Model S/X Computer

[ny3ranger]

I know in 2016 tesla updated the computer with an NVIDIA processor. Does any owner know if their have upgraded the computer to a newer chipset from 2016. In the computer work a year is a long time.

Thinking about getting a model S but feel the computer on model 3 is probably a newer one then the one in the S/X.

 

[carbonx]

I know in 2016 tesla updated the computer with an NVIDIA processor. Does any owner know if their have upgraded the computer to a newer chipset from 2016. In the computer work a year is a long time.

Thinking about getting a model S but feel the computer on model 3 is probably a newer one then the one in the S/X.

From what I've heard it's the same. The late 2016+ Model S should be future proof from that aspect as they've marketed it to have full self driving at some point. Worse case, the chips are swapable in case Tesla needs to upgrade them.

 

[chillaban]

In 2016, Tesla upgraded (or rather, added) an Autopilot 2 ECU that is based off a newer nvidia Drive PX processor. It does not control the touchscreen or the instrument cluster. Those in the S/X are still the same nvidia Tegra 3 / Tegra 4 processors as before.

The model 3 does indeed have a newer MCU (touchscreen) computer, based off Intel Gordon Peak (part of the Atom family, nothing terribly exciting but definitely better than a Tegra 4). It's not known when the Model S/X may be getting this. It's part of the rumored S/X refresh that has been talked about for over a year.

 

[Snowstorm]

The current model S/X main touch screen is barely smooth enough to use. Feels like running the ill-fated ipad 3 (do you remember that one?). I really wish there can be an retrofitting upgrade that would allow us to swap that out to a new touch screen. It was ground breaking when it was revealed 2013, but in 2018, it is really showing its age. It is still way better than most cars today, but it have so much more potential if it would just have a bit more processing power.

 

[Canuck]

From what I've heard it's the same. The late 2016+ Model S should be future proof from that aspect as they've marketed it to have full self driving at some point. Worse case, the chips are swapable in case Tesla needs to upgrade them.

Future proof? Ha! Now that made me really laugh out loud, as did the swapable chips part. That won't happen but even if it did, it won't get you FSD since it still lacks redundancy, and you will never see them rewire vehicles and add the hardware now on the Model 3/S/X. You only need to read what you agreed to when FSD was marketed to you, and was even bolded (because Tesla knew it was a huge caveat). In case you missed it, you agreed to this:

Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available, as this is highly dependent on local regulatory approval.

It was marketed "dependent upon... regulatory approval" and I think we all know, including Tesla (as of mid-2017), that "regulatory approval" will require redundancy. That occurred mid-2017 according to Tesla's spokesperson:

“The internal name HW 2.5 is an overstatement, and instead it should be called something more like HW 2.1. This hardware set has some added computing and wiring redundancy, which very slightly improves reliability, but it does not have an additional Pascal GPU.” The spokesperson also said that all Model S, Model X, and Model 3 vehicles being ordered today (mid-2017) come with the new hardware.

Tesla tries to downplay this change adding redundancy, changing the name from 2.5 to 2.1 (sorry Tesla but a rose by any other name is still a rose) but whatever you call it, you can't downplay the importance of "redundancy" in relation to computer hardware and wiring. That's huge, especially to regulators which is obviously (at least in my view) why Tesla introduced it. As to upgrading AP2.0 to AP2.5 -- there's absolutely no way in my view that Tesla will add the hardware and wiring redundancy as a retrofit. We know from history Tesla does not do that even for more minor retrofits, since the service centers are already overwhelmed just dealing with the S/X -- and now with the Model 3 coming out in greater numbers forget about it. Tesla will simply say "we marketed it subject to regulatory approval and unfortunately the regulators will not approve AP2.0" then they will try to sell you on a new vehicle. It works much better for them to do it this way. There will also be no legal recourse given what was agreed to -- but I'm certain funds will be refunded for those who paid in advance for FSD subject to regulatory approval once it's confirmed the regulators will not approve it.

As I've said before, mid 2017 is going to be a huge date for FSD, at least in my view. It will make the Model 3 more desirable than pre-AP2.5 Model S/X to many people looking for FSD vehicles, which is the future.

 

[MarkS22]

Future proof? Ha! Now that made me really laugh out loud, as did the swapable chips part. That won't happen but even if it did, it won't get you FSD since it still lacks redundancy, and you will never see them rewire vehicles and add the hardware now on the Model 3/S/X. You only need to read what you agreed to when FSD was marketed to you, and was even bolded (because Tesla knew it was a huge caveat). In case you missed it, you agreed to this:



It was marketed "dependent on local regulatory approval" and I think we all know, including Tesla (as of mid-2017), that "regulatory approval" will require redundancy. That occurred mid-2017 according to Tesla's spokesperson:
Tesla tries to downplay this change adding redundancy, changing the name from 2.5 to 2.1 (sorry Tesla but a rose by any other name is still a rose) but whatever you call it, you can't downplay the importance of "redundancy" in relation to computer hardware and wiring. That's huge, especially to regulators which is obviously (at least in my view) why Tesla introduced it. As to upgrading AP2.0 to AP2.5 -- there's absolutely no way in my view that Tesla will add the hardware and wiring redundancy as a retrofit. We know from history Tesla does not do that even for more minor retrofits, since the service centers are already overwhelmed just dealing with the S/X -- and now with the Model 3 coming out in greater numbers forget about it. Tesla will simply say "we marketed it subject to regulatory approval and unfortunately the regulators will not approve AP2.0" then they will try to sell you on a new vehicle. It works much better for them to do it this way.

The 3 also has a second, redundant, power steering motor. I believe this is also crucial for regulatory approval. (Not to mention the interior “selfie cam.”)

I just got a 3 this week to replace my S coming off lease. I won’t get an S until the next refresh which hopefully has the same redundancies as the 3.

I love the S but moving between the 3 and S makes the S feel really dated.

 

[Canuck]

The 3 also has a second, redundant, power steering motor. I believe this is also crucial for regulatory approval. (Not to mention the interior “selfie cam.”)

I agree -- both are huge for FSD.

So much for the S/X will always lead with hardware, etc.

 

[verygreen]

Those in the S/X are still the same nvidia Tegra 3 / Tegra 4 processors as before

They are actually Tegra 2 / Tegra 3.

 

[doubleohwhat]

They are actually Tegra 2 / Tegra 3.

Any idea what the CPU load is? Just curious.

 

[DIL]

I love the S but moving between the 3 and S makes the S feel really dated.

Can you say more about this?

 

[verygreen]

Any idea what the CPU load is? Just curious.

depends on what you are doing. Load average with screens on on cid hovers around 5 nowadays (but that also takes slow io into account on Linux).

I never paid too much attention what it was on ic. Some current root users might want to comment here.

 

[navinsiri]

Can you say more about this?

The UI on the 3 looks a lot newer and is generally nicer to use with the exception of some missing features like the energy graph, but I'm sure that is something they can easily add with a future update. The touch screen also moves around like a smartphone. If you look at the map on the S/X, animations move around in steps and isn't really smooth. Animations on the 3 are very smooth and it's very quick to respond to touch.

 

[Snowstorm]

The 17" display in the Model S/X has low "FPS". It feels that it doesn't have enough compute horse power to have a smooth and fluid experience. There is nothing wrong with the graphical design, it looks good, just doesn't feel good when interacting because it is too slow. Taps takes a fraction of a second to respond (enough to feel unresponsive) and pan/drag is choppy. It is really a shame that we can't just swap out a chip for few dollars and dramatically upgrade the main non-driving part of the experience.

Heck, I would pay $$$ to upgrade the MCU if it works like as smooth as an iPAD pro, with full browser and app support.

 

[jlund]

I’m hoping that, like the LTE upgrade, this is offered as a future paid option. I think part of the reason they offered the cellular hardware upgrade is because it made their lives easier. It allowed cars to upload more data back to the mothership and offer updates to cars without wifi. A similar situation might happen for their MCU/IC computers.

Tesla is in a bit of a unique situation:

• Almost all of the in-car entertainment and information is through the displays.
• They have to support these cars for years.
• Current computers are already years old and arguably pushed to the limits

Given that, I see one of two things happening:

1. They fork the codebase. For all cars with Tegra 2/3 processors, they get a legacy experience. For new cars, they get a modern slick one.
2. They offer paid upgrades during your next service trip. Priced fairly reasonably.

I think which way they go largely depends on how feasible it is to upgrade.

 

[MarkS22]

Can you say more about this?

I posted my initial impressions here: Model 3: First 100 Miles Impressions from a P85D Owner

Hopefully this answers your questions.

 

[ny3ranger]

Okay but a S/X bought today can do 2.5 autopilot? Nothing to do with the MCU thats the screen? Or is the model 3 also "better" at autopilot?

 

[verygreen]

Okay but a S/X bought today can do 2.5 autopilot? Nothing to do with the MCU thats the screen? Or is the model 3 also "better" at autopilot?

Model3 has interior cam to watch the driver. Once they detect driver's eyes open up real-wide they hit the brakes In 3 months maybe 6 months definitely, that's it.

 

[SeminoleFSU]

Given that, I see one of two things happening:

1. They fork the codebase. For all cars with Tegra 2/3 processors, they get a legacy experience. For new cars, they get a modern slick one.
2. They offer paid upgrades during your next service trip. Priced fairly reasonably.

I think which way they go largely depends on how feasible it is to upgrade.

No offense but you haven't been watching Tesla very long if you think #2 is a possibility... They've allowed us to upgrade LTE, but I highly doubt we'll get to replace the MCU.... they'll probably want you to buy a new car for something like that.

 

[SeminoleFSU]

Model3 has interior cam to watch the driver. Once they detect driver's eyes open up real-wide they hit the brakes In 3 months maybe 6 months definitely, that's it.

I was thinking more along the lines of watching to make sure the eyes are focused on the road and if they aren't alert/chime and even disable autopilot for the remainder of the journey (behavior similar to today) if it sees you taking your eyes off the road too much... or closing them starting to fall asleep, etc. That seems like phase 1 functionality to me.

 

[chillaban]

I was thinking more along the lines of watching to make sure the eyes are focused on the road and if they aren't alert/chime and even disable autopilot for the remainder of the journey (behavior similar to today) if it sees you taking your eyes off the road too much... or closing them starting to fall asleep, etc. That seems like phase 1 functionality to me.

This camera is not the correct approach or mount angle for alertness tracking. Facial alertness tracking (gaze tracking) is done using active IR cameras mounted near the steering column or the instrument cluster pointed to the driver's eyes. Pure passive gaze tracking using a camera mounted at that location would be virtually unprecedented, and beyond the difficulty of detecting water droplets on a windshield which took over a year to have a beta-quality implementation...