FSD on MX Better than FSD on MS. Why??

[Wingsy]

Model 3 in April 2019. Sold the M3 and got a MS Plaid in August 2021. Got a MX LR in March 2022.

Just got back from a trip from NC to FL & back and I can tell ya, after 1400 miles in the MX the FSD on the MX is WAY better than FSD in my MS. I've had only ONE phantom braking event in all that time, whereas in my S it pumps the brakes every few miles... sometimes every few feet. (And yes, I've done a couple of camera calibrations.)

Now the question is, why?

 

[Mardak]

I've noticed FSD Beta generally performing better on the Model Y than the Model 3 except in very narrow situations, e.g., lateral traffic calming. My guess is the higher position of the main cameras on the X/Y might perceive the world better than vehicles with the camera nearly a foot lower. Or maybe there's more Model Ys in general and training data might be biased?

 

[islandbayy]

Model X Should have the best performing FSD and Autopilot of the 4 vehicle lineup, Mostly due to the vehicle being higher, better visibility.

 

[clydeiii]

I never experience PB on M3 with FSD beta, so…no one knows.

 

[Wingsy]

I've noticed FSD Beta generally performing better on the Model Y than the Model 3 except in very narrow situations, e.g., lateral traffic calming. My guess is the higher position of the main cameras on the X/Y might perceive the world better than vehicles with the camera nearly a foot lower. Or maybe there's more Model Ys in general and training data might be biased?

You might be on it. The height above ground for the X forward cameras are 7" higher than on the S and the side cameras are 5" higher.

I wish there were 2 cameras on each corner of the vehicle and one in the windshield and 1 out the back. If nothing else it would give us a good cross traffic view when backing out of a space. Also triangulation to distant objects would be more precise.

 

[Dewg]

I would love to get two identical cars to do testing with. Two Model Ys, approx same year, both with the same software version. Then test FSD on the same route. There has to be some hardware issue, because I'd bet both cars would have different experiences, especially with PBs.

 

[Wingsy]

I would love to get two identical cars to do testing with. Two Model Ys, approx same year, both with the same software version. Then test FSD on the same route. There has to be some hardware issue, because I'd bet both cars would have different experiences, especially with PBs.

Personally what I think is... measuring distance to objects, especially at long range, is so iffy that even the slightest variation in reflectivity of an object can result in a wide variation in range, that the system just can't get it right every time. The signal is so very close to the noise. I can drive on the same road within minutes of each drive and braking may or may not occur at the same location. Then there are the gross errors, like attempting to go around a stopped school bus while unloading children, literally slamming on the brakes when there is absolutely nothing in its path, and more. I think we're years away from having a stress-free drive. I'll keep testing, but only when I'm alone and in light to zero traffic.

 

[EVRider-FL]

Model 3 in April 2019. Sold the M3 and got a MS Plaid in August 2021. Got a MX LR in March 2022.

Just got back from a trip from NC to FL & back and I can tell ya, after 1400 miles in the MX the FSD on the MX is WAY better than FSD in my MS. I've had only ONE phantom braking event in all that time, whereas in my S it pumps the brakes every few miles... sometimes every few feet. (And yes, I've done a couple of camera calibrations.)

Now the question is, why?

Which firmware versions are you comparing? That makes a difference too.

 

[clydeiii]

also OP is the route you’re taking the same? So many variables.

 

[Reddy Kilowatt]

My results are just the opposite.

The PB on the X is so bad its rarely used. My S never has issues, except on coming traffic when it slows down. We have not recalibrated the cameras on either car.