Alright Tesla -- switch out the crippled unit and get me my FSD.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD may require a hardware upgrade...
- Thread starter boonedocks
- Start date
-
- Tags
- Model S
Alright Tesla -- switch out the crippled unit and get me my FSD.
Waiting for more hard specs, but it seems they seriously increased the performance to achieve a given functionality, looks they realized they needed even more power than they thought. Now take into account that the current AP2 ships with 1/4 of the horsepower that Nvidia said in 2016 would require for full FSD (2x dual Drive PX2's). What does that mean terms of time? Looks like no hardware available to upgrade your AP2 to real FSD until 2019.
Tesla seems to become more bold with their exaggerations, not less.
...and until the software is implemented and works reliably (not a dumb demo), the compute power required is likely to be adjusted upwards again after 2018's hardware generation.
I wonder if Tesla tried to recognize revenue on the ~$5k per car over AP1 functionality.
Oh yeah... if people think this generation of hardware is anywhere near capable of FSD, they are dreaming. As we start to have workable, real-world software that is potentially capable of FSD, we are going to start to realize the actual computing power needed to achieve it with NTSB level of reliability. Sure, you can make an FSD car with this new Nvidia drive unit, and it may fail 1 out of 1000 times. Even a 1 in 10,000 or 1 in 100,000 failure rate is WAYYYYYYYY too unreliable for 1 million vehicles on the road, much less 10 million or more.
Tesla hasn't been able to get the automatic wipers working with the AP2 cars yet. Not only is it possible more processing power will be needed, Tesla may need to make changes to the sensor suite to operate in FSD mode safely. And additional hardware (V2V/V2I?) might be needed to get FSD regulatory approval.
Still wonder about audio sensors - so that the FSD software can hear warning signals - such as train whistles, emergency vehicles, police whistles - and respond, like human drivers can even if they can't see where the sound is coming from.
Still wonder about audio sensors - so that the FSD software can hear warning signals - such as train whistles, emergency vehicles, police whistles - and respond, like human drivers can even if they can't see where the sound is coming from.
It won't be this. it will be proprietary once they find out what is really needed.
Why didn't they just called it super EAP+ and not FSD.
It's truly remarkable how well Tesla has mastered the over promise under deliver mantra.
It's truly remarkable how well Tesla has mastered the over promise under deliver mantra.
Last edited:
3Victoria
Active Member
It only needs to be better than the average driver to save lives. Why do you think NVidia knows how much compute power is necessary ... if they do then they already have programmed. Computer sellers will always inflate what you need - its more profitable.
philgrocks
Member
Tesla did define. See attached.
Comparing with the SAE spec would indicate L5 or at least L4. SAE Levels of Driving Automation | Center for Internet and Society
The disclaimer is about when and not if.
Attachments
I very much dislike this much used statement. Have you seen the average driver? Average drivers are terrible. The current incarnation of AP2 is more like a teenage driver wearing a blindfold. I am much, much safer than your average driver. I am far above average, therefore likely far safer than any incarnation of AP or FSD. Why would I abdicate my superior driving skills to something less? Seems silly and dangerous.
I already have a system in my Tesla far superior to anything Tesla will have for years to come. It's called my eyes, ears, and my brain. I can see and anticipate situations that no Tesla technology is equipped to handle. I can bring my car to a stop in situations that Autopilot would never be able to handle. I can recognize hazardous objects on the backs of trucks that I need to avoid - Autopilot can't do that. I can take proactive measures to change lanes when I can see a dangerous driver ahead. And a million other things Autopilot can only dream about.
Why would I want to use Autopilot and put myself into the hands of a LESS SAFE and less capable river than myself? Just so I can check my email?
The disclaimer cites the timeframe is "highly dependent on local regulatory approval." I get that. My question is when will Tesla actually submit for said approval? And for those people that leased cars with the FSD option, I'd be concerned that they'll see FSD before the lease ends. It likely won't be a regulatory issue. It just won't be fully baked, like AP is now.
I do think that Tesla will owe some people a refund. I think Musk felt this was an acceptable risk, and their attorneys have thought through the implications. A refund doesn't affect P&L, just cash.
Wouldn't it be great if no refund was needed?
So based on the above I assume you have not purchased the autopilot or full self-driving options and thus have no reason to even be on this thread?
Title should have been:
FSD Will Require a Hardware Upgrade...
and none will get it, you'll have to buy a new car.
FSD Will Require a Hardware Upgrade...
and none will get it, you'll have to buy a new car.
dhanson865
Well-Known Member
It only has to be better than the average driver because there are hundreds of millions of drivers (possibly over a Billion). Half of those are worse than the average and that still leaves hundreds of millions of drivers that should switch to AP asap. And that is still more cars than Tesla can sell in any reasonable time frame.
Once AP cars become common, new systems only have to be safer than the old system + the remaining human drivers which will raise the bar (assuming the worst human drivers switch first)
So why should you use it? You shouldn't if you don't want to. But you aren't the only driver on the road.
Well put. If Tesla (or whoever) saturates the "current below average human driver" market with FSD vehicles that are "current average human driver" equivalent, then they've raised the average by 50%. So it's "better".
That said, I -- like Elon -- think the "trust hurdle" is much higher for regulatory and social buy-in.
One more thing: for anyone who considers himself/herself a "better than average" driver, consider investing in a "worse than average" driver you know by (a) gifting him/her a HPDE class [some come with provided cars] and/or (b) gifting him/her an FSD upgrade for his/her vehicle rather than your own.
Similar threads
