It does not "fix", it logs less so it delays it. Every eMMC will die, just a matter of when.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Likely MCU Failure (MCU1 eMMC)
- Thread starter Jkaufm32
- Start date
It does not "fix", it logs less so it delays it. Every eMMC will die, just a matter of when.
amiral_sub
Member
Yes that's why I wrote "partially".
I did no understand by how many the fix divides the log writing
I did no understand by how many the fix divides the log writing
That's true for every flash part. This is why cars and other long-lived products should always use a slot-based MMC.
[* bangs head against wall *]
I had a blackout 2 weeks ago which would not reboot until the 6th try. It was during a new update install so Tesla blamed that. Just before that I had the scrambled video over the whole climate control popup. I could still make changes to the AC if I remembered where the buttons were, but it did not go away until after the update installed.
So I may want to stream a bunch now to push this one over the edge before my 4 year warranty expires in March 2021. It seems this problem might be accelerating due to the larger and more frequent software updates since AP2 in early 2017. Mine is a January 2017 build with 29K miles and had no log monitoring and energy savings always set. I stream only about 30% of the time, the rest is local radio which I assume does not write to the eMMC, but maybe it does.
So I may want to stream a bunch now to push this one over the edge before my 4 year warranty expires in March 2021. It seems this problem might be accelerating due to the larger and more frequent software updates since AP2 in early 2017. Mine is a January 2017 build with 29K miles and had no log monitoring and energy savings always set. I stream only about 30% of the time, the rest is local radio which I assume does not write to the eMMC, but maybe it does.
IndyToronto
Member
A few weeks ago, I did notice the slacker streaming audio would display a consistent corrupted button that was unreadable. A soft reset cleared it.
About every couple of days, when driving, I would notice the MCU icons would change LTE to 3G, and one bar and then the streaming player would stop and display an error. The geographic location is random and on roads where I know the phone works consistently with good reception.
Other than that, I have not had a black MCU ever.
If this is what is happening, I'm surprised to see this at 30k miles.
I design chips for a living. Many years ago when flash was still new, a customer in japan was using our chip with a tiny external flash memory that was designed to store user settings. An over zealous test engineer on their side included a stress test that would continuously write and delete large amounts of data non-stop. They started to report device failures within days that was at 100%. When we looked at their testing we explained that if they really wanted to know how many safe write cycles there are with the flash component, all they had to do was read the datasheet for that component. Writing hundreds of thousands (or millions) of times to any same flash area is guaranteed to make it fail.
In the industry today, noone does this anymore except at the beginning to establish that indeed the 100,000 or 1,000,000 write cycles are actually working. Along with the wear levelling that delays this failure to (dozens of years). I'm surprised Tesla with all their smart engineers did not realize that excessive write cycles is not appropriate use of flash memory.
How can I stress MCU1 on normal reasonable use so that it would fail within the warranty period so I can get a proper replacement from Tesla?
About every couple of days, when driving, I would notice the MCU icons would change LTE to 3G, and one bar and then the streaming player would stop and display an error. The geographic location is random and on roads where I know the phone works consistently with good reception.
Other than that, I have not had a black MCU ever.
If this is what is happening, I'm surprised to see this at 30k miles.
I design chips for a living. Many years ago when flash was still new, a customer in japan was using our chip with a tiny external flash memory that was designed to store user settings. An over zealous test engineer on their side included a stress test that would continuously write and delete large amounts of data non-stop. They started to report device failures within days that was at 100%. When we looked at their testing we explained that if they really wanted to know how many safe write cycles there are with the flash component, all they had to do was read the datasheet for that component. Writing hundreds of thousands (or millions) of times to any same flash area is guaranteed to make it fail.
In the industry today, noone does this anymore except at the beginning to establish that indeed the 100,000 or 1,000,000 write cycles are actually working. Along with the wear levelling that delays this failure to (dozens of years). I'm surprised Tesla with all their smart engineers did not realize that excessive write cycles is not appropriate use of flash memory.
How can I stress MCU1 on normal reasonable use so that it would fail within the warranty period so I can get a proper replacement from Tesla?
OK, I think I just misunderstood your post. I'm trying to keep the information here clear and detailed. I want to avoid the term "fix" since the update was pushed to many emmc that are pretty worn, and the life expectance for those emmc is extended, however most of those are severely worn and responding much slower to read and writes. They are likely witting acceptable limits. I did have a customer report their MCU was very unresponsive and slow to draw maps (timed in minutes) and did not have other failure signs. He did notice a huge improvement once upgraded chip was installed.
Streaming will wear more. If your already getting blackouts I would think your close to the edge of total failure... However if you have current updates, logging will be reduced, and that caused much more writing than streaming does... If it does not fail within your warranty I suggest you contacting someone to replace your eMMC as soon as warranty expires.
Your car is 2016, the mcu is often running even if your not driving. All that time the last 3 years your emmc has been busily writing, and writing, and writing. You are showing many of the failure symptoms. Look at the post above this. I just responded to a very similar condition to yours.
It is slot based, they just have a lot on that particular card. /s
But, it really shouldn't cost that much to replace. The daughter card should be no more than $200 in volume. And that's being verrrrrry conservative.
Decibelle
Member
Hi Tony,
I you change the chip in my car, will I have to update the car to a newer software version or can I still be on my current version?
I you change the chip in my car, will I have to update the car to a newer software version or can I still be on my current version?
Replacement would be exactly the same as you have now. Just on a bigger chip. Preventive upgrades should be super smooth and just like nothing happened. It's the recoveries that get crazy...
whitex
Well-Known Member
That price you gave is to upgrade your bus pass to a Tesla with MCU2. Upgrading from say 2017 Model S with MCU1 to 2020 Model S with MCU2 does not start at $80K.
whitex
Well-Known Member
Even with everything that is on that card, the replacement should be sub $200. You can buy similar modules today (newer Tegra, more memory) for $150. No, they are not compatible with Tesla, but the same type of parts just newer.What's a CPU or two among friends, eh?
That is exactly my plan. I do have the latest software with logging reduced. If it somehow lasts another 15 months without warranty replacement, then it will be in the mail to you that day. In the meantime, I just need to listen to more music and start logging with TeslaMate I guess
.
It truly is an issue - mine did just die (2015 April) - out of warranty by around 6 months.... the service center hinted there was some engineering firmware that might have possibly revived it to make the car drivable for some more time until spare parts could get here, but alas, it was stone dead. Lucky no pin to drive so I didn’t have to tow the car there, but could actually drive it (car can’t charge though .
I’m still happy as far as one can be right now, they seem to be able to fix it within a week which is ok - unfortunately mcu1 - lets see if refurb or not.
. I’m still happy as far as one can be right now, they seem to be able to fix it within a week which is ok - unfortunately mcu1 - lets see if refurb or not.
You're ok with paying 2500?It truly is an issue - mine did just die (2015 April) - out of warranty by around 6 months.... the service center hinted there was some engineering firmware that might have possibly revived it to make the car drivable for some more time until spare parts could get here, but alas, it was stone dead. Lucky no pin to drive so I didn’t have to tow the car there, but could actually drive it (car can’t charge though
I’m still happy as far as one can be right now, they seem to be able to fix it within a week which is ok - unfortunately mcu1 - lets see if refurb or not.
Pursuant to a backchannel conversation:
Rework requires a good bit of technique and skill. Maybe there are youtube videos.
My guy uses a JBC infrared soldering station.
You'll also need a stainless mask with the correct number of pads and pad-pitch, and solder balls of the right size (ask for a sample from Kester), and liquid flux of the right type. (I believe my guy uses NC-256)
For example...
Rework requires a good bit of technique and skill. Maybe there are youtube videos.
My guy uses a JBC infrared soldering station.
You'll also need a stainless mask with the correct number of pads and pad-pitch, and solder balls of the right size (ask for a sample from Kester), and liquid flux of the right type. (I believe my guy uses NC-256)
For example...
Hi all - I've browsed around this thread (not all 19 pages, however), looking for more details/instructions on what is required to address the eMMC issue. I gather that TonyT provides this service, but not clear on what is required to access the board(s) that need to be modified, what it costs, etc. Can someone point me in the right direction?
I have a 2014 S85, although no symptoms (yet).
Thanks!
I have a 2014 S85, although no symptoms (yet).
Thanks!
Similar threads
- Question