going to take years.
What the government is in charge of is approving whether we can step out of the car or not. It may or may not preceed the time when Tesla has fully functioning FSD, but the chip will be very useful well before either of those milestones.
Things like whether or not it requires input to reach its destination and how much input will be required can be improved at any time without any restriction from legislation as long as the "who is responsible" issue remains "the driver". We may also see self-parking authorized garages before we see approval for use on public roads.
They may even allow the driver to take responsibility for disabling nags before they allow truly human-free driving - again, it would be up to the driver to take responsibility for anything the car does even if they've disabled the nags and the driver would be backing up that decision with their own life as well.
Any of these may be improved or enabled by that chip well before any laws are changed.
dhanson865
Well-Known Member
MXWing said:
Disagree that it requires software change, Example is 8086+8087 same code faster execution.
They could optimize for the new chip or have different code downloaded to cars with the new chip eventually, but they don't have to do so on day 1 if the new chip can emulate the old chip or if the code type is general purpose enough that it runs hardware agnostic.
Hardware Abstraction Layers are a thing. It's why a new graphics card can come out years later and speed up and improve graphics on an old game.
If they already moved to x86 why move away and not keep x86 on board? MCU2 + NNchipwhatever. Sure you replace the whole board, but the new board has to have a CPU of some sort, I'm saying the NN chip is comparable to a dedicated GPU in a laptop, external to the CPU but still on the motherboard either directly or by way of socket or even a riser.
Last edited:
Similar threads
