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

Help me time my M3 purchase.

This site may earn commission on affiliate links.
I am trying to time my purchase as well.
I have a 2016 Nissan Leaf that is on lease. The lease expires in May, and I'm ready to turn this piece of $h1t in and upgrade to a Long Range Model 3!!

Estimate for my desired configuration says 2-4 weeks. I'm hoping to order in early April, so if it takes anywhere from 4-8 weeks, I will be able to return the Leaf and take the new Tesla in May - and still beat the tax credit reduction at the end of June. A win/win!!

Thinking of long range RWD blue with white interior, AP but no FSD yet. First time I've ever had a $50,000 car. Nervous, but excited like a kid at Christmas.…!!
 
It isn't needed vs it doesn't make any difference are two entirely different things.

I don't need a calculator to do math but it sure makes it quicker.

Exactly this.

To us computer nerds the HW 2.5 vs HW 3 is like the diff between an integrated graphics card vs the latest Nvidia. It doesn't matter if I'm only running some program with low demand, having powerful hardware matters (perhaps more than wheel size or paint color)
 
I understand that Elon Musk has said that HW3 is included with FSD, and I understand that everyone who has purchased FSD wants the HW3 upgrade.

What gives me pause and has stopped me from pulling the trigger on FSD is the simple fact that it does not say “includes free upgrade to HW3” on the order page for FSD. You sort of have to ask yourself, if that was truly Tesla’s intention, why would they omit that officially? And why wouldn’t they also mention this fantastic free upgrade on their website every chance they got? It would certainly have sold at least one more vehicle (to me), and I suspect to many others.

I also know that Elon Musk has a demonstrated history of saying things he believes/hopes to be true at the time and then walking them back later when they turn out to be perhaps not as true.

I certainly hope that everyone does get the upgrade they are expecting. My hunch is that even Tesla does not yet know if they can deliver FSD on HW2.5. I bet they are trying, however, since it will save them quite a bit of money on the upgrades.

I am going to wait to purchase until HW3 is confirmed in the vehicle, since even if Tesla is able to get it to work on 2.5 I bet it will run much better on 3.
 
I understand that Elon Musk has said that HW3 is included with FSD, and I understand that everyone who has purchased FSD wants the HW3 upgrade.

What gives me pause and has stopped me from pulling the trigger on FSD is the simple fact that it does not say “includes free upgrade to HW3” on the order page for FSD. You sort of have to ask yourself, if that was truly Tesla’s intention, why would they omit that officially?

Because most buyers have no idea WTF "hw3" is in the first place and they'd just create a ton of confused customers.

Instead they list the features you are buying. There's no need to worry about the HW behind it.

For the techie folks who DO know what it is, Elon has clearly, explicitly, repeatedly, consistently stated, clearly, that HW3 comes free to all FSD buyers.



And why wouldn’t they also mention this fantastic free upgrade on their website every chance they got? It would certainly have sold at least one more vehicle (to me), and I suspect to many others.

Because, again, the only reason it matters is to enable FSD features

HW3 doesn't magically add 20% horsepower to the car, or make it magically have a heated steering wheel it didn't have before.

Literally the purpose of the thing, the only thing it's needed for, is to enable the features already listed as the future FSD features

It'd be like a restaurant throwing up in big letters on the menu "STEAKS COME WITH SHARP STEAK KNIVES!"

At best folks would mostly ignore it, at worst it'd confused people and make customers go "Wait- are you saying your steaks are so tough they need special knives?"

Instead they just tell you about the awesome steak you're going to get to eat.



I also know that Elon Musk has a demonstrated history of saying things he believes/hopes to be true at the time and then walking them back later when they turn out to be perhaps not as true.

Can you cite any examples of his promising a physical deliverable on which one might condition a purchase from his company that turned out not to be true?

I don't mean "We think X will be ready by this date" and it's late.

I mean "If you buy this, you physically get X" and then you don't.


My hunch is that even Tesla does not yet know if they can deliver FSD on HW2.5.

Then it'd be really weird they spent a couple of years, and a ton of money and development, on creating HW3 in the first place, and are now actively installing it in new S/X cars (with 3s likely coming soon) huh?

In fact that'd make 0 sense at all.



I bet they are trying, however, since it will save them quite a bit of money on the upgrades.

It really wouldn't.

Once you're producing 500k boards anyway (roughly the annual car run rate they're aiming for by end of 2019) producing an extra maybe 20-30% is a pretty tiny cost.... ditto the 30 minutes labor at the SCs they've said the swap takes.



I am going to try to cancel the M3P order I made on March 17th (delivery date scheduled, but not matched to a VIN yet) tomorrow and then place my order again once HW3 is officially announced.


Prices go up April 2nd, so enjoy paying 3% more for literally no reason at all (see again your current car would get the HW3 swap free if you buy FSD, and doesn't need it if you aren't)



Exactly this.

To us computer nerds the HW 2.5 vs HW 3 is like the diff between an integrated graphics card vs the latest Nvidia. It doesn't matter if I'm only running some program with low demand, having powerful hardware matters (perhaps more than wheel size or paint color)


It really does not though for a low demand program.

Integrated graphics will run minesweeper just as well as the fastest Nvidia card will.

The difference only matters when you try and run something the lower spec can't already run full speed.

(this is somewhat similar to why "big brakes" don't actually stop a car any faster- they're not the limiting factor in the first place in any normal situation- the tires are).

But that's not even happening here- HW3 isn't going to be running the same code as HW2.x long term anyway.

One of the things explicitly mentioned when HW3 was discussed on the investor call was HW3 will enable much larger neural nets for the advanced coming-later-this-year FSD features... it won't do anything for the autopilot folks going forward, because other than the stuff rolling out now that 2.5 can already do (advanced summon primarily, to make EAP feature complete) they're not getting more features going forward.

I'd expect primary development going forward will be on FSD/HW3 neural nets that simply don't run at all on older HW, and 2.x code will basically be bug fixes and minor tweaks (somewhat similar to what happened to HW1 cars when 2 came out).
 
May all this happen exactly as you predict Knightshade.

I still have strong doubts. I stand by my original conviction to wait to buy when I know exactly what it is I am buying—that is, when it’s in black and white on a sales contract.
 
It really does not though for a low demand program.

Integrated graphics will run minesweeper just as well as the fastest Nvidia card will.

The difference only matters when you try and run something the lower spec can't already run full speed.

(this is somewhat similar to why "big brakes" don't actually stop a car any faster- they're not the limiting factor in the first place in any normal situation- the tires are).

I'm not asking this to be argumentative...I'm just asking to understand better...

Are you saying the safety features and autopilot are already at the same technological maturity level as minesweeper?? In other words, they are the best they will ever be and can never ever get better?

Or are you saying the limiting factor in the safety features and autopilot today is something other than the computer running it? If so, what do you believe the limiting factor is?? Programming?? Environmental observation devices??
 
I understand that Elon Musk has said that HW3 is included with FSD, and I understand that everyone who has purchased FSD wants the HW3 upgrade.

What gives me pause and has stopped me from pulling the trigger on FSD is the simple fact that it does not say “includes free upgrade to HW3” on the order page for FSD. You sort of have to ask yourself, if that was truly Tesla’s intention, why would they omit that officially? And why wouldn’t they also mention this fantastic free upgrade on their website every chance they got? It would certainly have sold at least one more vehicle (to me), and I suspect to many others.

I also know that Elon Musk has a demonstrated history of saying things he believes/hopes to be true at the time and then walking them back later when they turn out to be perhaps not as true.

I certainly hope that everyone does get the upgrade they are expecting. My hunch is that even Tesla does not yet know if they can deliver FSD on HW2.5. I bet they are trying, however, since it will save them quite a bit of money on the upgrades.

I am going to wait to purchase until HW3 is confirmed in the vehicle, since even if Tesla is able to get it to work on 2.5 I bet it will run much better on 3.
This is the sensible approach. You cant time stocks, and you also cant time tesla. You want hw3, wait until its out then buy. You want the 3750$ rebate, put your order in 1 may the latest, to give tesla 8 weeks to fullfill the order. You dont want buyer remorse, because you pay to much for fsd preordered, wait until fsd is out. Yes, you might pay more, in the end, but I think it will sting more, when you think , you have made a good deal, to only see it turn in to something that you cost money, if only you had waited.
 
I'm not asking this to be argumentative...I'm just asking to understand better...

Are you saying the safety features and autopilot are already at the same technological maturity level as minesweeper?? In other words, they are the best they will ever be and can never ever get better?

.... that analogy doesn't really work...

If they "improved" minesweeper to have brighter colors, it'd still run exactly as well on integrated graphics as it does on the newest Nvidia high end GPU.

The program simply doesn't demand very much of hardware... and has no code dependent on specific HW.... so faster hardware does not help.

Running the 2.x NN on HW3 isn't going to show much improvement either- since by definition it was written with the HW limits of 2/2.5 in mind.

The big "jump" in HW3 is running MUCH LARGER NNs that 2.x can't run at all.... this was explained in some detail on the investor call.

Which is why future development will likely be on the HW3 version going forward (exactly when the bulk of that development switch happens is unclear, but there's no reason to doubt it'd happen- not like they went back and did a ton of HW1 code updates once HW2 existed)



Or are you saying the limiting factor in the safety features and autopilot today is something other than the computer running it? If so, what do you believe the limiting factor is?? Programming?? Environmental observation devices??


"safety features" is pretty much AEB and blindspot warning... HW2.5 should have plenty of horsepower to do JUST those things. You get those even if you didn't pay for any driver assist features anyway, and I wouldn't expect them to be significantly different on HW3 cars either, since the sensor suite hasn't changed.

AP today is just TACC and autosteer- which generally work very well... and given they've moved all the "other" EAP features out of the new AP, again there should be plenty of spare cycles if they find tweaks to make to improve them at all- though again I expect we won't see a ton more development in that regard.



I'm not really sure, given the defined functions of those features (TACC, autosteer, AEB, and blndspot) what limits you really think they've got, computer-wise?



Certainly some of the more advanced features not part of either (but with some overlap for legacy EAP-only owners) are bumping into computational/HW limits.... that's part of why HW3 exists in the first place.

My expectation as mentioned is EAP-only on 2.5 won't be significantly improved going forward (you'l get advanced summon, and probably a few other tweaks/improvements first half of this year, but after that coding efforts will largely be in the HW3 NNs)

FSD cars, which will include the HW3 upgrade for 2.x vehicles, will see increasingly better behavior running this much larger and more advanced NNs that 2.x simply can't do- which will enable things like NoA in much more complex situations like city streets.

(I don't think you get anywhere near level 5 this way- but I could definitely see an L2 system that works most everywhere, and being able to give you L3/L4 in some limited domains once regulation isn't a mish-mash-mess of 50 different states laws.)






May all this happen exactly as you predict Knightshade.

I still have strong doubts. I stand by my original conviction to wait to buy when I know exactly what it is I am buying—that is, when it’s in black and white on a sales contract.

You realize the AP HW version isn't in the sales contract, no matter when you buy the car, right?
 
"safety features" is pretty much AEB and blindspot warning... HW2.5 should have plenty of horsepower to do JUST those things. You get those even if you didn't pay for any driver assist features anyway, and I wouldn't expect them to be significantly different on HW3 cars either, since the sensor suite hasn't changed.

AP today is just TACC and autosteer- which generally work very well... and given they've moved all the "other" EAP features out of the new AP, again there should be plenty of spare cycles if they find tweaks to make to improve them at all- though again I expect we won't see a ton more development in that regard.



I'm not really sure, given the defined functions of those features (TACC, autosteer, AEB, and blndspot) what limits you really think they've got, computer-wise?

I don't even have my M3 yet, so I can't even comment first hand beyond a 5 mile drive in my friend's M3.

I guess I find it hard to believe that auto-emergency braking is already perfected. That blindspot warning is perfected. That their is no way they can ever benefit from a CPU upgrade.

I read on this forum that autosteer struggles on some roads.

Maybe a future iteration of TACC does more predictive analytics of the specific vehicle in front of you based on road conditions and whatever it has observed to date? Maybe it does that for all of the vehicles surrounding you. Maybe it understands exactly what type of car it is and what those vehicle characteristics could do? Clearly this is all programming, but maybe more possible on a more advanced CPU.

In your post you even say "I wouldn't expect them to be significantly different".... which kind of seems like a bit of backtracking from you.

I just think it's hard to say with certainty that there will be 0 benefit. That's my only point.
 
I don't even have my M3 yet, so I can't even comment first hand beyond a 5 mile drive in my friend's M3.

I guess I find it hard to believe that auto-emergency braking is already perfected. That blindspot warning is perfected. That their is no way they can ever benefit from a CPU upgrade.

What functionality do you think is CPU limited in either feature, specifically?

Both are really far more sensor limited than anything else as far as improvements (most other companies use side radar for blindspot detection for example- Tesla does not, and a faster CPU doesn't fix that)


I read on this forum that autosteer struggles on some roads.

Only time I've seen it struggle at all is either:

Very high curve where it can't see where it's going. Again not a CPU problem.... and when there's no visible road lines and has no car to follow (and that only generally on rural, non-divided, roads where AP is explicitly not intended for use anyway)

But if you can cite specific situations "more CPU" would help, what are they?



Maybe a future iteration of TACC does more predictive analytics of the specific vehicle in front of you based on road conditions and whatever it has observed to date? Maybe it does that for all of the vehicles surrounding you. Maybe it understands exactly what type of car it is and what those vehicle characteristics could do? Clearly this is all programming, but maybe more possible on a more advanced CPU.

But virtually none of those things are "TACC" functions... (and they're pretty vaguely defined.... what, functionally, does "understanding what type of car" do that it doesn't do now, in practical terms? the system already IDs between cars, trucks, and bikes for example)

TACC uses the front camera and radar to see what the car in front of you is doing (and the one in front of it with scatter from under the front car)... and uses that info to maintain requested set speed in relation to forward traffic and follow distance.... I've seen no suggestion in anything it does that it's CPU limited at all. Have you?



I expect more CPU to help a lot with things the car doesn't do now- because it can't do it with HW2.x...like handle complex urban/city driving with a LOT more input from the cameras on pedestrians, cyclists, animals, and both oncoming and cross traffic.

But just "stay in your lane on divided highways and adjust speed based on radar/front camera", which is the current AP offering? That worked pretty well back on HW1.
 
If you define the Autosteer/TACC functionality as narrow as "stay in your lane on divided highways and adjust speed based on radar/front camera"... then you are probably right.

You say none of those things I mentioned are TACC functions. Why not? "Traffic aware cruise control". All of the things I mentioned are traffic awareness related. Expand your mind bro...

haha..but in all seriousness.. I was just asking and trying to be educated. Thanks for your commentary!