Tesla Is Making Use Of The Open-Source Coreboot Within Their Electric Vehicles

[discordian]

Look up "tivoization". This case is no different. It's a device you own, running free software, that artificially blocks you from exercising the freedom to modify it.

Speaking in terms of the car world, as the owner, you should definitely be able to tune your car yourself (or take it to a third party shop for that), period.

Good idea, users hacking a self driving car with a ton of burning-like-hell lithium. Do you have to turn in your home to insurance companies to drive one?

[Vistaus]

AGL - Automotive Grade Linux may be the greatest, most successful move for general Linux adoption and sustained evolution after supercomputers.

I mean, they got like half the big car companies in the world paying big cash to devs, to do dead-serious work on the whole software stack, stability and security concerns are an integral part of their worries, and they really seem to be happy and supportive of FOSS principles as the work evolves... and now even some open hardware principles seem to be seeping in... into car companies!

It looks plain awesome!

Half of the manufacturers? Au contraire, my friend. A lot of manufacturers actually use QNX instead of AGL.

[marlock]

If I understood correctly, at the moment QNX has only partial overlap with AGL, and manufacturers can sign up to work with AGL regardless of using QNX in their vehicle... even using parts of both in the same vehicle (eg: AGL-based infotainement + QNX critical systems)

What I saw to make me say so:

https://utbblogs.com/truth-blackberry-qnx-agl/

Which makes sense since QNX is more well established already while AGL is a relatively recent initiative, not yet past all safety critical certifications, etc... but doesn't mean AGL isn't generating widespread interest and engagement in development from many big automotive players.

Note how I said nothing about the amount of vehicles being sold with AGL onboard, because at the moment there are very few in any part of the vehicle software stack... but that's likely to gradually change as AGL matures, and a lot of companies are working on it... which is what I already see as a win for Linux as a broader ecossystem, even if it doesn't instantly dominate the market in all levels...

ps: one cool thing about FOSS in general and Linux specifically: even if it never pans out, the work done until then will have already benefited Linux as a broader whole... and AGL has been careful to contribute upstream to ensure so.

[tildearrow]

I wonder why Tesla designed their own hardware instead of just using SoC from Qualcomm, Marvell, Broadcom, etc.

Lack of fitness for their purposes.

[starshipeleven]

I wonder why Tesla designed their own hardware instead of just using SoC from Qualcomm, Marvell, Broadcom, etc.

Neither make a half-way powerful enough system for a self-driving car. Only one is NVIDIA and knowing them they were probably ridicolously expensive and not focused enough on what they want to do.

Most modern AI-grade hardware accelerators aren't a GPU but something custom and designed jfor the type of algorithm

[uid313]

Neither make a half-way powerful enough system for a self-driving car. Only one is NVIDIA and knowing them they were probably ridicolously expensive and not focused enough on what they want to do.

Most modern AI-grade hardware accelerators aren't a GPU but something custom and designed jfor the type of algorithm

But Google's TensorFlow runs on Qualcomm Snapdragon mobiles.
But yeah maybe just Nvidia fills the niche of offering high performance SoC.
I don't think Intel, AMD or POWER have any AI offerings. The things from Marvel, NXP, Qualcomm and Samsung might not have anything powerful enough.

[starshipeleven]

But Google's TensorFlow runs on Qualcomm Snapdragon mobiles.

And NVIDIA PhysX can run on CPU. The fact that it can run does not mean that it has useful performance for every real-life load you can throw at it. Driving a car is a very big load. Snapdragons can run image or sound recognition or other stuff like that that is still AI, but not at the scale required by driving a car.

Also SnapDragons have no dedicated AI coprocessor, it's all run on the CPU or GPU.

But yeah maybe just Nvidia fills the niche of offering high performance SoC.

Yes but their SoC's "AI" capability is still provided by a general-purpose GPU with Cuda and all that stuff.

For Tesla and anyone doing a custom design, the hardware would be more similar to a Parallela https://www.adapteva.com/parallella/ where you have a weak general-purpose ARM CPU doing basic stuff like I/O and a dedicated custom co-processor optimized to run whatever AI algorithms they want to run. Of course the Parallela is still a weak device for the job, but it would be similar, just bigger.

I don't think Intel, AMD or POWER have any AI offerings. The things from Marvel, NXP, Qualcomm and Samsung might not have anything powerful enough.

Intel is shopping for AI tech startups and they bought the second one so far, but no there is no real decent "custom AI" offering yet, if you want high performance you have to design your own.

[pal666]

Meanwhile all Tesla's software is locked down, and they are literally charging you $2000 for a software fix that allows the car you already bought go faster.

but you bought slower car, why are you whining now? intel sold software locked cpus, tesla sold software locked cars because apparently it was cheaper to sell some cars with less profit than to additionally produce smaller battery pack

[pal666]

Speaking in terms of the car world, as the owner, you should definitely be able to tune your car yourself (or take it to a third party shop for that), period.

and then provide all warranty by yourself when your tuned battery dies in a month

[pal666]

car with a ton of burning-like-hell lithium.

what are you smoking? gasoline is burning like hell, not lithium.
illustration https://www.youtube.com/watch?v=ub5NjC7np7E
btw, largest tesla battery weights much less than ton, and lithium takes up few percent of its weight. for example "A 70kWh Tesla battery uses 63kg of Lithium Carbonate Li2CO3, of which 19% or 12kg is Lithium." so you have 12 kg of benign lithium vs dozens of kg of explosives(you know ice cars work on fast explosions, don't you?)