There will be NO HW4 upgrade for HW3 owners

[Midpack]

Given their track record, I'm not sure why anyone would buy FSD upfront anymore if they can just subscribe at $99/199 per month.

And while I'll admit I thought they'd be a lot closer by now, I don't remember when Tesla guaranteed a date for FSD to be fully capable, so I'm not sure I understand those who've decided times up. Good luck.

 

[Ramphex]

Given their track record, I'm not sure why anyone would buy FSD upfront anymore if they can just subscribe at $99/199 per month.

And while I'll admit I thought they'd be a lot closer by now, I don't remember when Tesla guaranteed a date for FSD to be fully capable, so I'm not sure I understand those who've decided times up. Good luck.

It’s not even worth the subscription price, unless you’re new and you just want to play with it to get it out of your system. But that’s just my opinion.

 

[vinnie97]

I thought I was going to get FSD for the cyber truck, but I'm increasingly not interested.

 

[jeremymc7]

I guess wait and see if it has 11 / 12 cameras and see how much more interesting it is to hen we get there.

 

[Midpack]

I thought I was going to get FSD for the cyber truck, but I'm increasingly not interested.

I assume you mean get an FSD subscription for $199/month. I can’t imagine any reason/advantage in paying $15K upfront.

 

[Kognos]

Apparently there will be no free upgrade Elon Musk kills hope of Tesla retrofitting new Autopilot/Self-Driving hardware

Which seems very unfair given I paid 10k and other folks here paid even more. Honestly that makes me want to pursue some sort of recourse.

This is actually one of the reasons I advocated for FSD license to be tied to the user and not the car.

Unless you believe Elon Time - and FSD is "done soonish" - there is no way the hardware will meet expectation by the time the product is completed. There are a ton of little things that haven't been sorted out yet to truly make this a level-5 car.

And as technology goes, not everything is just plug-and-play. This hard-barrier was bound to happen, especially with near-decade development cycles.

So if the license were tied to the user and not the car, you can take your license and move it to a new Tesla which has HW4/5/x and problem is solved with significantly less friction. Considering each car automatically comes with all the hardware necessary, this should be a completely seamless experience.

 

[jeremymc7]

So if the license were tied to the user and not the car, you can take your license and move it to a new Tesla which has HW4/5/x and problem is solved with significantly less friction. Considering each car automatically comes with all the hardware necessary, this should be a completely seamless experience.

Totally agree and would actually cause me to not only buy more Teslas but to buy them more often. Hey I like to upgrade unnecessary. But given the current pricing I’m certainly not likely to get another Tesla anytime soon.

But the other factor is the the digital purchases running rampant these days. Tesla wants / expects you to just do a $15K purchase; or digital subscription. Either way it’s just another revenue stream and another way to nickel and dime the customer for everything.

Extra charges for electric seats, ability to use Apple CarPlay or Google Audio, faster performance, better audio on the stereo.

Either these digital type of subscriptions need be much cheaper. $1/5/mth depending. Or maybe $20-50 for FSD, but not at its current beta performance.

Otherwise keep it old school and let us purchase each, and for old screen prices. Not hundreds to thousands to turn each on.
L

 

[vinnie97]

I assume you mean get an FSD subscription for $199/month. I can’t imagine any reason/advantage in paying $15K upfront.

This sounded attractive at one time, but not with a headaches of dealing with it on a daily basis. I'm not sure I can justify $200 a month either. With a single activation fee, I at least wouldn't have a monthly reminder of how far it falls short.

 

[eevee-fan]

Tesla Model Y Hardware 4 Teardown Reveals Cheaper Ram, Missing GPU and Camera Connectors

Get an expert's insights on the new Tesla Model Y HW4, as we dissect the intricate details shared by renowned 'Tesla Hacker,' Green. From physical characteristi

www.notateslaapp.com

Different board size, looks like Tesla is trying to use the 'same' board for all models (Not just S/X).

 

[borland]

Tesla Model Y Hardware 4 Teardown Reveals Cheaper Ram, Missing GPU and Camera Connectors

Get an expert's insights on the new Tesla Model Y HW4, as we dissect the intricate details shared by renowned 'Tesla Hacker,' Green. From physical characteristi

www.notateslaapp.com

Different board size, looks like Tesla is trying to use the 'same' board for all models (Not just S/X).

There are still plenty of early Model 3's with FSD option, that were never updated with HW3 AutoPilot computers.

Due to high cost of HW3 semiconductor foundry/fab work, when/if Tesla runs out of HW3 update board inventories, it will be much less expensive for Tesla to make compatible update boards with HW4 component parts.

 

[dgatwood]

Due to high cost of HW3 semiconductor foundry/fab work, when/if Tesla runs out of HW3 update board inventories, it will be much less expensive for Tesla to make compatible update boards with HW4 component parts.

It's not just a case of changing the board layout to accommodate connectors here. If I understand correctly, HW4 is a single, unified computer that replaces both the FSD computer and the MCU computer. So they would have two choices: Either design a new board from scratch that removes the MCU bits (expensive) or find a way to mount the new, larger board where the MCU is and make modifications to the ventilation and camera wire routing (probably with an extension to bring them up from the FSD computer bay to the MCU bay).

The latter would be a better option, because it will fix the upgrade hassle once and for all, and if they end up being forced to do this upgrade, I'm assuming that they would make the cars compatible with the boards rather than the other way around. Otherwise, they have to keep doing that over and over again, generation after generation, until it actually is good enough.

 

[gearchruncher]

Either design a new board from scratch that removes the MCU bits (expensive)

There are over 2 million cars with HW3. The design cost is insignificant if it saves even a few dollars in each installation.

In the end, labor time for installation is the biggest thing you need to minimize if you're doing 2M of something.

 

[stopcrazypp]

It's not just a case of changing the board layout to accommodate connectors here. If I understand correctly, HW4 is a single, unified computer that replaces both the FSD computer and the MCU computer. So they would have two choices: Either design a new board from scratch that removes the MCU bits (expensive) or find a way to mount the new, larger board where the MCU is and make modifications to the ventilation and camera wire routing (probably with an extension to bring them up from the FSD computer bay to the MCU bay).

The latter would be a better option, because it will fix the upgrade hassle once and for all, and if they end up being forced to do this upgrade, I'm assuming that they would make the cars compatible with the boards rather than the other way around. Otherwise, they have to keep doing that over and over again, generation after generation, until it actually is good enough.

My understanding from green's post is the hard coolant connectors are in a different place and those will present the biggest challenge to retrofits (may require replacing coolant lines). The board integration actually wouldn't really matter as long as the input/outputs are compatible, given all the signal wires are flexible.

 

[dgatwood]

There are over 2 million cars with HW3. The design cost is insignificant if it saves even a few dollars in each installation.

In the end, labor time for installation is the biggest thing you need to minimize if you're doing 2M of something.

It depends on how many more times they have to replace it before it actually works. If doing a compatibility rework adds an extra $100 in (actual) labor costs, but you do that only once, times 2 million cars, that's a $200 million cost. If they can't pull off FSD according to spec until HW8, then they'll end up replacing them five more times. Divide that $200 million by five, and your total design and tooling and SKU management per upgrade would have to come in under $40 million. And four-ish hours of labor is a rather cynically high estimate for any rework. An hour of labor would be more like $10 million, which seems quite low for something that has to go through such extensive testing and validation before it ships.

 

[enemji]

How about replacing the new fender cameras on the HW3?

 

[gearchruncher]

An hour of labor would be more like $10 million, which seems quite low for something that has to go through such extensive testing and validation before it ships.

$10M divided by 2M is $8. You think Tesla only pays $8 an hour for techs? Their incremental and opportunity cost for an hour of labor is well over $50, probably closer to $100.

$10M is silly high to re-lay out a board that already has a schematic and then make a new housing and test it. They aren't designing something new, they are re-packaging it.

Tesla's never going to think beyond HW4 anyway. They thought HW2 was going to be FSD and that was now 7 years ago. HW 5 in their mind is 5+ years away, and thus you can assume a bunch of cars will be worn out by then. Putting extra labor in a car now in order to maybe save labor in the future is not the way a company like Tesla thinks at all, nor how they drive quarterly results.

 

[gearchruncher]

How about replacing the new fender cameras on the HW3?

Why? Do we have any evidence that this is limiting performance of the algorithm right now?

 

[enemji]

Why? Do we have any evidence that this is limiting performance of the algorithm right now?

Yes. It is very limited in making a decision when it comes to T intersections. It keeps on creeping and creeping and creeping and creeping….

 

[gearchruncher]

Yes. It is very limited in making a decision when it comes to T intersections. It keeps on creeping and creeping and creeping and creeping….

And you know snapping in the HW4 cameras will fix this?

 

[dgatwood]

$10M divided by 2M is $8. You think Tesla only pays $8 an hour for techs? Their incremental and opportunity cost for an hour of labor is well over $50, probably closer to $100.

You forgot to multiply by 5. That $10M is still after amortizing the labor cost across five updates, just like the $40 million was. The top of their pay scale for techs is somewhere around $40 per hour, but the average is considerably less than that. And you can't consider opportunity cost, because they can spread the work across as long a time period as they want to.

$10M is silly high to re-lay out a board that already has a schematic and then make a new housing and test it. They aren't designing something new, they are re-packaging it.

And redesign the cooling hardware. And separate out the MCU hardware from the FSD hardware. And validate that the new FSD hardware works correctly when talking to MCU1 and MCU2 and MCU 2.5. And possibly deal with different numbers of Ethernet connectors, etc. that may be present on the old hardware, plus things that don't even have ports anymore (RADAR), plus who knows what else.

It's not nearly as simple as you're making it out to be. Also, by doing the retooling, they'll have the opportunity to *sell* unified MCU upgrades to all those 2 million people once they pass the point where the FSD hardware is good enough. So it isn't entirely without potential upside.