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

[Drumheller]

72 Watts for computer... It's the Heater solution, recyle that for passenger toe warming.

For comparison: high end gaming GPUs can use up to 300Watts for the card while under gaming load. Perhaps more, I don't know the specs on all of the current gaming cards.

 

[scaesare]

Yea, so my question is what are the obvious improvements? And why, if 3.0 is sufficient ?

You can never be too rich have enough TOPS or too skinny consume little enough power.

 

[T-Will]

Will there be a Q&A?

 

[Cherry Wine]

I'm always impressed with the level of detail that Elon knows. Most CEOs do not get that deep into the weeds.

It is impressive, but at the same time comes across to me as a bit micro-managerial.

 

[Nocturnal]

I enjoy these sorts of things. When you work with idiots and you are a person of above average intelligence you begin to forget that there are people out there that are far more intelligent than you are. It's a nice reminder that I'm actually just an idiot.

 

[sbradbury]

I wish Musk had kept quiet about the power efficiency of a competitor chip. He insisted it had a larger power envelope when it has about half according to what I could find:

NVIDIA DRIVE Xavier, World’s Most Powerful SoC, Brings Dramatic New AI Capabilities | NVIDIA Blog

And when you need 7 of those to bring it to the same compute performance what happens to the power?

 

[gene]

EM body language is not very comfortable.

I think he is holding back to allow the others to speak.

 

[Cherry Wine]

Substantive increase in stock price.

 

[Nocturnal]

Yea, so my question is what are the obvious improvements? And why, if 3.0 is sufficient ?

Lower power usage, future proofing etc.?

I think he is holding back to allow the others to speak.

I'm obviously no where near his level, but when I'm in a situation where I'm letting underlings (I don't care for that word) present something it's very hard for me to sit still and not jump in. Probably the same thing.

 

[T-Will]

Can someone ask if the FSD 2.0 hardware can be retrofitted (for a cost)?

 

[SOULPEDL]

SP up to $263.75, it's happening!!!

Juice after market close maybe.

 

[Drumheller]

Yea, so my question is what are the obvious improvements? And why, if 3.0 is sufficient ?

The obvious ones I can think of:
- move to a smaller manufacturing process
- Reduce power consumption

And I'm sure they are learning things already for the software testing that they can add to the next design to make it more powerful and more efficient.

 

[ZeApelido]

Don't Osborne the chip Elon!

 

[scaesare]

EM body language is not very comfortable.

I read it more as "You guys really need to understand this."

 

[Silent Ludicrosy]

Bullish AF. Anyone think VW is designing their own chip?

 

[anon]

The NVIDIA Pegasus Self-Driving kit actually has around 130 TOPS of INT8, according to their website, but it does require 4 chips in total, I think likely at around 200W total power consumption. So Tesla's chip is clearly better, but not quite to the extent that they tried to make out. Maybe they are just out of date and haven't bothered to keep up on NVIDIA's latest stuff.

 

[humbaba]

And when you need 7 of those to bring it to the same compute performance what happens to the power?

True, but it is half of the 7x -- just let your competitors be. He didn't need to say that.

 

[Doggydogworld]

Kick-ass chip!

What surprised me is that they created not just a NN accelerator, but integrated it with a full SoC, video processing units and a number of beefy 64-bit ARM cores.

I.e. it's a full replacement for all Nvidia chips AFAICS.

Maybe one of their old chips. He said they only licensed a modest GPU as they (correctly) expected the need for post-processing to diminish as the NN improved.

 

[JRP3]

EM body language is not very comfortable.

Typcial

 

[dailo]

It is so awkward up there...