Processing Speed of Touchscreen in Model X/S vs. Model 3

[Jayfaz]

This video is a great comparison of the processing speed of the touchscreen in Model X/S versus Model 3. When averaging the speed of different tasks on each model, the Model 3 average 4.44x faster than Model X/S.

Given that Model X/S are considerably more expensive than the Model 3, the processing speed simply should not be slower as it is the primary interface in the vehicle. What is Tesla going to do about this?

When I test drove an X last week, the Tesla "Copilot" said that they would be improving this via a software update. Seems to me that this is a hardware issue and the only way to improve it would be to update the processing chip. Please tell me if I'm wrong.

 

[davidc18]

We have been waiting for a hardware update of the MCU since 2013

 

[Snowstorm]

Software can help some if it is written more efficiently, but what it is likely more effective and easier to swap for a faster processor. Unfortunately I don’t think the MCU is design with this in mind so an entire MCU swap may be needed. I hope it won’t be too expensive as it is really a shame.

 

[verygreen]

Software can help some if it is written more efficiently, but what it is likely more effective and easier to swap for a faster processor. Unfortunately I don’t think the MCU is design with this in mind so an entire MCU swap may be needed. I hope it won’t be too expensive as it is really a shame.

Actually the cpu is mounted on a daughterboard so swapping it is not too hard.

What's somewhat hard is you need to create a new daughterboard with a different cpu to plug into that same socket (right side of the picture. The metal housing is for the fan).

Daughterboard is mostly self-contained, it has on-board emmc and other storage, and the socket mostly only has various i2c/gpio lines and ethernet.

 

[owen.shift]

What's somewhat hard is you need to create a new daughterboard with a different cpu to plug into that same socket.

But is it possible? What would the CPU for such an upgrade be? And would it work with the rest of the MCU (without a firmware update that would require Tesla's support)?
What would be the total cost to manufacture such an upgrade board?

 

[verygreen]

But is it possible? What would the CPU for such an upgrade be? And would it work with the rest of the MCU (without a firmware update that would require Tesla's support)?
What would be the total cost to manufacture such an upgrade board?

I am not in pcb manufacturing so I cannot quote you any numbers.
Technically just a daughterboard replacement looks very possible to me having my somewhat limited knowledge about the system.

Obviously when we are talking about the daughterboard - it needs to come from Tesla because they have all the info, and they will bring matching firmware with it.

CPU could be pretty much anything as long as it does not consume more than the power rated for the current tegra is (so probably a lot of modern much faster options). Since Tesla already uses x86 on model 3 it only makes sense it'll be x86. And this also matches well with other x86 cid development traces I was able to find elsewhere.

Now I am not saying they are going to make it just a daughterboard replacement as opposed to a whole cid replacement (and I don't know if they decide to make it retrofittable or not either!). But assuming they care about already sold cars, doing a new daughterboard is a way to make all old cars work with it where as they'd need to make multiple cid versions for older generation hardware in the case of whole cid replacement which is probably less desirable.

 

[jaguar36]

When I test drove an X last week, the Tesla "Copilot" said that they would be improving this via a software update. Seems to me that this is a hardware issue and the only way to improve it would be to update the processing chip. Please tell me if I'm wrong.

It certainly could be done just with software. The Tegra is more than capable of handling the tasks, its just a matter of optimizing the software. Optimization is expensive though, and the question is if Tesla will bother. Considering they haven't for the last 5 years, and in fact have made it worse, it seems highly unlikely that they would at this point.

Much more likely that they will just introduce new hardware and ignore all the older cars. (which just to put into perspective is how every other car maker works)

 

[owen.shift]

I am not in pcb manufacturing so I cannot quote you any numbers.
Technically just a daughterboard replacement looks very possible to me having my somewhat limited knowledge about the system.

Obviously when we are talking about the daughterboard - it needs to come from Tesla because they have all the info, and they will bring matching firmware with it. [...]

Tesla will never offer such an upgrade for already sold cars; for the same reason they will never offer a battery upgrade: they want to sell new cars.
I was asking to see if it was possible to create such an upgrade card independently. If it brings MCU performance on par with Model 3, I am sure that more than enough Model S and X owners would pay between $1,000 and $2,000 for it to make that a pretty lucrative business.

 

[verygreen]

Tesla will never offer such an upgrade for already sold cars; for the same reason they will never offer a battery upgrade: they want to sell new cars.
I was asking to see if it was possible to create such an upgrade card independently. If it brings MCU performance on par with Model 3, I am sure that more than enough Model S and X owners would pay between $1,000 and $2,000 for it to make that a pretty lucrative business.

Ok, I am pretty sure it's technically possible, but there are a few kinks in this all.

1. If you want future firmware updates from Tesla to work - you need to either emulate the old cpu and use vanilla firmwares - the speed boost is unlikely to be as good as for a native implementation, but still something, I suspect. Or you might reimplement one version and be stuck with it - obviously not great, esp for ap2+ cars (and if you plan to update it from time to time - that's more ongoing costs).
2. I suspect the development costs still would be somewhat massive so not sure at $2k it would be economical, even if you pay off some existing/former Tesla employee to sell you some of the missing details instead of fully reverse-engineering them.
3. Then there's the whole warranty issue.