Computing platform reveal: Intel and nVidia

[anticitizen13.7]

Tesla Model 3: first look at new dual computing platform Tesla developed for Autopilot and MCU

Liquid cooled module w’ 2 boards: Intel Apollo Lake for multimedia system and nVidia GPU for Autopilot.

 

[diamond.g]

Will be interesting to see exactly which chip they chose from Intel.

 

[TexasTeslaRacing]

Isn't this really a Mobile Eye platform since Intel bought them out?

 

[diamond.g]

Isn't this really a Mobile Eye platform since Intel bought them out?

The MCU is running Intel Atom which is a line of system on chips that has existed for quite a few years. I don’t think it has anything to do with MobileEye.

 

[JeffK]

Correct, Intel Atom has existed for nearly ten years for general purpose, low power, applications. They used to suck pretty bad, but from the videos of Model 3 the new ones seem pretty snappy.

 

[Cosmacelf]

I wonder ... did Tesla extract a ‘concession’ from Intel to allow Tesla to buy Mobileye chips again? I know Tesla changes technology like underwear (ie, often), but swapping out the MCU wouldn’t be done lightly. You would think there would be a good reason for the switch. Cost? Performance? Access to Mobileye chips? I’m just suspicious about Tesla’s inability to get parity with AP v1 still, I wonder if they decided to have a fall back.

 

[Mkorpal]

I wonder ... did Tesla extract a ‘concession’ from Intel to allow Tesla to buy Mobileye chips again? I know Tesla changes technology like underwear (ie, often), but swapping out the MCU wouldn’t be done lightly. You would think there would be a good reason for the switch. Cost? Performance? Access to Mobileye chips? I’m just suspicious about Tesla’s inability to get parity with AP v1 still, I wonder if they decided to have a fall back.

This has nothing to do with Mobileye.

 

[Twiglett]

This has nothing to do with Mobileye.

Exactly. ^^^
It’s pretty obvious from reading the article that the AP hardware is still Nvidia.
The big change is using an Intel proc for the media system etc.

 

[Cosmacelf]

Exactly. ^^^
It’s pretty obvious from reading the article that the AP hardware is still Nvidia.
The big change is using an Intel proc for the media system etc.

I didn't say they were looking to dump nVidia for AP (although there are rumors they will eventually do just that when they build a custom chip set), I said that maybe they will go back to augmenting nVidia with Mobileye until they can get AP on nVidia working better than AP 1.0. That was Tesla's original plan before Mobileye got pissed off at them and declined to sell them chips. Now that Intel owns Mobileye, the strategic calculation changes and no doubt Intel would love to see more Intel chips in Tesla and Tesla would have been a fool to not extract a concession to allow them to buy Mobileye chips going forward should they choose to now that Tesla is using Intel for the MCU.

The only question is whether the AP board for Model 3 ALSO contains the Mobileye chip or not - remember than Tesla has ALREADY done the hardware engineering for this - there is actually a spot on the existing Model S/X Nvidia board for a Mobileye chip.

 

[diamond.g]

Correct, Intel Atom has existed for nearly ten years for general purpose, low power, applications. They used to suck pretty bad, but from the videos of Model 3 the new ones seem pretty snappy.

Oh it still stinks, just not as bad as ARM does (ignoring Apple SOC).

 

[diamond.g]

I didn't say they were looking to dump nVidia for AP (although there are rumors they will eventually do just that when they build a custom chip set), I said that maybe they will go back to augmenting nVidia with Mobileye until they can get AP on nVidia working better than AP 1.0. That was Tesla's original plan before Mobileye got pissed off at them and declined to sell them chips. Now that Intel owns Mobileye, the strategic calculation changes and no doubt Intel would love to see more Intel chips in Tesla and Tesla would have been a fool to not extract a concession to allow them to buy Mobileye chips going forward should they choose to now that Tesla is using Intel for the MCU.

The only question is whether the AP board for Model 3 ALSO contains the Mobileye chip or not - remember than Tesla has ALREADY done the hardware engineering for this - there is actually a spot on the existing Model S/X Nvidia board for a Mobileye chip.

I may be mistaken, but doesn’t the Mobileye chip exist where the camera is housed? This unit appears to be housed where the AP brain is housed which means it isn’t in the camera housing.

 

[scottm]

Don't really care if Tesla reconstitutes the Commodore 64 factory line in efforts to get a working computer in the car.

So long as the car does what it's supposed to do, reliably... I'd be happy.

 

[voip-ninja]

I didn't say they were looking to dump nVidia for AP (although there are rumors they will eventually do just that when they build a custom chip set), I said that maybe they will go back to augmenting nVidia with Mobileye until they can get AP on nVidia working better than AP 1.0. That was Tesla's original plan before Mobileye got pissed off at them and declined to sell them chips. Now that Intel owns Mobileye, the strategic calculation changes and no doubt Intel would love to see more Intel chips in Tesla and Tesla would have been a fool to not extract a concession to allow them to buy Mobileye chips going forward should they choose to now that Tesla is using Intel for the MCU.

The only question is whether the AP board for Model 3 ALSO contains the Mobileye chip or not - remember than Tesla has ALREADY done the hardware engineering for this - there is actually a spot on the existing Model S/X Nvidia board for a Mobileye chip.

You'd have to explain how a general purpose Intel Atom CPU has anything to do with auto-pilot... because it does not have the horsepower and is not designed for that application.

Now if you want to argue that Tesla threw Intel a bone by purchasing Atom for the multi-media system to make inroads with them for some future AP hardware purchase, only a handful of execs at Tesla would have any insight into that.