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

Let the hacking begin... (Model S parts on the bench)

This site may earn commission on affiliate links.
I have an MCU1 on the bench. Powers up fine and the touch screen works. I can't get the IC to display. The IC has power, and the wake line is high. I can confirm the LEDs are lit inside the IC (red for 12v, and the 2 green ones for 5v and 3.3v). I've also got the 4-pin ethernet cable connected (the original Tesla cable) between the IC and MCU1. Any else I may be missing? MCU1 is not rooted and is running v11 software (2022.8.10.5). I have nothing else connected to MCU1 or the IC. I haven't tried the brake On Sw line (it's unconnected), but since the MCU1 display is on, assume IC should be too. Has 10 service alerts, but expected with nothing connected. None of the alerts seem to be IC-related (brakes, charging, BMS, etc.)
Did you use the IC from the same car as the MCU?
 
Thanks all for your suggestions.

I bought MCU1 & IC (Gen-2 2015) as a set from one 2015 car. Still haven't figured out getting IC to display. Using power, CAN chassis 6, Wake line, and Ethernet. Tripple checked all my wiring. IC is powering up and since the IC's green LEDs light, I know the wake line is high.

Before connecting it up, I did try powering up the IC with the wake line at 12v. I also pressed the mystery button which might have been a mistake. It's only accessible when the back is off. It might do something like erasing the flash memory which would be bad. I'm just too curious!

I also have an MCU2-era 2017 wiring harness. The only thing I'm using from it is the Fakra Ethernet cable between the MCU1 and IC. I doubt the wiring changed, but I'll try and ohm it out to be sure the wiring is still correct using the MCU1 to Gen-2 IC connection.

It's also possible the IC is just defective. Both displays have lots of leakages, but that will not stop them from working.

Interesting that MCU1 can get an LTE connection without an antenna attached (one bar). I don't like that and will add a 50-ohm terminator to hopefully kill the signal.
 
Thanks all for your suggestions.

I bought MCU1 & IC (Gen-2 2015) as a set from one 2015 car. Still haven't figured out getting IC to display. Using power, CAN chassis 6, Wake line, and Ethernet. Tripple checked all my wiring. IC is powering up and since the IC's green LEDs light, I know the wake line is high.

Before connecting it up, I did try powering up the IC with the wake line at 12v. I also pressed the mystery button which might have been a mistake. It's only accessible when the back is off. It might do something like erasing the flash memory which would be bad. I'm just too curious!

I also have an MCU2-era 2017 wiring harness. The only thing I'm using from it is the Fakra Ethernet cable between the MCU1 and IC. I doubt the wiring changed, but I'll try and ohm it out to be sure the wiring is still correct using the MCU1 to Gen-2 IC connection.

It's also possible the IC is just defective. Both displays have lots of leakages, but that will not stop them from working.

Interesting that MCU1 can get an LTE connection without an antenna attached (one bar). I don't like that and will add a 50-ohm terminator to hopefully kill the signal.
the "mystery button" is just a reset button, thankfully, so you're good on that front.

You must be pretty close to a tower to be getting LTE on a bare MCU. If you're worried about it, you could disconnect the LTE board while you're tinkering. (probably already too late though, its likely it has already received the fastcharge kill signal and will need to be reenabled for DC charging to work with that gateway.

Did you wire up your CAN signals and wake line according to the documentation I posted above? Bench · Lunars/tesla Wiki
If so, you probably need to send the wake command to IC to get it to display anything. I don't know exactly what the wake command is, it's somewhere out there. Though, it will probably wake on its own if installed into a vehicle along with the CID.
 
Interesting that MCU1 can get an LTE connection without an antenna attached (one bar). I don't like that and will add a 50-ohm terminator to hopefully kill the signal.
Pretty sure i heard this from someone else as well that there's 3rd antenna in MCU not just mirrors..
Also, quiet possible Tesla purposely did it to have access even if mirrors are disconnected.
After all, our cellphones have internal antennas too, not hard to add one to MCU..

You must be pretty close to a tower to be getting LTE on a bare MCU. If you're worried about it, you could disconnect the LTE board while you're tinkering. (probably already too late though, its likely it has already received the fastcharge kill signal and will need to be reenabled for DC charging to work with that gateway.
😂
 
  • Disagree
Reactions: NevG
Good to know about the reset button! It was an unknown when I wrote my IC article: Instrument Cluster – TeslaTap

I'm about 1/2 mile from the cell tower. I've minimized keeping the MCU1 powered up very long (< 5 minutes or so). I'm waiting on coax connectors to terminate the cellular properly. If that fails, I'll go in and pull the card. This is just for my entertainment and is not going to be put into a car so not that worried about DC charging disablement, but don't really want Tesla to see it appear (likely too late).

On the docs, confirmed the wiring with the Lunar's bench and the schematics too. All agree with each other and what I've done. I did confirm the hardware wake line must be at +12v for the internal +5v and +3.3v power to become active and it is active as wired. I expect there is a further display command either via CAN or Ethernet.

In earlier testing with the IC alone, I was able to get some data via a Nmap scan over Ethernet - so I'm fairly sure the IC's Nvidia processor powers up and is functioning.
 
Pretty sure i heard this from someone else as well that there's 3rd antenna in MCU not just mirrors..
Also, quiet possible Tesla purposely did it to have access even if mirrors are disconnected.
After all, our cellphones have internal antennas too, not hard to add one to MCU..


😂
Yes it has, If you power the MCU on the bench and connect a metal wire to the correct place, you will get a stronger signal
 
  • Informative
Reactions: brainhouston
I picked up a used MCU2 without a screen on the cheap and am considering setting it up on the bench before I attempt to upgrade my MCU1. Anyone have extra connectors they'd like to sell me? I'm assuming to build a bench setup I'll need....

MINIMUM
a main touchscreen and ribbon cables
a good 12v power supply
power supply connector

MAYBE
instrument cluster screen and cable w connectors

OPTIONAL
bluetooth connector and cable
wifi connector and cable

FOR FUN
has anyone looked into upgrading the MCU2 eMMC? It is 64G right, like the factory upgraded MCU1? Basically, getting someone with a rework station to desolder and resolder new eMMC chip(s)? I am curious as to which replacement chips would work, 128G? 256G?

DIFFICULT
I'll be rooting around (pun intended) the forum to see about finding someone to root the unit and help configure it before I install it. Might need to root my 2022.8.10.11 MCU1 to pull the config file. Not sure what firmware was last on the MCU2 but at least 2020.40.102.2.

Also, I assume that if I do the MCU swap myself I might want to do the following at the same time as I believe thy require config changes:
  • XM FM Tuner addition (my standard audio car has no current XM and likely no antenna and antenna cabling)
  • CCS upgrade
I've followed this thread since inception but am rereading it now to refresh.

By the way, what are we calling the new iteration of MCU3/Z that has the GPU on the same PCB, MCU3.1? MCU3.5? MCUZ+?
 
Last edited:
I picked up a used (S/X) MCU2 without a screen on the cheap...
Lets get it out of the frame. Tools needed? Torx T10 driver. 6mm socket. Deep 7mm socket.

First unscrew the 12 Torx bits circled in red. All the same. Then the 2 Torx bits circled in green. Those 2 are shorter.

Image1a.jpeg


Lift off the CPU fan unit, lean it aside. Use the 6mm socket to remove the three standoffs circled in purple. Slide away the bar held by the last two Torx screws in the previous step.

Image2a.jpeg


Flip the frame over. Use a deep 7mm socket to remove the four nuts circled in the blue.

Image3a.jpeg


Now where are those memory chips?

Image4a.jpeg
 
Last edited:
Last edited:
Process of elimination...

TITAN LE940B6-NA (daughtercard) - multi-band LTE with 3G/2G fallback
blox NEO-M8L-04B-00 D9702587274 2203 0500 11 - GPS? Accelerometer?
NXP TJA1043 60000052 TnD24521 - CANBus
MAXIM MAX9260GCB/V KOXZ 2124 - Deserializer
(INTEL) i 12 2143AYP WGI210CS - Ethernet ICs DEVICE CON
ANALOGIX ANX1122I 2034 NA429AB - low power DisplayPort 1.2 to 24bpp dual-/single-channel LVDS translator
MARVELL 88EA6321-TFJ2 PAG0050.08 2124 A0P TW - automotive gigabit Ethernet switch
TOSHIBA 358748IG 2105HGL LP8258A TAIWAN - Camera Interface?
SPC5668GVMG 0N61C CTAH2101D - Unknown
FTDI 2051-C DHYAQG1 FT4232HQ - USB 2.0 Slave to Quad Channel UART / Serial Converter
(LG) RBHP-B216C LC 2202 2B004 H/W VER B (FCC ID YZP-RBHP-B216C) - T(V4.2) + WLAN(802.11a/b/g/n/ac) 2x2 MIMO Module

Maybe the memory is under the CPU heatsink. I really don't want to take the heatsink off...

Image.jpeg

Image.jpeg

Image.jpeg
 
Yea, memory will be close to the CPU, and if CPU is under the heatsink, so will the memory. Additionally, memory also should also have a heat sink in automotive, but I think Tesla solves this problem with cabin overheat protection. Here, for example, is a Porsche OTA module, notice how all the memory, emmc, and CPU/SoC chips sink to the heat-sink. .
1678002900448.png


The heat-sink is the entire bottom of the enclosure for the ECU:
1678002992755.png


PS> yea, I've moved on to reverse engineering Porsche, very interesting to see how other manufacturers do things as compared to Tesla.
 
  • Informative
Reactions: Droschke
Yea, memory will be close to the CPU, and if CPU is under the heatsink, so will the memory. Additionally, memory also should also have a heat sink in automotive, but I think Tesla solves this problem with cabin overheat protection. Here, for example, is a Porsche OTA module, notice how all the memory, emmc, and CPU/SoC chips sink to the heat-sink. .
View attachment 914003

The heat-sink is the entire bottom of the enclosure for the ECU:
View attachment 914004

PS> yea, I've moved on to reverse engineering Porsche, very interesting to see how other manufacturers do things as compared to Tesla.
It's counter intuitive, but non-volatile memory actually degrades less during writes when they are hot, so actively cooling them actually is not a good thing. They do like to be stored cold, but a heat sink doesn't really help things in this regard that much for storage (given cooling system would be off when car is sleeping, so eventually everything reaches ambient anyways).
https://www.eeweb.com/industrial-temperature-and-nand-flash-in-ssd-products/

As such it is not recommended to heat sink the NAND flash portion (video mentions NAND flash running at 25C will have half the life of flash running at 40C), even though heat sinking the controller might make sense if it's a performance bottleneck (which they say usually it is not):

Cabin overheat AFAIK, had nothing to do with the eMMC. Instead it was to address the screens of the original Model S, which was not rated for as high a temp rating as automotive screens (so it fails after a while, the fluid/bubbling problem that affects so many Model S).

The eMMC was failing because they only had 8GB and were writing tons of useless logs to it. By the time they tried to address it, it was too late for most of them (thus the recall). There's little indication temperature had anything to do with it.
 
Last edited:
It's counter intuitive, but non-volatile memory actually degrades less during writes when they are hot, so actively cooling them actually is not a good thing. They do like to be stored cold, but a heat sink doesn't really help things in this regard that much for storage (given cooling system would be off when car is sleeping, so eventually everything reaches ambient anyways).
https://www.eeweb.com/industrial-temperature-and-nand-flash-in-ssd-products/

As such it is not recommended to heat sink the NAND flash portion (video mentions NAND flash running at 25C will have half the life of flash running at 40C), even though heat sinking the controller might make sense if it's a performance bottleneck (which they say usually it is not):

Cabin overheat AFAIK, had nothing to do with the eMMC. Instead it was to address the screens of the original Model S, which was not rated for as high a temp rating as automotive screens (so it fails after a while, the fluid/bubbling problem that affects so many Model S).

The eMMC was failing because they only had 8GB and were writing tons of useless logs to it. By the time they tried to address it, it was too late for most of them (thus the recall). There's little indication temperature had anything to do with it.
Heatsink is not really considered active cooling. Shared heat sink heats up the EMMC to the same temperature as the rest of the chips (SoC, RAM, couple of others), so the EMMC chips warm up faster than in the Tesla but also don't reach the same highs. Additionally, I'm only starting to dig into this (just picked up my Taycan 3 weeks ago, then drove it home for 9 days to get to know it), but I don't think Porsche is writing much to the OTA module EMMC at runtime, other than during OTA.
 
I picked up a used MCU2 without a screen on the cheap and am considering setting it up on the bench before I attempt to upgrade my MCU1. Anyone have extra connectors they'd like to sell me? I'm assuming to build a bench setup I'll need....

MINIMUM
a main touchscreen and ribbon cables
a good 12v power supply
power supply connector

MAYBE
instrument cluster screen and cable w connectors

OPTIONAL
bluetooth connector and cable
wifi connector and cable

FOR FUN
has anyone looked into upgrading the MCU2 eMMC? It is 64G right, like the factory upgraded MCU1? Basically, getting someone with a rework station to desolder and resolder new eMMC chip(s)? I am curious as to which replacement chips would work, 128G? 256G?

DIFFICULT
I'll be rooting around (pun intended) the forum to see about finding someone to root the unit and help configure it before I install it. Might need to root my 2022.8.10.11 MCU1 to pull the config file. Not sure what firmware was last on the MCU2 but at least 2020.40.102.2.

Also, I assume that if I do the MCU swap myself I might want to do the following at the same time as I believe thy require config changes:
  • XM FM Tuner addition (my standard audio car has no current XM and likely no antenna and antenna cabling)
  • CCS upgrade
I've followed this thread since inception but am rereading it now to refresh.

By the way, what are we calling the new iteration of MCU3/Z that has the GPU on the same PCB, MCU3.1? MCU3.5? MCUZ+?
I think you need to remove heatsink to find eMMC. But i'm not sure if it is worth upgrading it. The only reason why would you upgrade would be increase in eMMC performance. For example, tesla probably uses v5.0 or even older eMMC versions, but just upgrading chip to 5.1 will should improve performance significantly.

Now CPU is another story, as I've posted in another post, the fastest CPU you can solder without any issues is probably Pentium J4205, which is lower base frequency, but significantly higher boost frequency:
Atom CPU in MCU2 - - Product Specifications | Intel
Pentium J4205 - - Product Specifications | Intel
Also note that Atom does support ECC Memory and Pentium doesn't, which shouldn't be a problem as ECC ram can work with non-ecc controller. you can also check how much RAM there is actually soldered. Both CPUs support up to 8GB, which can be an easy ram upgrade.


Another thing which is easy to replace is microSD card, you can get something high performance with High Endurance, which will avoid wearing it out after some time. microSd is used for maps only, so my guess you'll not see much performance increase as it will preload maps, but worth testing.