4 ms·

My guess is that the hardware engineer that originally designed the board picked a reasonable part for a traditional auto OEM. Then at some point software engin

My guess is that the hardware engineer that originally designed the board picked a reasonable part for a traditional auto OEM. Then at some point software engineering decided to log all the things. Nobody actually looked back at what the storage system longevity would be with this new lifecycle. It's a dumb mistake, but not exactly an uncommon one. I made the same mistake early on and I have contract engineers that propose inappropriate storage solutions to me still on projects. Some very insistently. I think a whole lot of people that work on the embedded side of engineering treat storage as a magic perfect box.

amluto 6y ago
The hardware folks picked fancy stuff that was not intended for automotive use. Look up the screen yellowing for an example — no manufacturer would sell a screen in the size Tesla wanted that was rated for hot car interior temperatures, so Tesla went with one that wasn’t.
- rektide 6y ago
  1.4 drive writes per day is a shit ton of a lot of writes to hit a drive with daily. this is a computational not an automotive concern, & a pretty frelling heinous oversight at that.
  - o-__-o 6y ago
    To be expected when not-a-car company makes cars :) (For instance GM has a strict list of test scenarios for their onstar integration)
    - maxerickson 6y ago
      They still screw up. I like this example because it's their infotainment screen failing prematurely: https://cadillacsociety.com/2019/10/28/second-lawsuit-filed-over-cadillac-cue-screen-issues/ https://cadillacsociety.com/2019/10/28/second-lawsuit-filed-...
      - o-__-o 6y ago
        Sure but that’s not a software issue, that’s poor material selection
  - amluto 6y ago
    I wonder if a “cycle” is actually a full drive write. If it is, then they have ~11GB/day, which is nuts, at least in the pre-dashcam models. But I can easily imagine that a cheap eMMC part has poor wear leveling and eats a “cycle” after much less than 8GB.
  - NathanielK 6y ago
    It's not, the issue was there is no wear leveling, so it's 1.4 writes per day. This is approximately 2 trips a day, 5 days a week. Reading between the lines I believe the car "eats into" the 3000-5000 cycles everytime it is cycled on.

amluto 6y ago

The hardware folks picked fancy stuff that was not intended for automotive use. Look up the screen yellowing for an example — no manufacturer would sell a screen in the size Tesla wanted that was rated for hot car interior temperatures, so Tesla went with one that wasn’t.

rektide 6y ago
1.4 drive writes per day is a shit ton of a lot of writes to hit a drive with daily. this is a computational not an automotive concern, & a pretty frelling heinous oversight at that.
- o-__-o 6y ago
  To be expected when not-a-car company makes cars :) (For instance GM has a strict list of test scenarios for their onstar integration)
  - maxerickson 6y ago
    They still screw up. I like this example because it's their infotainment screen failing prematurely: https://cadillacsociety.com/2019/10/28/second-lawsuit-filed-over-cadillac-cue-screen-issues/ https://cadillacsociety.com/2019/10/28/second-lawsuit-filed-...
    - o-__-o 6y ago
      Sure but that’s not a software issue, that’s poor material selection
- amluto 6y ago
  I wonder if a “cycle” is actually a full drive write. If it is, then they have ~11GB/day, which is nuts, at least in the pre-dashcam models. But I can easily imagine that a cheap eMMC part has poor wear leveling and eats a “cycle” after much less than 8GB.
- NathanielK 6y ago
  It's not, the issue was there is no wear leveling, so it's 1.4 writes per day. This is approximately 2 trips a day, 5 days a week. Reading between the lines I believe the car "eats into" the 3000-5000 cycles everytime it is cycled on.