Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
Considering that they could design a headlight unit with a side-facing camera, the only answer must be cost.
Perhaps the additional images are a problem for system throughput.. Tesla could always add a 2nd fender camera or even better add the cameras high up.
One thing I've noticed when Omar (Whole Mars Catalog) publishes his videos is he never has a badly obstructed view intersection.
 
Is that radar as an option like "we may or may not ship this thing with radar" or radar as an option like "customers may or may not select to get radar"

The first sure same as S/X Tesla is still making up their mind.... the second would be a massive change in how the company operates regarding driving HW.


BTW random question- remember a while back Tesla china deleted a backup steering motor or something, and dismissed concerns as it wasn't needed until autonomy?

Did highland add that back, or are they still admitting in HW autonomy isn't gonna be a thing right now?
 
  • Like
Reactions: navguy12
Is that radar as an option like "we may or may not ship this thing with radar" or radar as an option like "customers may or may not select to get radar"

The first sure same as S/X Tesla is still making up their mind.... the second would be a massive change in how the company operates regarding driving HW.


BTW random question- remember a while back Tesla china deleted a backup steering motor or something, and dismissed concerns as it wasn't needed until autonomy?

Did highland add that back, or are they still admitting in HW autonomy isn't gonna be a thing right now?
Sounds like option as in, Tesla might install it, so get approval ahead of time.

They depopulated the second motor driver in the steering rack during the chip shortages. Newer vehicles have more redundancy on the power and data topology so it's likely fully populated. The improved semiconductor supply helps too.
 
So we don't yet know if Highland has HW3.5 or HW4?

3.5 would have HW3 computer with HW4 cameras but upgradeable to HW4 computer.
It has HW 4.0 as seen in the parts catalogue from Norway
1699649294812.png
 
Or they take the new for 2023 5nm design already at TMSC and just shrink it.


I mean, I guess good for power consumption purposes... but if the capabilities are substantively different your choices become:

Run the same code on both and waste the extra available performance on the more powerful computer (they did this initially when HW3 came out- but eventually moved to....

or
Create Yet Another Class of Owner running a different, more performant, set of code on the newer HW


Choice 2 wasn't a big deal in a world that existed until HW4 came out, where if you became an FSD owner you got upgraded to the "current" computer.

But in this new you're stuck with whatever computer you've got world it'll become an issue over time....

(I suppose they might be doing planned upgradability for HW4 owners to go higher--but the fact they explicitly abandoned this approach with HW4 makes that less likely-- and doesn't help the majority of the fleet stuck at HW3 max)
 
I mean, I guess good for power consumption purposes... but if the capabilities are substantively different your choices become:

Run the same code on both and waste the extra available performance on the more powerful computer (they did this initially when HW3 came out- but eventually moved to....

or
Create Yet Another Class of Owner running a different, more performant, set of code on the newer HW


Choice 2 wasn't a big deal in a world that existed until HW4 came out, where if you became an FSD owner you got upgraded to the "current" computer.

But in this new you're stuck with whatever computer you've got world it'll become an issue over time....

(I suppose they might be doing planned upgradability for HW4 owners to go higher--but the fact they explicitly abandoned this approach with HW4 makes that less likely-- and doesn't help the majority of the fleet stuck at HW3 max)
Sure
I was just saying 4.5 might be at compute parity with 4.0, so not dusting previous generations more.

Although... could process and design improvements result in a HW4+ that is power compatible with HW2/HW3 units?
 
Sure
I was just saying 4.5 might be at compute parity with 4.0, so not dusting previous generations more.

Although... could process and design improvements result in a HW4+ that is power compatible with HW2/HW3 units?


I mean, it certainly could.... but that'd make the apparent intentional decision not to design HW4 for backward compatibility even weirder... (and trickier if you're designing 4.5 to be a drop in swap for 2/3 AND 4)
 
That summary is misleading
Tesla was previously at 7nm with Samsung
The changes listed are just due to the improvement of TMSC's 3nm process.
1.04 means 4% increase, not a doubling.


"A17 Pro is manufactured using TSMC's latest 3nm process (N3), with a transistor count of 19 billion. This is the first time that TSMC's 3nm process has been used on a top chip. The 3nm process will have 70% more transistors than the 5nm process, with the same power consumption. The speed can be increased by 15%, or the power consumption can be reduced by 30% at the same speed."
 

"A17 Pro is manufactured using TSMC's latest 3nm process (N3), with a transistor count of 19 billion. This is the first time that TSMC's 3nm process has been used on a top chip. The 3nm process will have 70% more transistors than the 5nm process, with the same power consumption. The speed can be increased by 15%, or the power consumption can be reduced by 30% at the same speed."


"The electronic chips are critical components, especially since there are only two companies that can manufacture the most advanced chips: Samsung and TSMC. These two giants duel for silicon supremacy, with the Taiwanese company having a slight edge over the Koreans. Tesla used both manufacturers at times. Early Autopilot chips were manufactured by Samsung using 14 nm lithography, which was later upgraded to a 7 nm process.

However, Tesla switched sides to TSMC with the Hardware 4 Autopilot chips, as the Taiwanese had a more advanced 5 nm technology. It could also deliver the quality and production scale Tesla needed for its current production needs. However, as the EV market leader is preparing the next-generation Autopilot, it will use both Samsung and TSMC chips in its computers. Clearly, Tesla anticipates a significant increase in vehicle production by the time HW5 arrives, which corresponds with the start of next-generation vehicle production.

In July, we wrote about the deal Tesla signed with Samsung to produce HW5 Autopilot chips using their 5 nm node. According to Korean media, the production was expected to start in 2026. At the time, Tesla had already signed a similar partnership with TSMC for the same purpose. Now we know more details about the TSMC contract.

The Taiwanese chip giant is currently using a more advanced 3 nm technology called N3E, which has attracted leading chip designers such as MediaTek, AMD, NVIDIA, Qualcomm, and Intel. TSMC will start mass production of its most advanced 3 nm technology, dubbed N3P, next summer, and Tesla has been listed among its customers.

Rumors suggest that Tesla plans to utilize the N3P process for its next-generation self-driving chips. It's also apparent that Tesla placed a huge order, potentially turning it into one of TSMC's most important clients. The updated N3P process provides a 5% improvement in performance, up to 10% reduction in power consumption, and a slight increase in chip density compared with N3E, which is in production today."