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

GROUP BUY - MCU-1 to MCU-2 Upgrade, Testing interest

This site may earn commission on affiliate links.
Anyone have contact with Rich rebuilds. He opened the electrified garage and posted a video about the emmc failure. The guy in that video seems like he would be perfect for this

That's @Ingineer here on TMC. He could definitely do it, but he had to move his shop earlier this year and hasn't been heard from in quite a while, I think he's just overwhelmed with moving his entire operation and business with very little warning. @wk057 is another member who could do it, but I don't think he's interested at this time (I could be wrong). He said it's doable, but going to be really expensive. I think a few other folks here have the knowledge to do it, but not necessarily the shop or parts or hacking tools to do it on a large scale. I'm sure Rich Rebuilds could do the installation parts, but not the hacking parts or creating the required harnesses or pinout modifications.
 
There are ways to read the MCU1 emmc without (de)soldering it, no Toolkit needed either. There are few ppl that can do this. PM me if you want more info.

Reading (or writing) the MCU2 emmc is another story which I do not know much about (yet) but from what I've heard this can be done by ISP. This is actually on my to-do list, I only need to find time.

I actually have a MCU1 car, and a spare MCU2 on a bench setup, and the tools to read/write the emmc's and write gateway config to MCU2 etc.
So in theory I could test the TS proposed experiment on an actual car, but I really don't want to be first and possible have my car banned or anything...

The emmc chips differ by the way, MCU1 is 8GB Hynix H26M42001FMR or similar, MCU2 is 64GB Micron MTFC64GAJAECE-AIT.
Hi, I would like to know more info about this and actually has anyone of you wondered if there is any way to know how your Emmc is getting to its last cycle? Is there any way to know how long will your MCU last till it dies?thx for your opinion and help!
 
Here's a question from a computer geek who knows nothing whatsoever about the electronics in my AP2 Model S. I hope it's not so naive as to cause hearty laughter.
When Elon tweeted "Working on it" recently in response to questions about upgrading MCU1 to MCU2, didn't that imply that Tesla would offer this upgrade at SCs? If so, wouldn't they have to read the old eMMC and write the data into the new MCU2's eMMC, just as they do when replacing a failed MCU1?
OR, is this proposed research project aimed at finding a DIY approach to the upgrade?
 
... wouldn't they have to read the old eMMC and write the data into the new MCU2's eMMC, just as they do when replacing a failed MCU1?...
Tesla doesn't have to read the eMMC as Tesla put all the data there and can recreate it at any time. To DIY you need to read the eMMC as you don't have access to Tesla servers.
 

This is verging on criminal IMO. Tesla = tech business, right? Then any hardware designer knows about the limited emmc writes deal. This is perfect planned / designed in obselessence. Tesla have designed in a controllable way with clear statistical basis of being able to break your car after a certain number of hours of use. As the video states, there are so many easy and far better ways of storing log data - especially data that sounds like it doesn't serve much purpose in normal use of the car.

When you see such info as this, along with all Tesla's other 'sharp practices' it becomes VERY hard to trust the company at all. All those owners dreaming of MCU upgrades from 1 to 2 (one assumes at least partly in search of better reliability) are dreaming for nothing if the emmc is used in the same way.

Just an idea. I wonder, if the data written to the emmc really is just data logs, would the car still work if you disabled 'write' to the emmc so that it's contents never changed?
 
One snag would be that it would be pretty easy for a moderatety competent manufacturer (not sure if Tesla qualifies) to make a write - read back to check that the WE was still connected to the chip. :-(
 
The emmc is used for way more than logs. It stores all non volatile data like map tile cache, spotify cache, profile settings, nav history, etc. Making the emmc read only is not what you would want.
Currently the lesser evil solution to improve emmc life is to root the car and disable logging or have it log to ram. Still, with all the updates and cache writing to the emmc it will fail someday.

Hi, I would like to know more info about this and actually has anyone of you wondered if there is any way to know how your Emmc is getting to its last cycle? Is there any way to know how long will your MCU last till it dies?thx for your opinion and help!
Without having access to the logs (how ironic) it is hard to determine if the emmc is getting corrupted. You would notice it when it often resets out of the blue, is very slow to start, screen stays black, weird bugs noone else reports start to happen etc, etc.
When you do have access to the logs you will see lots of read errors like these:

Code:
cid kernel: [ 1814.117998] SQUASHFS error: zlib decompression failed, data probably corrupt
cid kernel: [ 1814.125083] SQUASHFS error: squashfs_read_data failed to read block 0x299ccd1
cid kernel: [ 1814.133038] SQUASHFS error: Unable to read data cache entry [299ccd1]
cid kernel: [ 1814.139567] SQUASHFS error: Unable to read page, block 299ccd1, size d6f9
 
Last edited:
The emmc is used for way more than logs. It stores all non volatile data like map tile cache, spotify cache, profile settings, nav history, etc.
That begs the question of are different user use case likely to cause one user's system to fail sooner than others. i.e.
a) people that road trip a lot and use the NAV/route_planner which would cause the map cache to be redone a lot more than a user staying in the same area and just 'reading' the cached map over and over
b) users constantly using Spotify vs phones music or FM radio
c) etc

Would the bug Tesla had to get the traffic data in millisecond frequency instead of like 120 second data cause that emmc to be written to (ie. is that where they stored it?
MCU1 performance improvement & browser improvement. Simple trick by Thomas The Tesla Tuner..
Ok, I'm inside (Mr Elon, please do not send an assassin ;) - The whole problem is not a lack of memory. It is caused by refresh function for traffic info, with passes processing to the main routine. It seems that somebody interpreted value of delay of a refresh to be in seconds (120, = 2 minutes), but it is in milliseconds. It eats 30% of spare MCU1 time. It tries to re-render it 6-7 times per second, with no new information (refresh period of traffic info via communication channel is set correctly).
My assumption - two differently skilled programmers did that.
Please, Tesla, correct this. Just add three zeros (120 -> 120000), recompile and send to us. And all will be ok.
Thomas
PS: Loading Tegra3 with this one passive cooler over 60% of load causes throttling. Who was a developer? Even ventilation holes, developed by engineers are covered from the bottom by shelf under the screen. No air circulation then.
 
Last edited:
The emmc is used for way more than logs. It stores all non volatile data like map tile cache, spotify cache, profile settings, nav history, etc. Making the emmc read only is not what you would want.
Currently the lesser evil solution to improve emmc life is to root the car and disable logging or have it log to ram. Still, with all the updates and cache writing to the emmc it will fail someday.


Without having access to the logs (how ironic) it is hard to determine if the emmc is getting corrupted. You would notice it when it often resets out of the blue, is very slow to start, screen stays black, weird bugs noone else reports start to happen etc, etc.
When you do have access to the logs you will see lots of read errors like these:

Code:
cid kernel: [ 1814.117998] SQUASHFS error: zlib decompression failed, data probably corrupt
cid kernel: [ 1814.125083] SQUASHFS error: squashfs_read_data failed to read block 0x299ccd1
cid kernel: [ 1814.133038] SQUASHFS error: Unable to read data cache entry [299ccd1]
cid kernel: [ 1814.139567] SQUASHFS error: Unable to read page, block 299ccd1, size d6f9

I guessed there would be way more than just logs. TBH, if it came down to the crunch, I would give up a lot of functionality compared with paying 2 - 5000 every 3 or 4 years, and if turning off writing to the eMMC achieved that, I would consider it. It might even be a way of locking your car out from any unwanted 'updates'. I'm sure there would be issues, but what's the point of exchanging oil services (oh, except I hear that some models need 3 yearly lube flush), exhaust systems etc of ICE cars with electronics that is pretty much designed to fail every 3-4 years? And I bet that eMMC expected service life degrades significantly with temperature, and we know that there are cooling issues.

Your second point is also very valid (about the somewhat random nature of faults that could show up due to the eMMC errors. It's almost like the duff SD flash memory cards that keep hitting the market where you don't know you have an issue until you start trying to access duff / non-existent areas of memory.) Rebooting of one sort or another may mask the underlying MCU problem to start with, but eventually the memory performance degrades to a point that it is impossible for the system to function. These issues are huge and fundamental. A more mature business that wasn't so desparate to show its viability would have to bite the bullet and replace all effected hardware. Tesla is not only out on a limb financially, but somehow it will have to try and escape from these technical obligations.
 
Hmm. Care to share more info.
Like what? Im not going to shame the nameless more then this, Im sure they have their reasons BUT I do think a Twitter bot that has tokens for all of our twitter accounts to ask every 12-18 hours per person might actually be the easiest way... and asking at a shareholders meeting.

By the way, FOR THOSE DONATING TO THE CAUSE... No money will be asked for until an actual plan is in place so you know exactly what you are buying into. (R&D, Twitter harassment etc ) lol
 
Last edited:
When I read the title of this post, I took it as; 'how many people would it take before Tesla would pay attention to seriously offering an upgrade service?'
Of course it would be expensive, no one seems to dispute that. BUT, how much would the market bear?
I'd love to see a survey about the consumers vs dollars.
If this were a TESLA provided service, included a warranty, listed the (current) features your vehicle would receive what would be the high end of your tolerance to pay? I think for me it would be upwards of $10k.
So if there were 999 more people like me, is $10M still something Tesla would sneeze at?

The last time I tried to create a survey on this forum, I made quite an abomination. I still have the nervous tick from desperately trying to pull it down.. So.... If someone else cares to do it... Just sayin' :)
 
I think for me it would be upwards of $10k.
So if there were 999 more people like me, is $10M still something Tesla would sneeze at?

Tesla probably are counting on getting at least half that back when people upgrade their rides, which they won't do as soon if they keep existing cars updated. (sustainable bubble popping)
Still need to keep making and selling new cars. All those used models that have to find homes just mess things up for selling nice shiny new ones.

All companies that make stuff need to carefully distance themselves from 'the old model' as fast as possible.
 
True. And I agree that the corporate strategy is likely "sell cars at any cost to customer loyalty". It's the 'new way' for companies today.
But, I'm sure I am not alone in that I have purchased my one and only Tesla. I have participated as an early adopter - with no one really to blame but myself. And Tesla has missed promise after promise of the great autonomy of self driving. Their software management is hobby level at best. Customer Service is in tatters, The steering wheel nag continues to negate any sense of self driving enjoyment.
I fully understand distancing from inferior products for newer models. But I don't understand distancing from customers and potential evangelists. What Tesla *should* say is... 'Thank you for helping Tesla get off the ground.. Instead of spending money to upgrade your hardware, can we offer you $10K off on a "shiny new" model S?'
I wonder how long before my S75D becomes a collector's item. If ever...
 
  • Like
Reactions: StarLog and am_dmd
I could chip in to a research project.

I actually have a MCU1 car, and a spare MCU2 on a bench setup, and the tools to read/write the emmc's and write gateway config to MCU2 etc.
So in theory I could test the TS proposed experiment on an actual car, but I really don't want to be first and possible have my car banned or anything...

If that's only thing holding you back, I would have this perfect 85-beater with 540 tkm on it to try this =).
 
  • Like
Reactions: LuckyLuke