The current M3 warranty for the drivetrain/electric motor for my 2018 M3 LR RWD is 8 years/120k miles.

How long do engineers think the drivetrain/electric motor will last? (Note: Edited to correct info on warranty).

How long do engineers think the drivetrain/electric motor will last? (Note: Edited to correct info on warranty).

0

## Comments

https://electrek.co/2018/10/15/tesla-drive-after-million-miles-test/

Say what you want about the car, but the customer communication and information accuracy has a lot of room for improvement.

+1

However, it’s worth mentioning that messaging has improved quite a bit in the past year.

whichever comes first, with minimum 70% retention of Battery capacity over the warranty period". If you purchased your vehicle after February 2019, the battery and drive unit warranty has ALWAYS been 100,000 miles, irrespective of what any salesman may have told you. Read the published literature.

Formula E engines are expected to last 2 seasons....

Not really comparable, but ....

Sorry for the literal answer... I found humor in it

@rehutton777 You missed the main point of my post. It wasn't about the length of the warranty coverage, it was about how long the engineers/experts think the drive unit/motor will last in comparison to the warranty.

——

I don't consider myself an engineer* or an expert, but my original answer was based off my experience across my Teslas. My Model 3 is approaching 70K miles now, and I believe that the mileage based warranty on Teslas is meaningless, the drivetrain is far more resilient than the warranty numbers suggest.

*I have a master's degree in electrical engineering, but I never used those skills professionally - life had something else in store for me

The curve is the probability of fault at a particular time. At T=0, when the product leaves the factory, the probability of error is relatively high due to latent manufacturing faults. This is why a lot of stuff that one buys has that 30-day warranty at the Big Box store; there's a lot of returns during that time when some nut-that-didn't-get-screwed-down right backs off and kills the hardware. Or a defect in the metal after it came out of the vat and was forged comes to bear, the piston rod breaks, and one needs a new motor. Early stuff.

As time goes on these failures drop off. One is then left with random failures, evenly distributed over time. This is usually figured as a Poisson distribution. The failure unit is in FITs: Failures in Ten to the Ninth Hours. A single resistor, for example, is usually figured to have a rating of 1 FIT. If one has a billion resistors lying around, then it's expected that, at room temperature, with the things operating, one resistor fails per hour.

So, say that one has a radio with a thousand resistors in it. That means that, just talking about the resistors in that radio, there's 1000 FITs. So, the failure rate would be 1000/1e9 = 1e-6 probability of failure per hour, on average, across all the radios made by this manufacturer. Or, looking at it another way, given that there's 8760 hours per year, the probability of this radio failing in a given year because of a resistor in it going bad is 8760*1e-6 = 8.76e-3, or about 0.87% per year. Not bad.

But, since we're talking about, say, a table radio, let's say that there's 10 ICs in it. FIT rates of IC can vary a lot; especially hot ones, and it's proportional to the number of I/O on the IC. Let's say these are middle-of-the road and there's each 100 FITs. That's 1000 FITs for the ICs, 1000 for the resistors, 1000 for the other passive components, 5000 for any moving parts, and then we got a vague back-of-the-envelope number for a given radio might be 8000 FITs. Probability of failure of said radio in a year would be 8760*8000/1e9 = 7%. And now you know why radios come with a one year warranty.

Mean Time Between Failures is simply 1e9 divided by the FIT rate. So, for our radio with 8000 FITs, the MTBF is 1e9/8000 = 125,000 hours, or about 14 years.

For those who are interested, FIT rates can be actually measured; there's something called the Arrhenius equation which relates failure rate to temperature. So one will take a large batch of components, crank the temperature to some huge number, then record the failures over a month or three. Then, once the failure rate is measured with this statistically significant number of parts, one can mathematically crank the temperature back down to some more sane number and get the FIT rate at 25C.

Where life gets interesting is when one is interested in building reliable hardware that has protection, so that if one thing fails, there's a backup, with a Mean Time To Repair once the original fault's detected. There's standards for telco gear, for example, that mandates 50 microminutes per year expected downtime. And you don't want to even think about the calculations done for NASA-certified long-haul satellites that don't have any repair possibilities; backups on top of backups on top of backups, with fail-safe all over.

Finally, at the end of this Poisson-distributed flat error rate, one runs into the end of the bathtub curve: Wearout, which is what it sounds like.

An example of an object that everybody knows about that's subject to the bathtub curve are incandescent light bulbs. Get a case of 1000 of them, and power 'em up. Likely that in the first few hours a few will die; then, they'll run for the length of time, roughly, stated on the box, with the rare but occasional random failure; then, since there are processes in play that cause the filaments to erode over time, more and more of them will begin to fail. For a bulb rated at, say, 1000 hours of operation, they'll all likely be dead (probability, remember) around 5000 hours.

But there are things that don't have a wearout mechanism. There are low-power transistors, for example, that, if one doesn't run the electrons around too fast, have no known wearout mechanisms. Start one up today and, assuming one can find a reliable power supply (ha!) it might be running 1000 years from now. Fun.

But if any of you were wondering why the Gen 3 computer in Model 3's have two redundant CPUs.. Now you know.