Autonomous Cars

[roon]

Windshield installers and random numbers aside... If the Tesla tech succeeds that would be proof of the effectiveness of relatively inexpensive machinery. Small optical cameras, sonars, and a PC. Mass production + time... I think we may see cheap options become available. Just my guess.

Interfacing with a vehicle's OEM computer may be a challenge, although Waymo and Uber use Chrysler vans and Volvo SUVs, respectively.

[TankMarvin]

And the power required to run the CPUs/GPUs used to drive the AI means it can really only be deployed in electric vehicles.

I know some of the early development cars carried 8-10 servers and probably had around 8-10kW sustained power draw

Tesla claim 72w for theirs.

https://cdn.vox-cdn.com/uploads/chorus_ ... _ai_15.jpg

Marketing slide.

That is just for a single chip, the associated hardware such as memory/storage, connectivity, etc will increase that into the 100s.
Then there will be multiple chips/packages; for both performing different tasks as possibly for resiliency/failover.

Production systems will be significantly less than the development vehicles, but still much higher than 72w.

[roon]

And the power required to run the CPUs/GPUs used to drive the AI means it can really only be deployed in electric vehicles.

I know some of the early development cars carried 8-10 servers and probably had around 8-10kW sustained power draw

Tesla claim 72w for theirs.

https://cdn.vox-cdn.com/uploads/chorus_ ... _ai_15.jpg

Marketing slide.

That is just for a single chip, the associated hardware such as memory/storage, connectivity, etc will increase that into the 100s.
Then there will be multiple chips/packages; for both performing different tasks as possibly for resiliency/failover.

Production systems will be significantly less than the development vehicles, but still much higher than 72w.

This is production. In the cars already. Samsung in Texas manufacture the chip. Memory less power than processor. Connectivy in what sense consumes power? "100s" is a number you made up. Processor is built as two concurrently running halves for redundancy--leading to their "shootable" claim. This is all freely available info released months ago. I cannot vouch for any of it other than to say: why would they start lying at this point in their history.

[TankMarvin]

Tesla claim 72w for theirs.

https://cdn.vox-cdn.com/uploads/chorus_ ... _ai_15.jpg

Marketing slide.

That is just for a single chip, the associated hardware such as memory/storage, connectivity, etc will increase that into the 100s.
Then there will be multiple chips/packages; for both performing different tasks as possibly for resiliency/failover.

Production systems will be significantly less than the development vehicles, but still much higher than 72w.

This is production. In the cars already. Samsung in Texas manufacture the chip. Memory less power than processor. Connectivy in what sense consumes power? "100s" is a number you made up. Processor is built as two concurrently running halves for redundancy--leading to their "shootable" claim. This is all freely available info released months ago. I cannot vouch for any of it other than to say: why would they start lying at this point in their history.

Processor = ~72W
8GB GDDR5 (for example) = ~20-25W. Given the size of the datasets and models the systems could be running multiple of this, so 20-25W or 40-50W or maybe even 80-100W just in RAM (or less RAM per server but more servers)
Solid state storage = ~2W per TB active power consumption. Given AVs are predicted to generate >10TB/day (but not store all of it) there would likely be muliple TBs for the current, working dataset, so say another 10W.
Networking/connectivity = ~10W - this would be internal connectivity - sharing data (100s Mb/s up to Gb/s) between systems in the car (such as acqusition of images from camera to processor or do another idenitical processor running in lockstep as part of a resilient system or to another system that performs a different function in the workflow) and also perdiocally syncing/sharing data with cloud/backend systems for model/training updates, analytics and accountability likely over a 5G antenna.
So that is somewhere between ~100-200W for a single computing unit (processor + memory/storge + connectivity)

Also, that is only one implementation. Just because one company claims a implementation with X power consumption doesn't mean there are other implementations with different features/performance that consume more.
My point was, in general, in the context of different technology providers, there will be popular, non-niche production grade systems on the market where the power requirements can and will be greater than that figure.


A single processor built as 2 concurrently running halves in not a resilient system - it might given 2 independantly calculated results for any given task to help detmine probabilities/weights when all components are functioning correctly, but there are single points of failure spacially (they are physically in the same place), electrically, thermally and logically.
A truely resilient system needs physically seprated components which do not so share any common infrastructure - therefore multiple computing units active in the vehicle (multiples of that 100-200W figure).

Why would a company in a capitalist market trying to sell a product in a rapidly expanding sector make any sort of claims about a product they are developing... Well...
I'm not saying their figures are wrong, just a little disingenuous as it does not paint the entire picture... and tech companies always "sell forward" that is part of demand generation - make the buying public think it's ubiquitous.

[roon]

How is a slide about processor power use which lists processor power use, disingenuous? You brought up "8kw-10kw" draw as your last point of reference. Just giving you a new one.

In terms of resiliancy/redundancy, they may need only meet human performance, or best it, ideally. The computer is in the dash and can survive and operate under high g loads. It's 10(?) eyes are distributed around the vehicle.

[subcritical71]

listen to the audio of the presrntstion and they went in to say 250w per mile. not sure how they came up with that though... maybe at highway speed(?)

These are Tesla designed units made specifically for the task. The CPU specs are quit impressive. HW3 is running at 5% load, where HW2.5 is at 80%.

[roon]

250w per mile

What's the time component?

If per hour that equates to total vehicle draw (i read 240-300 Wh/mi for Model 3 average use; 100 on hypermiling attempts)

Need to know peak vs avg draw of the FSD system, and what percentage of the baseline comparison figure constituted autopilot use.

[subcritical71]

250w per mile

What's the time component?

If per hour that equates to total vehicle draw (i read 240-300 Wh/mi for Model 3 average use; 100 on hypermiling attempts)

Need to know peak vs avg draw of the FSD system, and what percentage of the baseline comparison figure constituted autopilot use.

yeah, I am with you. watts per mile mean absolutely nothing without a time variable. Seeing the specs of the computer I would expect a steady power draw of at least 250W, only time will tell.

[Phil]

They don't have a reason to tarnish their reputation with this.

They want to sell cars. It's a business. A very lucrative one at that, with plenty on the line. Doesn't mean everything that is being claimed will actually become reality or that all targets are gonna be met. Further more, claims like this are nothing new. Which is sort of the point I was making. You seem to take everything they claim at face value while I prefer to critically question the numbers that are being thrown around.

I tried to highlight how unrealistic some of the numbers being thrown around are. Yes, fudging them here and there can make them appear better (i.e. adding fareless rides to the cost), but in the end, you won't beat depreciation (which you have so far ignored). How much does a Tesla depreciate?

I can't vouch for the following pictures credibility, but nothing I've seen so far would paint a more optimistic picture anyway:



At 100k miles, you are scraping at the 50% depreciation value. In my opinion, that number is very optimistic. Random browsing through the local used car website (we have a lot of Teslas on the road here in Switzerland) suggests a higher depreciation.

Funny thing though; While we argue about these numbers, have you ever considered how much time your car needs to travel to do 166k per year? How many fares that would be? A quick calculations suggests, assuming 5 mile rides, 33k rides per year, or 150 rides per day (assuming 220 working days a year). If you prefer to assume 20 mile rides, we still end up 37.5 rides per day. At that point, we're not even factoring in the duration it would take on average for a single ride. My guess is you'd struggle to have your car back in time after work before your autonomous taxi would hit even a quarter of the required rides to come close to make this some sort of a viable business venture. Assuming of course it will find that many people to transport around - which is really a stretch and in reality the biggest giant pink elephant in the room.

Yeah, good luck with that. Maybe if you bought a whole fleet of Teslas and create your own autonomous taxi business operating 24/7. Yes, I can already hear the slogan: "Free money!!". While you'll likely bankrupt yourself with the running cost that will inevitably hit you eventually, Tesla will make a fortune. Sounds like a genius plan for them, but for you? Not so much, I'm afraid.

[roon]

So plug in depriciation instead of writing paragraphs. $15k - $50k depending on model at 50% depriciation, assuming that figure is representative. Divide by mileage. I've been using 100k mi and seeing what that provides as a per-mile figure. Model 3 base model would charge +$.15/mi. P100D would charge +$.50/mi. If covering depriciation. Add to my last figure of approx .40-.50 USD per mile. Keep in mind their main goal, as you say, as "a business," is to undercut $2/mi rideshares.

I'm avoiding only in the sense of spending a few minutes writing a post. You can keep playing wisened curmudgeon. I can keep doing elementary math.

[Phil]

Which brings me to my next point; How expensive is Uber really? I'm seeing 1 dollar per mile figures (subject to increase depending on area and demand, obviously no different than an AV would be, assuming making money is the target). So how much are you undercutting them by, using elementary math?

All assuming of course that demand is actually there and that buying a simple Tesla could kick start your free money making taxi empire.

One thing this fun exchange has highlighted nicely though is that the ~$0.2/mile fee claim is absolute rubbish.

[Andres125sx]

Which brings me to my next point; How expensive is Uber really? I'm seeing 1 dollar per mile figures (subject to increase depending on area and demand, obviously no different than an AV would be, assuming making money is the target). So how much are you undercutting them by, using elementary math?

All assuming of course that demand is actually there and that buying a simple Tesla could kick start your free money making taxi empire.

One thing this fun exchange has highlighted nicely though is that the ~$0.2/mile fee claim is absolute rubbish.

Not completely, you´re comparing with cost per mile for normal users, while a taxi will do a lot more miles per year, lowering cost per mile, specially since the car is electric and have no maintenance like ICEs have.

Anycase I´ll insist, only problem with EVs is batteries. Once someone brings to the market one of the new technologies wich not only increase range, but also lowers price significantly, all this simple maths will be useless and none will question profitability of EVs/AEVs

[roon]

Which brings me to my next point; How expensive is Uber really? I'm seeing 1 dollar per mile figures (subject to increase depending on area and demand, obviously no different than an AV would be, assuming making money is the target). So how much are you undercutting them by, using elementary math?

All assuming of course that demand is actually there and that buying a simple Tesla could kick start your free money making taxi empire.

One thing this fun exchange has highlighted nicely though is that the ~$0.2/mile fee claim is absolute rubbish.

I already corrected the "rubbish" claim in the post following that one i.e. the .18 USD figure is an estimated cost per mile not the total cost of fare. That figure encapsulates consumables and some profit. That's the figure I've been using in addition to price per kW, additional profit, and depriciation.

So how much are you undercutting them by, using elementary math?

That was answered in the post you are replying to.

.15 + .4 is .55

.5 + .5 is 1

Don't fear the number pad, Phil.

[loner]

Huawei teams up with automakers to launch self-driving cars as early as 2021
https://www.rt.com/business/461683-huaw ... ving-cars/

[loner]

https://twitter.com/XHNews/status/1139180541268451328