Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/why-teslas-fsd-chip-is-a-big-deal-for-the-whole-automotive-industry.11351/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Why Teslas FSD Chip is a Big Deal for the Whole Automotive Industry

kloty

New member
Last week during a financial call Tesla has introduced a chip, which was created by Tesla themselves, they called it FSD, which stands for Full-Self-Driving. It is an exaggeration though, since experts say, that the capabilities of the design are sufficient for Level 2-3 autonomy, but definitely not 4-5, which would be full self driving. Nevertheless, it is a premiere that a car manufacturer is designing a chip, usually this is left to Tier-1 or Tier-2 suppliers like Infineon, Bosch or nowadays NVIDIA or Mobileye-Intel. The existence of the chip and the way it is advertised, might call a disruption in automotive world for several reasons.


In the past the IT world was ofter compared with cars, remember the infamous <a href=https://www.neowin.net/forum/topic/658078-joke-if-microsoft-made-cars/>„If Microsoft would build cars?“</a> joke. Now it is time to change the sides. Let’s look at the mobile phones market. In the past there were several Tier-1 companies, which provided application processors to mobile phones manufacturers, like TI, ST, Infineon, <a href=https://en.wikipedia.org/wiki/IPhone_(1st_generation)>Samsung</a>. Nowadays there are still Tier-1 companies, which are providing application processors like Qualcomm or MediaTek, but the top-notch phones have processors, which are created by the creators of the phones, like Apple with A-series processors, Samsung with Exynos or Huawei with Kirin. So now that Tesla has presented an own processor, there is a valid question, if other car manufacturers will also start designing their own processors in order to differentiate from the rest, show technical excellence and exclusivity. This might disrupt the whole established car supply chain.


Remember the days when it really mattered, which processor was in the PC you were buying. All technical details like processor speed, size of caches, bus and memory speed were important. The same was valid for cars, the cars were sold due to technologies like airbag, ESP (Electronic Stability Platform), power steering, ABS (Anti-lock breaking system), fuel injection system and so on. These days are gone long ago, most of the advertising for cars or PCs is not about technology, but about emotions, which are achievable using this technology. Having a long technical presentation during a financial call might bring the selling because of technology back, but now it is the electronics in the car, which really matter and ensures advantages of one manufacturer over the others. And since electronics became the most valuable asset in a car (in terms of financial value, but also in terms of competitiveness), the car manufacturers might not longer rely on the suppliers, but want to design the most important components themselves for the matter of differentiation.


One of the interesting parts of the presentation was the fact that the new computing platform has replaced the old one, which was based on NVIDIA. This is hardly possible in other cars, to simply (I don’t know the details, so might exaggerate here) replace the electronic heart of the whole system. So at least internally Tesla has reached some kind of standardization, which has yet to be achieved by other car manufacturers. This is absolutely required in the future, since the cycles of car renewal and car electronics renewal are completely different, a car might outlive several generations of car electronics, so there must be a standard, how the different generations of car electronics can be plugged into the car and the user might experience a new level of autonomy with the new hardware. Maybe even a completely new market for autonomy accelerators might be created here, when there are several different manufacturers of accelerators competing and exchanging the accelerator would be as easy as exchanging of car radios was in the past.






 
Good observations and a not unreasonable hypothesis. Picture on Tesla is a little murky though. It appears they are depending on Chinese technology for the Model 3 autopilot ECU, on which the US has recently denied an exemption on high tariffs:

U.S. rejects Tesla bid for tariff exemption for Autopilot 'brain' - Reuters

Is this a case where Tesla already plans to design out the Chinese chip, i.e. they really are building self-sufficient expertise, Apple-style?
 
Good observations and a not unreasonable hypothesis. Picture on Tesla is a little murky though. It appears they are depending on Chinese technology for the Model 3 autopilot ECU, on which the US has recently denied an exemption on high tariffs:

U.S. rejects Tesla bid for tariff exemption for Autopilot 'brain' - Reuters

Is this a case where Tesla already plans to design out the Chinese chip, i.e. they really are building self-sufficient expertise, Apple-style?

From reading the article it seems that the problem is not the chip but PCB and maybe packaging.
 
A chip does nothing unless it is programmed, so the software running on the hardware along with the sensors providing continuous inputs is what makes a self-driving car feasible. So let's see what Tesla can do with its own chip and software system. My contacts that own a Tesla are fiercely loyal to that brand, way more than any other auto brand that I can ever recall, almost cult-like, or Apple fan-boys.
 
The fan-boy characterization seems apt. When the major car makers and many others are saying true autonomy is still a long way off, and when there seems to be a rising level of desperation in Musk's claims versus delivery, it is difficult to have confidence they alone hold the keys to this nirvana. Perhaps only they have found the true path where no-one else has, but I wouldn't bet on it.
 
The fan-boy characterization seems apt. When the major car makers and many others are saying true autonomy is still a long way off, and when there seems to be a rising level of desperation in Musk's claims versus delivery, it is difficult to have confidence they alone hold the keys to this nirvana. Perhaps only they have found the true path where no-one else has, but I wouldn't bet on it.

It seems to me desperation too, but it is hard to tell when musk is involved.

He maintains that LIDAR is not needed that will keep the cost down and Tesla's today just need a software upgrade to reach full autonomy (level 5).

[video=youtube;dEv99vxKjVI]https://www.youtube.com/watch?v=dEv99vxKjVI[/video]
 
My 2c - Tesla made a pretty good case during their Autonomy Day presentation that they had the right "raw materials" to get autonomous driving right, though not immediately. But for me, fleet data set trumps the cool hardware. That Tesla been able to co-optimize chip hardware with machine learning using real fleet data and shadow mode opens the door to much faster advancement (how long have we seen the Google-mobiles creeping around the Bay Area - 8 years ??). They have been able to bring all three together (fleet data set, optimized hardware, machine learning) since they "own" all three.

[video=youtube_share;-b041NXGPZ8]https://youtu.be/-b041NXGPZ8?t=2992[/video]
 
Last edited:
Back
Top