Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
Well yeah, it'd have to be given we know HW3 has been out of compute, even using both nodes as effectively a single node (and taking the performance hit doing that entails), for years now.

Eventually they'll need to be able to run the full stack in a single node after all for redundancy if they want any hope of L4 or better so 3x would be the BARE minimum better to have any chance at all of that goal... and it's entirely possible even 5x won't be enough.

As I've pointed out before, until someone actually solves this, nobody knows how much compute is "enough"
Yep, the infamous project scope creep is alive and well with FSD. Elon continues to push his internal switch to intellectual dishonesty mode.
 
HW4 is a downgrade, using more processing power…
Downgrade to HW3? Do you just mean that the driving experience on HW4 currently isn’t as good as HW3 because the software is still in development? That is frequently the case with Tesla. HW2 wasn’t as good as HW1 initially.

And every next revision takes more power. When Tesla releases a HWx+1 that uses less power than HWx, you’ll know their software is complete. It’s going to be a while.
…Might as well just use a Xeon simulation predicting the road ahead…
I assume you’re kidding about that. Anyone done the math on an appropriate GPU benchmark running on a new CPU, coat aside?
 
Last edited:
  • Like
Reactions: petit_bateau
Downgrade to HW3? Do you just mean that the driving experience on HW4 currently isn’t as good as HW3 because the software is still in development? That is frequently the case with Tesla. HW2 wasn’t as good as HW1 initially.

And every next revision takes more power. When Tesla releases a HWx+1 that uses less power than HWx, you’ll know their software is complete. It’s going to be a while.

I assume you’re kidding about that. Anyone done the math on an appropriate GPU benchmark running on a new CPU, coat aside?
It's a fine balance, if Ai one day should become aware it will making human errors just like it's creators. Heat is collision and should be included in a benchmark when measuring the ceiling. NVIDIA and the likes have mislead a lot of their benchmarks for financial gain. DVR once ran 4 cameras faster than a half decent gaming PC trying to run a webcam. Like I mentioned it's a fine balance try to master stability. Under clocking would be the only solution for EV. Now let begin the maths and check the results as we try figure out how bees prevent collisions flying in a swam.
CPU Technology
 
Last edited:
  • Informative
Reactions: Nrkl
The 4nm TSMC process (Factory) in Az is on hold. Hope that wasn't supposed to source Tesla. The stated reason is shortage of labor. (Once again, I'm reminded of how difficult it is to manufacture in USA.)


I get this funny feeling that AI will turn into a war with chips at the center, used for leverage or constraint on the competition. Although I'm glad Tesla didn't try to build their own Fabs, and glad to see multiple sources (Samsung/TSMC), it is still a risk factor none the less. It also makes me question the real reason for the TSMC delays. (Yes, some was safety related, skills, pay differences..) but this was all known before they entered our Desert State.

I smell something fishy here. But at least there's plenty of sand here for everyone to play!
 
You clean it...same with your current reverse camera or the front parking camera on almost every single car outside of Tesla.
I’ve driven five winters, so far, with the current hardware on my May 2018 build TM3.

The rear camera does get effected, but it is not a critical path item when (even simple) TACC is used.

When the forward radar was still being employed, any coating of any freezing precipitation cancelled all cruise control operations in this car.

Salt spray on (for example) HWY 7 between Peterborough ON and Ottawa ON will coat the front end of the car in a matter of minutes.

And it does so in such a manner that if the auto wipers don’t keep up with it and then they do a sweep, the resulting streaking cancels all TACC (ADAS) systems until it is rectified (with many uses of the washer fluid button).

Your “you clean it” and “every other car” answer doesn’t change the issue: the ADAS systems in these cars don’t work when covered in salt spray.

If every other car on the road in eastern Ontario between mid December and mid March suffers a loss of simple cruise control operability because of salt spray on a front bumper camera, I would have expected to hear something about it by now.
 
I’ve driven five winters, so far, with the current hardware on my May 2018 build TM3.

The rear camera does get effected, but it is not a critical path item when (even simple) TACC is used.

When the forward radar was still being employed, any coating of any freezing precipitation cancelled all cruise control operations in this car.

Salt spray on (for example) HWY 7 between Peterborough ON and Ottawa ON will coat the front end of the car in a matter of minutes.

And it does so in such a manner that if the auto wipers don’t keep up with it and then they do a sweep, the resulting streaking cancels all TACC (ADAS) systems until it is rectified (with many uses of the washer fluid button).

Your “you clean it” and “every other car” answer doesn’t change the issue: the ADAS systems in these cars don’t work when covered in salt spray.

If every other car on the road in eastern Ontario between mid December and mid March suffers a loss of simple cruise control operability because of salt spray on a front bumper camera, I would have expected to hear something about it by now.
it will still use the front windshield camera that has wipers.

I think you are overthinking this.
 
  • Helpful
Reactions: navguy12