Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

This site may earn commission on affiliate links.
For example see this and this. A non-trivial amount of die space will be used for decoding and microcode sequencing on modern CPUs.

Note that the diagram you cited is not showing the cost of the 'microcode layer' - no such "layer" exists anymore: most instructions are directly decoded and the large decoder areas your diagram is showing are multiple units of very complex x86 decoders, little of which die space is used for the microcode functionality itself.

Note that there's no direct relationship between "microcode" and "uOPs" ("micro-ops"), although the naming is confusingly similar and microcode sequences are (presumably) coded directly in uOPs.

In modern x86 CPUs the (indeed update-able) microcode is an exception mechanism that implements legacies, certain system functionality and bugfixes, such as interrupt handling or system calls - although the very latest micro-architectures accelerate even the system call instructions largely directly, without involving the microcode. Over 99.9% of the instructions that actually execute on an x86 chip on typical workloads never go through the microcode and are hard-coded.

"Microcode" is a very specific term in this context that means the small programs written in uOps, and which are invoked by the decoder only in special circumstances. Most of the die space in the image you cited is used up by the decoder that maps the instruction stream to uOps.

(I suppose we are largely in agreement, I just wanted to specify the context of what I wrote.)
 
You'll have to settle for stale info: ;)

carsonight 6 hours ago

"I am told that most of the batteries being produced by GF1 right now are SR. I'm also told the Grohmann machine is producing one every 45 seconds."

That's a very interesting leak if true:
  • ~80 SR battery packs per hour,
  • ~1,920 packs/day,
  • ~13,440 packs/week.
Presumably these are peak rates, but still very impressive. I suppose cell supply and Fremont assembly throughput has become the primary bottleneck, as hypothesized by @neroden for a long time.

Note that initially the Shanghai SR packs will be made in GF1, so they need this extra pack production capacity.
 
While being able to sleep (or work) in the car while it is driving itself unattended is the scientific and social Holy Grail of FSD, the actual economic FSD Holy Grail is a much more modest technological and legal milestone:
  • "The Tesla FSD car/taxi has to be able to self-drive to its next destination without any passengers on board at all, with high success rates, regulatory permission in at least one major market and low legal exposure."
On the actual trip where there are passengers in the car, one of the passengers is the designated driver with a driving license. (Just like the renting guest is responsible for an AirBnb home while using it.)

IMO this is the technological milestone that unlocks trillions in taxi, ride sharing and business fleet logistics revenue and this is the milestone enables the Tesla Network - and I agree with Elon that this could happen next year already. (!)

Seems to me however that there will be a high correlation between 'unsafe for a person to go to sleep in the car' and 'unsafe to other vehicles and pedestrians'. So I'm not clear how much easier it is to build a car to drive empty then to drive with passengers.
 
Now I heard about Tesla network, I want to try it and I call in Tesla Robotaxi. Do you expect me to jump in it, and take on liabilities of driving in to me unknown car, with unknown commands, without test drive, without introduction, just to save 10 or 20% off the ride fee vs. Uber? Not very realistic. And if I don't take on liabilities and oversight, what is the purpose of safety driver?

You mean, like you would be allowed to drive an unknown car that you rent from a car rental company or from Turo? They are only asking for your driver's license and your signature on a contract.

How it might work with the Tesla Network I described in more detail in a previous post:

So in the initial form (but I'm really just guessing here), I'd imagine something like this:
  • Tesla Network could self-insure and the customer EULA could act as a binding contract to make the driver liable for damages that they cause. They'd display it prominently, there might even be a short voice message before the driver can operate the car, and voice recognition would require the driver to answer "yes, I agree", and maybe sign their name on the screen or something. They'd be required to have a valid driver's license, and they have to be legally able to drive at that moment. (True for 90%+ Americans.)
  • Tesla Network might also require binding arbitration process before having to sue a customer for material damages. This usually makes it much cheaper and more reliable to recover small claims (which most traffic accidents really are).
  • "Always On Sentry Mode" recordings would offer robust evidence that regular insurance companies almost never have - so even during a major accident or major damage they'd have objective evidence from several cameras, plus telemetry and an internal camera recording. This is why I think they rolled out Sentry Mode so quickly.
  • Owners could be offered various convenience features, such as a new "loaner" car the moment there's a customer-induced accident with the car. I.e. it won't stop generating revenue just because a customer crashed the car.
So I think they can reduce their actual liabilities and legal costs immensely via these technological measures, plus they'd observe it during the prototype phase to see how much of an actual problem it is. The key observation is that video and audio evidence is both a powerful deterrent against abuse, and excellent evidence in any dispute resolution process.

Tesla also being a car company gives them unparalleled degrees of freedom to make this both profitable and convenient.

In addition to that I also outlined a mechanism for drivers who are uncertain whether they can drive a Tesla to go through a test drive in essence, before they sign up as drivers for the Tesla Network: this could involve a Tesla Network employee sitting in the car for the test drive and signing off on the driver after it.

I.e. depending on how much of a problem it is for inexperienced people to drive a rented Tesla in practice Tesla has a wide scale of options to balance between sign-up convenience and actual risks of inexperienced drivers.
 
  • Like
Reactions: UncaNed
Seems to me however that there will be a high correlation between 'unsafe for a person to go to sleep in the car' and 'unsafe to other vehicles and pedestrians'. So I'm not clear how much easier it is to build a car to drive empty then to drive with passengers.

Correlation yes, but also a significant risk and liability reduction factor:

There's a couple of fundamental differences when an FSD car is driving without passengers:
  • About half of the liabilities are occupant health damage costs in any traffic accident, especially if the Tesla car is at fault.
  • When driving between destinations without customers, an empty FSD car can be driving super conservatively: yield, keep to the speed limit, factor in weather conditions, etc. etc. Since the violence of traffic accidents scales with energy, and energy scales with the square of speed, just dropping the speed to the speed limit has a significant effect on insurance costs and liabilities. The FSD car will also not be annoying customers by driving in a timid, easily bullied fashion.
  • When driving between destinations without customers, an empty FSD car can drive the statistically safest route - i.e. it can do easy highway driving instead of that shortcut over complex urban terrain. It can use the Waymo tactics of avoiding risky left turns that LIDAR doesn't handle well, it can avoid residential zones when possible, etc.
  • If the FSD car is uncertain about any traffic situation it can just safely slow down, pull over if possible safely and generally ask for help - this is not a practical option for a 'taxi driver' FSD. 'Remote safety drivers' are possible technologically and could help a 'stuck' FSD car. The final fallback would be for a Tesla Network technician to get fetch the car in person.
All of these factors accumulate into significant differences both in terms of how many 9's it has to be able to achieve, in terms of regulatory approval, and in terms of liabilities and insurance costs/risks.

The introduction of level 5 FSD is basically an exercise in risk management, and all these factors reduce risk.
 
  • Like
Reactions: Richt and tschmidty
You mean, like you would be allowed to drive an unknown car that you rent from a car rental company or from Turo? They are only asking for your driver's license and your signature on a contract.

How it might work with the Tesla Network I described in more detail in a previous post:



In addition to that I also outlined a mechanism for drivers who are uncertain whether they can drive a Tesla to go through a test drive in essence, before they sign up as drivers for the Tesla Network: this could involve a Tesla Network employee sitting in the car for the test drive and signing off on the driver after it.

I.e. depending on how much of a problem it is for inexperienced people to drive a rented Tesla in practice Tesla has a wide scale of options to balance between sign-up convenience and actual risks of inexperienced drivers.

It's an idea. However, it seems like by the time they had any sort of sizable user base set up that way, Tesla would have gotten umpteen million miles of validation from owners to show the level of safety, thus rendering the backup driver moot.
 
  • Like
Reactions: tschmidty
Let's consider scenario: I have license, am urbanite, have some previous experience, but I don't drive, or drive/drove some old Toyotas.

Now I heard about Tesla network, I want to try it and I call in Tesla Robotaxi. Do you expect me to jump in it, and take on liabilities of driving in to me unknown car, with unknown commands, without test drive, without introduction, just to save 10 or 20% off the ride fee vs. Uber? Not very realistic. And if I don't take on liabilities and oversight, what is the purpose of safety driver?

Or do you want to onboard and qualify me through a short training program? How quickly do you think would Tesla network spread with such hefty enrolment process?

No, you need proper full FSD. Before full FSD, while preparing rollout, there could, and should be a period of mandatory safety driver presence, but they'll have to be paid, Uber-like contractors; this function can't be crowd-sourced.

What if you think of the "robotaxi" fleet less as an uber ride, and more as a 'rent a car that comes to you'? It's a fricking car ffs, if you have a license you should be capable of driving it. I can certainly see this as a stepping stone to robotaxis. From a regulatory and liability stand point I think the idea is interesting.
 
GF1 is mostly making SR battery packs, and is doing so at a rate of more than 1 per minute when the machine is running.

Yes, I thought that's what it means but it makes zero sense.


Why are more SR packs being produced than they can construct the cars? This will saturate their storage facility in like 2 days.

Or are there other uses such as for power walls?
 
B.C. passes law to increase sales of zero emission vehicles.

https://www.google.ca/amp/s/www.cbc.ca/amp/1.5155274

10 percent by 2025? They’ll be over 7 percent by the end of this year. Demand is already outstripping supply. We put in an order and 2000 deposit for a Leaf SL Eplus that we won’t see till 2020. If everybody ramps up (Hyundai, Kia, Nissan and the big boy on the block Tesla) I can see us easily reaching that goal early. Model Y and Tesla pickup will take us a long way to the 2040 100 percent law.
 
Smells a bit funny to me. On the linked page it is also stated:
"
Due to the license agreement change with our data vendor, Short Interest related data is no longer available on GuruFocus website."


And the same site mentions at a different page that institutional ownership is close to 77% (i'm not the expert, but the latest 13F had institutional ownership at 57%). And guru focus has insider ownership for TSLA at 0.33% (not sure if Elon Musk is considered an insider?).
Tesla Institutional Ownership (TSLA)


So I noticed these things as well... and went to Nasdaq for a 56.7% number

Tesla, Inc. (TSLA) Ownership Summary

Is it possible that there are completely fake financial sites dedicated to TSLA FUD?
 
  • Like
Reactions: neroden and dc_h
Yes, I thought that's what it means but it makes zero sense.


Why are more SR packs being produced than they can construct the cars? This will saturate their storage facility in like 2 days.

Or are there other uses such as for power walls?

When it is running.
The machine is built for the speed they need in the future when they are not limited by cells.
3 US & GF3, Y, maybe pickup.
 
What if you think of the "robotaxi" fleet less as an uber ride, and more as a 'rent a car that comes to you'? It's a fricking car ffs, if you have a license you should be capable of driving it. I can certainly see this as a stepping stone to robotaxis. From a regulatory and liability stand point I think the idea is interesting.

That's exactly how I see it as well. Was just about to post that

I would rather send my car off to regular customers, in an AirBnB fashion than send it to strangers all day long... It is more likely that I would send my personal car off to be rented by a regular customer, (car shared) for a half day or so... than to have the car drive the general public around for a few dollars a ride.

I can see joining a group of people, for example, my daughters water polo team families, buying a car to transport the girls to practice and games. (Buying a car... or timesharing the car...)

Yep, This "RoboTaxi" idea is hard to understand, most transitional (er... revolutionary) change is...
 
  • Like
Reactions: Mike Ambler
So I noticed these things as well... and went to Nasdaq for a 56.7% number

Tesla, Inc. (TSLA) Ownership Summary

Is it possible that there are completely fake financial sites dedicated to TSLA FUD?

I’m reasonably sure I’ve seen this gurufocus site pushing out really obvious FUD before. Like the kind of stuff you find on TSLAQ.