Roadfly Home | Features | Car Review Videos | Car Reviews | Cars For Sale | Used Car Parts Classifieds | Forum | Car Review Archives | Forum Archives Index


Allen Schiano
08-02-2002, 01:10 PM

Well I've been analysing Jason's data more as the months roll by. I've added the data for that last two months so that we're up to date. The URL attached shows a JPG of the Excell graphs. It would be nicer to let you manipulate those graphs in Excel but I don't have a clue how to do that via the web!

Anyway, the new data still follows the failure analysis I did earlier. You can really see the differentiation of the 'normal' engine failure mode from the failure mode experienced by the Oct-Nov-Dec 01 cars. Of course, all this gets more definitive as more cars fail - what a lousy thing to want!

Here's what I continue to see:

1) M3 engine can fail at anytime in their lives. Of course we only have data there up to 20K miles so far. But you can see in the cars before the suspect period - and the one 2002 build date failure - that so far we have about 1 car per 1000 miles 'box' failing for cars not in the suspect period.. I call this the 'normal' failure rate. Not being able to compare failure rates for bottom-end bearing failures for other cars, I can't tell if this is unusual (I have had such a failure on my wife's car in this way) or if the M3 engine is like others. These failures should also include driver error failures. But it seems for now that again about 1 car per 1000 mile block at this time.

2) For the months of Oct, Nov, and Dec 2001 there is definitely a different distribution - and therefore likely failure mode - for those cars. I count 27 cars for example failing during those three months AND failing between 4000 and 7000 miles - 3 'bins' worth.

In comparison to the 'normal' failure rate of 3 cars in those three bins, we have a factor of 9 in difference(!) in what could be interpreted as 'signal to noise', i.e., unusual to usual. That in scientific terms is a strong signal of a different mode of failure.

3) The peak failure rate is really peaked at 5 to 6K of mileage at failure time. One of the interesting new failures (#65) is from that period by has twice the 'average' mileage, 13,400. Hence it might either be a 'normal' failure even though the car is from the 'suspect' period' or it might be that cars are still failing for the 'suspect' reason but they are also happen later in the engine's life. Too early to tell. If there are more like it, i.e., 'Suspect' build date plus more miles, say above 10K, then we can expect the 'Suspect' failure mode to occur to many more (or all) cars built during the suspect time (Sorry StoneWalk!). The other option is that we see less and less 'Suspect'build date cars failing with advanced miles. So it was a 'birth defect' after all - you either got it and it killed you or you were spared.

4) Lastly, it does seem that cars built after Dec 2001 are much less suspect to failing than the suspect mode cars. They should fail like the cars before the suspect period if this theory holds. At one car per 1000 mile box AFTER they reach 6-10K miles, we should have a few failures after about 6-12 months (the same time since the first failures in 2001 cars) since their build date. So cars from Jan to March should have a few 'normal' failures by now. We have one Feb car already.

To show that those cars are not failing like the 'Suspect cars' we should NOT see a peak in the failure curve versus engine mileage like we currently see for the suspect cars.

In summary, it still strongly looks like something went wrong with cars built in mid Oct to mid Dec 01 (I use mid since there are far fewer Oct + Dec failures to Nov failures). It would be nice to hear this from BMW.

08-02-2002, 01:25 PM
not sure why unless they were not producing any 6 spds( NOT). The lack of any Feb01 failures with the first build for US( knock on wood) is interesting for those who believe first build cars have higher failure potential..mine is a Feb 01. Maybe more attention to detail for those first build since I have near zero issues with my LSB coupe.

Allen Schiano
08-02-2002, 01:28 PM
related to 'suspect' failures. Hence why the talk of mis-shifting has gone away - you can't mishift SMG, in principle (according to BMW :) ).

08-02-2002, 01:31 PM
I believe it's independent.

November was the first month that SMG was available in the US as an option. Pent up demand meant more than 75% of _all_ M3's in that month were ordered with SMG. Thus it's no suprise that 75% of the failed motors in November were also SMG.

e E36's S52 engine..
08-02-2002, 01:34 PM

08-02-2002, 01:38 PM
given the small number of available vins, it's hard to draw any conclusions, but I spent 5 min and looked at the distribution of vin number diffs.

Generally outside of the nov, the vins are separated by 150 to 350, while during Nov, the separation is around 100 sometimes less.
If one of the machines carving the part in question was not configured correctly, we "may" expect some periodic appearance of failures if the parts are produced about the same speed and mixed to the line as they come out.
If the parts are collected and used in a batch, we "may" expect some bunching of failures in vin diffs, perhaps shown in Nov.

Can't tell either way of course, and not sure if this kind of look makes sense or not, but we may see something interesting if more VIN's are available.

Jason, sorry to bother you with this, but for owners who do not want to reveal vin's to public yet you have on file, could you show us the distribution of differences in a way that does not reveal their real numbers?

08-02-2002, 01:40 PM
Maybe the crankshaft turning machine which was not properly set up was the one used to mate more SMGs than others. But all things being equal you would have expected some 6 spd representation in the grouping after the SMG became available. What is missing here is the exact breakdown of 6 spd to SMG during these specific production months though I still doubt that there were very few 6 spds produced in Nov and Dec 01. Interesting finding without good explaination. Thanks for the review.

Allen Schiano
08-02-2002, 01:54 PM

Old M3 Driver
08-02-2002, 02:02 PM
I haven't heard of any M Coupe/Roadster engine failure. They are essentially the "same" engine with the M C/R being slightly detuned and with different software. So, if the M3 is driven not to exceed 7400 rpm (or M C/R redline), will engine failure be less likely? I doubt any M3 owner never exceed 7400 rpm ;)

08-02-2002, 02:18 PM

Roadfly Home | Car Reviews | Forum Archives Index