Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

FSD Beta for MCU1 cars

This site may earn commission on affiliate links.
I have HW3 computer (upgraded last month) but MCU1… I purchased the car used from Tesla which advertised it as having FSD which was a $15k option at the time I purchased it.
Then you're good to go.

 
Then you're good to go.

What part about MCU1 do you continue to not understand? Its been explained countless times.

Tegra aka MCU1 is not and will never be compatible with FSD. Doesn't matter if you have HW3.0, 4.0, or 52.0...it won't work.
 
  • Disagree
  • Like
Reactions: t3sl4drvr and Krash
What part about MCU1 do you continue to not understand? Its been explained countless times.

Tegra aka MCU1 is not and will never be compatible with FSD. Doesn't matter if you have HW3.0, 4.0, or 52.0...it won't work.
MCU1 has been barely limping along since late 2018 or so anyway. It's time for Tesla to admit that they aren't capable of making MCU1 be usable again, and just upgrade all MCU1 hardware to MCU2, or at least for anybody who paid for FSD. I mean, even if they upgraded every MCU1 car, it would still be less than a quarter million cars in total, and the total real-world manufacturing cost for MCU2 is probably no more than a couple of hundred dollars. For fifty million or thereabouts, the whole problem goes away. That's probably less than the cost of the staff required to make FSD happen on MCU1. And if they do it just for people who bought FSD, it's probably single-digit millions, which is almost definitely less than it would cost for the extra staffing.
 
MCU1 has been barely limping along since late 2018 or so anyway. It's time for Tesla to admit that they aren't capable of making MCU1 be usable again, and just upgrade all MCU1 hardware to MCU2, or at least for anybody who paid for FSD. I mean, even if they upgraded every MCU1 car, it would still be less than a quarter million cars in total, and the total real-world manufacturing cost for MCU2 is probably no more than a couple of hundred dollars. For fifty million or thereabouts, the whole problem goes away. That's probably less than the cost of the staff required to make FSD happen on MCU1. And if they do it just for people who bought FSD, it's probably single-digit millions, which is almost definitely less than it would cost for the extra staffing.
Much easier just to ignore it and upgrade the ones who send an official letter requesting it.

I'll give them a couple weeks and then I'll have a lawyer write a strongly worded letter. Time to put the $10/month lawyer subscription to work. I'd recommend everyone else do the same.

And no, I didn't get the lawyer subscription for this...it's for my rental property. Ala carte letter probably run $200.
 
  • Like
Reactions: Tesomega
Much easier just to ignore it and upgrade the ones who send an official letter requesting it.

I'll give them a couple weeks and then I'll have a lawyer write a strongly worded letter. Time to put the $10/month lawyer subscription to work. I'd recommend everyone else do the same.

And no, I didn't get the lawyer subscription for this...it's for my rental property. Ala carte letter probably run $200.
Just post a copy for everyone to copy with their details… :)
 
What part about MCU1 do you continue to not understand? Its been explained countless times.

Tegra aka MCU1 is not and will never be compatible with FSD. Doesn't matter if you have HW3.0, 4.0, or 52.0...it won't work.
Settle down. People here confuse the FSD computer with the infotainment computer all the time.

Can't you get an upgrade to MCU2? I've read others here get the upgrade, some at no charge and some at cost.
 
  • Disagree
Reactions: dbldwn02 and Krash
…Tegra aka MCU1 is not and will never be compatible with FSD. Doesn't matter if you have HW3.0, 4.0, or 52.0...it won't work.
I disagree and have never heard a technical reason why MCU1 wouldn’t work. Now I think it’s possible that FSD Beta never gets released for MCU1, but likely not because it can’t, but rather because the trouble involved in dumbing down, porting the visualizations and testing them is so much work. So when FSD comes out of Beta, do I think Tesla would release an FSD version without the ICU visualizations on MCU1? Maybe. But if they don’t, I don’t think it will be because they couldn’t make it work, I think it would be because they just don’t want to.

I’m sure there are those on this forum that can enlighten us on how HWx takes inputs from the cameras and controls the steering accelerator, regen and braking. Is the MCU really part of that loop? The light input to engage it and modify settings clearly would not strain the Tegra. Do people really think that the Navigation input or some mix of GPS and Nav requirements would overwhelm the Tegra?

…upgrade to MCU2? I've read others here get the upgrade, some at no charge and some at cost.
I have never heard of anyone getting an MCU2 upgrade for free. HW3? Yes. Cameras? Yes. MCU2? No.
 
Last edited:
I disagree and have never heard a technical reason why MCU1 wouldn’t work. Now I think it’s possible that FSD Beta never gets released for MCU1, but likely not because it can’t, but rather because the trouble involved in dumbing down, porting the visualizations and testing them is so much work. So when FSD comes out of Beta, do I think Tesla would release an FSD version without the ICU visualizations on MCU1? Maybe. But if they don’t, I don’t think it will be because they couldn’t make it work, I think it would be because they just don’t want to.

I’m sure there are those on this forum that can enlighten us on how HWx takes inputs from the cameras and controls the steering accelerator, regen and braking. Is the MCU really part of that loop? The light input to engage it and modify settings clearly would not strain the Tegra. Do people really think that the Navigation input or some mix of GPS and Nav requirements would overwhelm the Tegra?


I have never heard of anyone getting an MCU2 upgrade for free. HW3? Yes. Cameras? Yes. MCU2? No.
What does it cost? Why not upgrade it?
 
I disagree and have never heard a technical reason why MCU1 wouldn’t work. Now I think it’s possible that FSD Beta never gets released for MCU1, but likely not because it can’t, but rather because the trouble involved in dumbing down, porting the visualizations and testing them is so much work. So when FSD comes out of Beta, do I think Tesla would release an FSD version without the ICU visualizations on MCU1? Maybe. But if they don’t, I don’t think it will be because they couldn’t make it work, I think it would be because they just don’t want to
.
6 years ago, Autopilot was in BETA, it currently remains in BETA. Do you think the explanation for why FSD Beta may not work makes sense?
 
6 years ago, Autopilot was in BETA, it currently remains in BETA. Do you think the explanation for why FSD Beta may not work makes sense?
I think the reason FSD Beta doesn’t run on MCU1 is because Tesla doesn’t bother to release a version for it. I think my own explanation makes sense to me. Or are you asking about someone else’s explanation?
 
I think MCU1 cars are fully capable of FSD, though I doubt visualizations will be possible which are vital at this time to inform the 'tester' of vehicle intent. I'm curious though, if someone turns off FSD visualizations that is in the beta; Does it work?
 
I disagree and have never heard a technical reason why MCU1 wouldn’t work. Now I think it’s possible that FSD Beta never gets released for MCU1, but likely not because it can’t, but rather because the trouble involved in dumbing down, porting the visualizations and testing them is so much work. So when FSD comes out of Beta, do I think Tesla would release an FSD version without the ICU visualizations on MCU1? Maybe. But if they don’t, I don’t think it will be because they couldn’t make it work, I think it would be because they just don’t want to.

One challenge is that the visualizations may be considered essential to monitoring the system for safety. And it's not strictly a porting question; there are three computer systems involved with MCU1 instead of two, so it gets a lot more complicated. :)


I’m sure there are those on this forum that can enlighten us on how HWx takes inputs from the cameras and controls the steering accelerator, regen and braking. Is the MCU really part of that loop? The light input to engage it and modify settings clearly would not strain the Tegra. Do people really think that the Navigation input or some mix of GPS and Nav requirements would overwhelm the Tegra?

The MCU is barely able to handle the camera input from HW3 as it is. With navigate-on-autopilot, there are a lot fewer opportunities for its slowness to cause serious problems than with FSD, where it has to handle a lot more turns in a much shorter period. So the answer is a definite maybe.
 
One challenge is that the visualizations may be considered essential to monitoring the system for safety. And it's not strictly a porting question; there are three computer systems involved with MCU1 instead of two, so it gets a lot more complicated. :)




The MCU is barely able to handle the camera input from HW3 as it is. With navigate-on-autopilot, there are a lot fewer opportunities for its slowness to cause serious problems than with FSD, where it has to handle a lot more turns in a much shorter period. So the answer is a definite maybe.
Why would FSD care about the camera output from HWx to MCU1 if the core of FSD runs on HWx? Why would the slowness of the MCU cause HWx to be able to make fewer turns?

I could see how there would be more dependence on the MCU if Tesla depended on map data, since the map comes from the MCU. But my understanding is that this data stream is pretty light.
 
Last edited:
Why would FSD care about the camera output from HWx to MCU1 if the core of FSD runs on HWx? Why would the slowness of the MCU cause HWx to be able to make fewer turns?

I could see how there would be more dependence on the MCU if Tesla depended on map data, since the map comes from the MCU. But my understanding is that this data stream is pretty light.

Light or not, the point is that a few seconds of latency makes little difference when you have ten miles to the next exit, but potentially a huge difference if you're turning in 100 feet.

To put this in context, my MCU1 car takes a good half a minute to render a simple zoom change on the map, which tells me that its ability to deal with map data in a timely manner is badly compromised. It has been this way since either the version 9 update in late 2018 or the version 10 update in late 2019. I can't remember which — only that it was at the end of a year, and it was one of the huge updates.