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

Memory cache full

This site may earn commission on affiliate links.
I have an appointment with Tesla tomorrow because after a 12v replacement my airbag light has stayed on. Over 2k miles and not sure if my airbags work or not (the debate on to drive or not to drive is not in question here)

I scheduled an appointment to have this fixed and was told that my slow
to boot MCU and IC was because the memory cache was “very full” they supposedly cleared it remotely and said things would be slow the first drive as everything reloads.

does this scream to anyone else Emmc failure imminent? Not sure if I should push further or wait it out? My warranty will be up mid may of 21.
 
As an S owner who has endured two dead/dying MCU's, I would definitely keep a close eye on the MCU behavior and, no matter what, as your warranty expiration date approaches I would make enough of a ruckus that Tesla at minimum replaces the Terga board (the daughterboard that contains the eMMC chip, which is the culprit in MCU failures of this type), if not your entire MCU. In this way, you'll have a trouble free experience for as long as possible after your warranty expires.

There are a bunch of long and very detailed threads on the subject of out-of-warranty MCU repairs, third-party MCU repairs, etc. if you want to torture yourself pouring through all of those. The summary of all of it is that the eMMC chip that Tesla used/uses isn't very high quality and it fails, which in the past required the entire MCU to be replaced. If you're under warranty they're going to replace/repair it however they decide. Out of warranty, and depending upon the model of your vehicle, Tesla is offering to replace the MCU with an MCU1 or a newer MCU2 for around $1200 or $2500, respectively. In response to griping about needing such an expensive repair for one little failed chip, Tesla has started offering in recent months a replacement of the daughterboard that contains the eMMC memory chip for around $500. This is less painful than replacing the full MCU, but still not a delightful prospect. And, finally, there are folks out there who will work on the MCU for you and replace the low quality eMMC chip with a higher quality one that also has more capacity so that the repair is presumably more future proof than what Tesla does.
 
  • Informative
Reactions: mambro
They described it inaccurately as "memory cache". That's literally in memory (RAM), but what they did was clear something in the flash chip, probably in the /var (changeables) directory.

If so, this screams software bug -- maybe logfiles are not being limited, cache or spool files not being reaped as they should be, etc.

Doesn't sound like a failing eMMC yet, but count on it, it will come.
 
  • Informative
Reactions: Alysashley79
If I wasn’t under warranty I’d just pay for the daughterboard tomorrow. (On an older car I really can’t justify MCU2) I’ve just never heard them say that they can remotely see that the cache is full and that they can wipe it. My concern is that it sounds to me like wiping cookies which just fill back up quickly as you use it. I just don’t want them to get me to the point just outside warranty so that I am then left needing to replace it.

all of this started on a road trip last month to Arizona. As we were arriving at our first super the car went into a cycle of repeated reboots (while driving) arrived at the super and it didn’t want to charge. I hard rebooted and we were on our way again. Remote said everything looked good so we continued. I don’t know how many times we had to reboot after that over the two weeks we were gone as I lost count. As I was there my 12v died. It was replaced and the software was rolled back because it was the day of the global Tesla outage and it was what the tech had stored on his mobile. (We were leaving to go hole the next day) all this kept happening all the way home.

we since got a new software update which solved a few of the issues...but not the slow to boot.
 
If I wasn’t under warranty I’d just pay for the daughterboard tomorrow. (On an older car I really can’t justify MCU2) I’ve just never heard them say that they can remotely see that the cache is full and that they can wipe it. My concern is that it sounds to me like wiping cookies which just fill back up quickly as you use it. I just don’t want them to get me to the point just outside warranty so that I am then left needing to replace it.

all of this started on a road trip last month to Arizona. As we were arriving at our first super the car went into a cycle of repeated reboots (while driving) arrived at the super and it didn’t want to charge. I hard rebooted and we were on our way again. Remote said everything looked good so we continued. I don’t know how many times we had to reboot after that over the two weeks we were gone as I lost count. As I was there my 12v died. It was replaced and the software was rolled back because it was the day of the global Tesla outage and it was what the tech had stored on his mobile. (We were leaving to go hole the next day) all this kept happening all the way home.

we since got a new software update which solved a few of the issues...but not the slow to boot.

A couple months ago I set up an appointment for slow laggy MCU that I reboot often to keep running smoothly (which I reported as frequent reboots). I specifically asked for EMMC board replacement in the service request. I got text saying that they remotely took a look and said that while the memory use is high, it’s not enough for EMMC replacement unde warranty. They said that they would clean my cache (sounds dirty) and I should see improvement, which I didn’t. So yeah they can remotely diagnose your EMMC chip and remotely wipe your cache and say that it’s fixed.
 
It’s all intertwined. Clearing the caches, including the nav map tiles, trip odometers, nav destination history, etc. will indeed help, but the root of the eMMC failure is a finite number of reads and writes, which kills the part, vs it actually getting filled up. The larger 64Gb part by default increases the life as you’re spreading out the reads/writes over a much larger array of addresses.
I’ll have more info tomorrow, but the key is to log/time stamp every time the MCU went black and required reboot to come back or spontaneously rebooted itself. That will help them review the logs.
 
Just responded to the main “Tegra Daughter Board” thread. Just got my car back after getting the board swapped and everything is working just fine. Not enough time behind the wheel yet to provide feedback on whether things overall are snappier, etc.
OP, since you’re in warranty until May, I would definitely make note of every time you get a black screen/every time you had to reboot manually. Will build your case for getting the board replaced under warranty.