Wiki Consolidated eMMC Thread (MCU repair) (Black Center Screen)

[TonyT]

Again, why should I accept a used part for anything?

You can ask them for a MCU2, new MCU1 likely does not exist anymore.

 

[TessP100D]

Thanks T. I really appreciate it.

 

[mspohr]

So I’m clear.... I bought a brand new car with all new parts. One of the parts included a design flaw that caused a safety that was formally recalled. Regardless of if my 2017 P100D is new or not, it is under warranty. So what you are saying is I don’t deserve a new part but a used or remanufactured part?

Read your warranty

 

[Akikiki]

If new parts to fix my MCU1 are not available, why would that be my problem? Wouldn’t Tesla need to figure that out without providing me with used parts?

You do know don't you, because we do, that you are being unreasonable about remanufactured parts?

Do you really think that if your drive unit or your battery failed that they would put a new one brand new off the assembly line one in your car? If you think they would, please stand off to the side and watch intently what happens. Just stand there and wait for it.

And back to our scheduled program. You do realize that if you don't want that remanufactured $`125 daughter board with the new Micron chip on it you can refuse it. Of course you will want to take your place over there, on the side with that other guy that's watching for his brand new battery or drive unit.

I'm not trying to upset you. I sincerely hope for you that you get what you want. This is just my sort of trying to be funny way to say that I don't think you will get it.

 

[Zuikkis]

The way I look at it is, I bought a new car and while,under warranty if a part needs replacing, it’s should be a new part

That's exactly what they are doing, replacing the faulty eMMC chip with a new part.

It's just that it is a bit difficult to replace and needs special tools and skills. To speed up the process, they take your old Tegra board, send it away to be repaired, and in the meanwhile install someone else's board to your car.

I guess if you feel it's really important you get your original part back, you can let the car sit a few weeks (months?) in the service center waiting for your own board to come back..

 

[TessP100D]

But is it really the chip that went bad or is it the size of the chip? Could the old chip have worked fine is a Tesla hadn’t cooked it by useless OTA updates? If all the 8g chips were going to fail 100%, does that mean under the same system the 64g chip will fail too, only it will take longer?

 

[tomas]

When one argumentative type doesn’t listen and starts making 50% of posts in a thread, it’s time to unsubscribe.

 

[TessP100D]

Argumentative type? Wait I see your out of warranty. Ok. I get it. Funny.

 

[brkaus]

But is it really the chip that went bad or is it the size of the chip? Could the old chip have worked fine is a Tesla hadn’t cooked it by useless OTA updates? If all the 8g chips were going to fail 100%, does that mean under the same system the 64g chip will fail too, only it will take longer?

OTA updates are nothing compared to the logging that they were doing on the past.

 

[mspohr]

But is it really the chip that went bad or is it the size of the chip? Could the old chip have worked fine is a Tesla hadn’t cooked it by useless OTA updates? If all the 8g chips were going to fail 100%, does that mean under the same system the 64g chip will fail too, only it will take longer?

You seem to be somewhat immune to information that contradicts your preconceived notions but here goes:
- The original chips wore out from too many write cycles. As each block wears out, it is marked as a bad block. If you have too many bad blocks, the memory capacity is reduced to the point where you get flaky MCU problems. The original Tesla software wrote a lot of data to the memory as it was logging different activities.
- Tesla fixed this "excessive" logging with a (not useless) OTA update a while ago. There is much less logging now and chips are failing at a lower rate but there are still a lot of partially worn out chips out there... hence the recall.
- The new 64 gig chip is 8 times as big and since the software is logging less data should not wear out like the old 8 gig chip.

 

[MP3Mike]

- The original chips wore out from too many write cycles. As each block wears out, it is marked as a bad block. If you have too many bad blocks, the memory capacity is reduced to the point where you get flaky MCU problems. The original Tesla software wrote a lot of data to the memory as it was logging different activities.

There is another failure mode: some of the 8GB eMMC chips are suffering from a controller failure and not blocks failing from excessive writes. (Though they would have eventually failed because of writes had the controller not failed first.)

 

[Akikiki]

But is it really the chip that went bad or is it the size of the chip? Could the old chip have worked fine is a Tesla hadn’t cooked it by useless OTA updates? If all the 8g chips were going to fail 100%, does that mean under the same system the 64g chip will fail too, only it will take longer?

Here. Please read this. I think it will answer your questions. MCU1 Flash Memory Analysis and Failures – TeslaTap

 

[TessP100D]

Ok I will read it all. Thanks you very much.

 

[M0rpH3x]

There is another failure mode: some of the 8GB eMMC chips are suffering from a controller failure and not blocks failing from excessive writes. (Though they would have eventually failed because of writes had the controller not failed first.)

Indeed, I find it interesting that they replaced the chip by another one with a very high temperature rate (-40ºC - 105ºC).

 

[appleguru]

Again, why should I accept a used part for anything?

Because the terms of the warranty say so?

 

[TonyT]

Indeed, I find it interesting that they replaced the chip by another one with a very high temperature rate (-40ºC - 105ºC).

Correct, however the controller issue was mostly on one specific hynix chip part number, I dont have handy currently. We are expecting the Micron chip to have a better controller and more write cycles to mitigate the issues.

 

[M0rpH3x]

Correct, however the controller issue was mostly on one specific hynix chip part number, I dont have handy currently. We are expecting the Micron chip to have a better controller and more write cycles to mitigate the issues.

are you implicitly suggesting that the more standard -40ºC - 85ºC can do fine? I would certainly expect so, this is no ICE car, and the residual heat is much less.

 

[TonyT]

are you implicitly suggesting that the more standard -40ºC - 85ºC can do fine? I would certainly expect so, this is no ICE car, and the residual heat is much less.

A 85ºC chip should be fine, and what was used for 8 years. A 105ºC should be better. 85ºC = 185ºF I've recorded dash temps over 140ºC in other cars, didn't try the S. I'll have to dig back but think I have some running temps from the mcu.

I'm stating we saw the specific hynix chip used around 2016 + had most of the controller failure issues, I just dont have that part number handy. Several different Hynix chips were used over the years . Unreadable chips due to controller failure were a small percentage of the boards we repaired. Inside the car can get pretty warm and the mcu is running all the time. 85ºC may be good enough but the 105ºC is better. Being this in inside the car I don't think ICE or EV will make a difference. Tesla cools the car when parked, but not all EV's do that.

 

[Akikiki]

@TonyT, saved this in case it needed to be referenced again.

Preventive eMMC replacement on MCU1 Post #378

From my observations I've seen the Hynix H26M42001FMR in 2012 and 2013 cars. The H26M42002GMR in ~ 2014 cars and the H26M42003GMR in 2015+ cars.
Please be very careful when de-soldering the chip, especially the oldest chips (H26M42001FMR) seem to corrupt or even die completely from the heating.
There are ways to read the eMMC chip before de-soldering though. Especially when you have an old chip or if it is failing already I would recommend having this done to safeguard your unique certificate files. Without those unique files the car will not have app functionality, updates, Spotify etc. (and Tesla will not give you these files, they will only be able to offer a +- €3000 replacement of the entire MCU).

Hynix H26M42001FMR in 2012 and 2013 cars.
Hynix H26M42002GMR in ~ 2014 cars.
Hynix H26M42003GMR in 2015+ cars.

 

[TonyT]

@TonyT, saved this in case it needed to be referenced again.

Preventive eMMC replacement on MCU1 Post #378

From my observations I've seen the Hynix H26M42001FMR in 2012 and 2013 cars. The H26M42002GMR in ~ 2014 cars and the H26M42003GMR in 2015+ cars.
Please be very careful when de-soldering the chip, especially the oldest chips (H26M42001FMR) seem to corrupt or even die completely from the heating.
There are ways to read the eMMC chip before de-soldering though. Especially when you have an old chip or if it is failing already I would recommend having this done to safeguard your unique certificate files. Without those unique files the car will not have app functionality, updates, Spotify etc. (and Tesla will not give you these files, they will only be able to offer a +- €3000 replacement of the entire MCU).

Hynix H26M42001FMR in 2012 and 2013 cars.
Hynix H26M42002GMR in ~ 2014 cars.
Hynix H26M42003GMR in 2015+ cars.

thanks, but why not direct link the post? Preventive eMMC replacement on MCU1