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

[MP3Mike]

On-chip redundancies are good for improving yield due to testable production defects, trying to detect runtime failures on-chip and then hope you have the right things being duplicated is more difficult. What if you get multiple errors on the chip, and both copies of the NN are affected? How do you decide which to use?

Actually, there is somewhat this problem with only 2 (and not 3 or some other odd number) of chips, too.

If they're using dual chips, it will be to protect against hard failure of a chip dying / locking up / overheating / etc, not so much against random cosmic ray bit flipping or electromigration causing one chip to produce wrong output (as the output might still look valid, and then you don't know which chip to use).

This is more redundancy along the lines of having dual steering motors or redundant braking boosters, etc, rather than 3+ computers having their output voted on as you would for spacecraft for example.

Elon answered this today:

Two, independent system-on-chip architecture, with each SoC having two NN accelerators that can perform simultaneous health-check calculations to protect against a soft error

So they have two chips that each have redundant cores. So they have 4 duplicate running processes to compare against on the FSD computer.

Of course he said the load for the current software, on the FSD computer, is 5%, or "10% with full fail-over redundancy". (Or 80% on the HW2.5 computer.)

 

[mrdoubleb]

Lots of TSLAQ hate in the replies.

It is absolutely ridiculous how they now parrot "where is the 8k?". LOL.

When did Tesla enter this agreement? February 25. Geez I wonder why Tesla filed 3 CT orders with the SEC that day granting confidentiality to some material information...

 

[madodel]

I am a big supporter of Tesla. I own one and bought stock. About 25% of my portfolio is in TSLA. I bought it last year at about $290, believing in the imminent rise of the stock due to increased production. "Short burn of the century"
It didn't materialize and since last quarters deliveries and especially production numbers I am getting cold feet.

My stock are under water and I am at loss what to do. My stock are for a study plan for my kidsI don't need to cash it for another year as my oldest daughter is only going to college next year. Shall I take my loss now before the stock tanks further or is there still hope they will recover. Some things that make me nervous:

- how come that M3 is still not sustainable at 5,000 per week - let alone 6,7, or even 8,000 a week?

- what is it with all these pictures with dust covered cars?

- I see people on YouTube ordering a car only to be called by Tesla within hours that they have a deal for a new 2018 M3. So there is a big stock?

- what is indeed the story with the solar roof? Absolutely nothing is happening there

- suddenly, after years, there is a big push in Software. I don't believe in coincidences. So this in combination with the bad sales and production figures makes me suspicious. Are they trying to change the focus?

-all these restless strategies around pricing of cars and software seems very nervous and not reassuring at all

I could probably think of more.

Am I alone this? Just wondering what other people think and are considering doing

As others have said TSLA is not a good short term buy. I've been in TSLA for about 6 years. I'm very long but expect major value increase in the next 5 years. Anyone that says they know for certain what will happen with TSLA in the next 6 - 12 months is not someone you should listen to seriously. Instead you have to listen to what may happen and gauge how much you can risk. If you have learned anything here it is that TSLA is very manipulated by shorts, bears and enemies who want it to fold. And apparently the Federal regulators who should be cleaning up Wall Street are in bed with the crooks.

You need to make your own decision as to what to do, but I'd advise following PapaFox's excellent Daily TSLA Trading Charts analysis here at Papafox's Daily TSLA Trading Charts for trying to figure out what may be going on. But even he is sometimes surprised at a daily result. No FUD, no hype, just data about what happened and what may happen. And never invest with funds you will need for something important soon. I'm a complete idiot and have gambled on some things people saw as sure things and lost money but I can afford the losses and write it off to a learning experience. Good luck with whatever you decide.

 

[BioSehnsucht]

Elon answered this today:



So they have two chips that each have redundant cores. So they have 4 duplicate running processes to compare against on the FSD computer.

Of course he said the load for the current software is 5%, or "10% with full fail-over redundancy".

That's interesting. I guess you can choose the 3 that agree (1 whole SoC plus half of another) if there's any disagreement, automatically kick out anything that detectably errored, and I suppose the chances of two different pairs of identical results (i.e. 2 vs 2) is statistically improbable ...

That's of course assuming the dual NNs per SoC are indeed the same NN, and not simply just using dual accelerators running different portions of the NN, which is a possibility if the SoC does not have the NNs on die but are SoC + two NN dies linked externally or on package. Using more smaller dies rather than fewer larger ones can be cheaper, depending on various factors - it will improve yield, but depending on the workload it may complicate things in terms of fully utilizing them.

In any case, I expect the results to be sufficiently safe against defects / runtime errors that the chances of a failure occurring in which we are able to after the fact determine the cause and that having an odd number (3+) instead would have been better, to be zero.

 

[JRP3]

Sorry you disagree, but here is my source.
There is no exact calculation for how much electricity it takes to drill, transport and refine a gallon of gasoline, but the accepted amount is around 8 kWh.

Debunked many times. It's 8kwh of energy, most of it in the form of heat, less than 1 kwh is actually electricity used.

 

[Off Shore]

@KarenRei I was expecting a Edgar Alan Poe reverence, but the picture works.


Contrary to some members thoughts, Tesla will not be out of the crosshairs in 2020 if/when Trump is replaced. There are currently 18 Democrats in the race currently. Presidential campaigns are very expensive and there are millions of Petro dollars available to "help".

Of course. Now turn yourself in at the closest retraining center for doubting your master's... err government's benevolent actions.

For that reason please check to see if your favorite candidate is in this group.

 

[RiverBrick]

Edit: Did some research and it looks like the Bolt gets closer to 2.9 ZEVs/each, so I updated the data above.[

Unfortunately, CARB uses the range on the UDDS cycle when calculating how many ZEV credits are earned for sale of a vehicle. It is true that the formula 0.01 x range + 0.5 would yield 2.88 if they used EPA range, but the sale of a single Bolt actually earns the cap of four credits or something near it. Someone can look up the actual value for the Bolt, but the UDDS range is generally 30%-50% higher than the EPA range, so you end up with roughly 0.01 x 238 x 1.4 + 0.5 = ~3.83

 

[KarenRei]

Maxwell offer extended to 15 May.

https://t.co/Q7dQskcrGa?amp=1

 

[Thekiwi]

To be honest, I think this is bending the narrative in an superbullish way.

If you think of it in production numbers, it means FCA will pay the same amount of more/less one week M3 production?

The FCA payment is 100% margin, should not be compared to car production with ~20% margins.

Using a minimum $200 million annual figure (the lowest “hundreds” of millions could be) - that would equal the equivalent of a $500 profit increase per car based on 400,000 units.

With the amount likely to be much higher than $200 million, I think its probably safe to say the FCA deal add something like the equivalent of a ~200 basis point increase in Tesla auto gross margin (eg 20% -> 22%) in 2019/2020

 

[Unpilot]

Nice after hours bump up

 

[KarenRei]

green on Twitter

"Well, I just thought Raven was mostly it, but nope, that was just scratching the top of it."

I so hope this is a "just one more thing" at the autonomy investor event or on the Q1 investor call

 

[CorneliusXX]

I’m glad that the SEC requested for the monthly reports (among increasing fines and other things) during the hearing to the Judge. It gives EM attorneys something to offer that was specifically requested of the Judge and does nothing to impede on Musk and Tesla (minus the time to quickly generate a monthly update). It’s a gimme IMO.

I'm not so sure this is the best idea. At some point Elon could tweet something he thinks is innocuous and the stock takes off. Then in addition to the agreement, you also have a complicit board. At least with the current setup only Elon can really get raked over the coals if he tweets incorrectly.

 

[MP3Mike]

Unfortunately, CARB uses the range on the UDDS cycle when calculating how many ZEV credits are earned for sale of a vehicle. It is true that the formula 0.01 x range + 0.5 would yield 2.88 if they used EPA range, but the sale of a single Bolt actually earns the cap of four credits or something near it. Someone can look up the actual value for the Bolt, but the UDDS range is generally 30%-50% higher than the EPA range, so you end up with roughly 0.01 x 238 x 1.4 + 0.5 = ~3.83

Great, so I was closer before the edit.

 

[KarenRei]

I'm not so sure this is the best idea. At some point Elon could tweet something he thinks is innocuous and the stock takes off. Then in addition to the agreement, you also have a complicit board. At least with the current setup only Elon can really get raked over the coals if he tweets incorrectly.

The metrics need to be clearly defined. If Musk states new info on Twitter that will affect profitability by more than $X, then the presumption is that it's material, and if either of those criteria fail, then it's presumed not. Something like that.

 

[Johnny Ma]

Whoops. Meant to include this:

My track record on predictions is terrible anyway, so I'm just gonna go ahead and call it and not care at all when I'm wrong: two more things are the S/X refresh.

 

[CorneliusXX]

Layoffs and store closures will be reducing expenditures, which was kind of the whole point of them.



Really?

So you mean we will see 70-100k deliveries nevermore?

 

[avoigt]

Wood doesn't say 'in 2019', but other than that that tweet could serve as a litmus test for how well informed she is on Tesla.

I mean, the actual amount paid by FCA to Tesla should become known later and then we can see how close the tweet came. Btw, 500M $ is about 440M €.

FC would need to pay the fine to the EU in 2019 therefore why should Tesla allow pooling with them pay later say in 2020?

That would be poorly negotiate out of a position of strength.

Although we do not know when exactly Tesla can account for it and Wood does not know either rungless she has seen the contract we should assume its in 2019. As of the amount, IMHO even if FC wind 100 Mio by doing the deal its good for them. Why should Tesla go far below the fine with likely many companies asking for that service.

In 2020 its gets more severe for automakers with higher fines if they do not have enough BEVs sold. That will allow Tesla to negotiate even better.... a cash machine...

 

[KarenRei]

So you mean we will see 70-100k deliveries nevermore?

On the morrow he will leave me, as my Hopes have flown before.”

Then the bird said “Nevermore.”

 

[Nocturnal]

FC would need to pay the fine to the EU in 2019 therefore why should Tesla allow pooling with them pay later say in 2020?

That would be poorly negotiate out of a position of strength.

Although we do not know when exactly Tesla can account for it and Wood does not know either rungless she has seen the contract we should assume its in 2019. As of the amount, IMHO even if FC wind 100 Mio by doing the deal its good for them. Why should Tesla go far below the fine with likely many companies asking for that service.

In 2020 its gets more severe for automakers with higher fines if they do not have enough BEVs sold. That will allow Tesla to negotiate even better.... a cash machine...

How does this work? Do they allocate an estimated MPG of the Tesla cars being "shared" and average that out against the Fiat cars?

The arguments against Tesla doing this make no sense to me. Keeping Tesla cash solvent and/or allowing them to grow faster through investment is far better for the goal of EV/low emission adoption than sticking it to Fiat and letting them suffer for their own failings.

 

[keydiver]

I cant see this posted:
Cathie Wood on Twitter
Does $500million sound right?

Wow, nothing but trolls responding to her Tweet. One the one hand, I'm glad to see that it isn't just Elon that gets bullied. But, on the other hand, it pulls my opinion of humanity even lower. Twitter really is a cesspool.