Right. When Tesla really gets going on the FSD upgrades, they’ll have a giant pile of “gently used” GPU-based modules. Why not stick these in new cars that don’t have FSD? They aren’t tires or water pumps or whatever that wear out. They’re rated for years of continuous use and frankly have a gentler environment than a home PC.
It sounds like you’re talking mostly repairs here–which is definitely another angle; Tesla could (unless they did a VIN lock as you suggest) just stockpile the modules and use them for repairs. That might be easier to get away with.
This was my thought, and I’d almost posted to say it earlier. Selling used parts as new, in a new car, would probably get them in lots of trouble. Selling refurbished parts for repairs happens all the time.
I read up a bit on remanufacturing and it does seem like this is the most likely scenario. At the least, putting remanufactured components on a new car would require some kind of disclosure (maybe I should read the paperwork I got from Tesla…). There’s probably enough demand from repairs (especially projecting out 10+ years) that there’s no need to use the modules on new cars.
Still, this does seem like a bit of a gray area. Obviously, it’s not a problem using recycled materials in a car. But at what point does recycling turn into remanufacturing? Does anything short of melting the product down into its component materials count as remanufacturing?
The actual parts cost of a computer with GPUs like that is probably $400-$800* or so. There’s actually multiple computers in that module. (and calling it a ‘computer’ is a bit of a misnomer because automotive embedded systems often have many additional components that a desktop computer doesn’t have, and they are on special PCBs, with special higher temp parts, with special conformal coatings and latching rugged connectors)
The problem is that you’re wrong about wear. Thermal cycles, vibration, component age, and moisture all cause some wear to the circuit boards. Yes, they will more than likely last 10+ years, but it is by no means a guarantee. Used parts carry a greater risk of failure.
*source : I work for a supplier of such devices now
The board design and module is Tesla’s, so I can’t really speak about that. But the GPU is qualified for a much longer life than a desktop one. And the thermal environment is much better on a Tesla, as counterintuitive as it might be. Tesla uses the same water cooling loop on their autopilot module as on their main battery and motor. It’s a really effective system and probably doesn’t get more than around 50 C in use. A desktop GPU when running flat out is liable to be at 80-90 C. The voltage is also higher on the desktop part, which causes greater electromigration and other “wear”. The auto part is run under much milder conditions, and for an ASIC, there is a very non-linear relationship between conditions and lifetime. A part that’s driven hard might only last a year of continuous use. But underdrive it a little to meet a 10-year life (say) and it’ll probably last forever in practice.
BTW, while I genuinely don’t know the numbers involved, I highly doubt we’re charging $400 or less. No one ever accused our company of lowballing on part prices.
Fair enough. I know the fully assembled modules without machine vision grade GPUs sell to the automotive companies for around the $400 figure. Don’t know what the premium chip cost for the Nvidia GPUs is.
Technically, Nvidia should just offer a chip swap. That is, accept back from a customer used or failed chips, which gives the IP license back to Nvidia, then just charge them the actual manufacturing cost of the chip. (which is likely less than $50-$100 per finished IC)
I mean, not just them, the whole industry should go to an IP model where the hardware is almost free in itself but the IP is what costs. That would be a logical progression.
I mapped the outlets in my garage, and I hope I’m in good shape. Eight of the 10 outlets are on a shared circuit with the master bath and bedroom. The other two are shared with the TV room. The good news is that the two outlets shared with the TV room are on a 20 amp circuit. One of the outlets in the garage is a 5-20P. I’d always hoped it was it’s very own circuit, but it is not. Tesla claims that is 4MPH charging instead of 3MPH for 15 amp.
My main concerns: is the circuit really on 12 gauge wire, and is there too much stuff on it to charge at 16 amps?
In the TV room I have the TV, receiver, Tivo, and wifi AP. Whenever nobody is using the wifi I’ll check and see what the draw is, but I don’t think it’s too much. On that same circuit in the garage is a freezer, which either pulls zero or one amp. I can always move the freezer to the other circuit.
My goal is to get 25-30 miles of charging in nine hours when it’s 20F outside. It’s been reported that 110V 20 amp can actually do 6MPH, instead of the Tesla’s official 4MPH, in which case I’m probably fine.
The other option is you can just swap the breaker for the TV room to 220 volts. Swap the receptacles to 220 volt receptacles. Then your TV, Tivo, receiver, and Wifi AP can plug into that receptacle and just use it because they are most likely using power supplies that are world standard. Buy a small $44 transformer from Amazon and use it to power up a 120 volt power strip if you need to.
That cuts your current draw in half. Then you can charge at just 12 amps and get 6+ MPH.
I’d thought about that, and if it was just the TV I could probably do it, but it’s every outlet in that room. All it takes is one time forgetting and plugging in the vacuum… (of course if I really did that I’d swap for 6-15 outlets, or whatever).
I tested, and the “off” draw of the TV, etc. is .45 amps, and watching TV it’s 1.25 amps. I guess I should plan on moving the freezer to a different circuit.
Of course, I might find out that it’s $250 to move the dryer outlet to the garage, in which case there’s no point in bothering with 110 charging. I can’t justify getting an electrician out to do an estimate until I at least have a delivery date, though. All I can do until then is make plans and bother you people.
You said it’s just 2 outlets shared with the TV room. And yeah, just swap those to an outlet meant for 220. And then that transformer I linked, or a bigger one, can be used to power a power strip that you can plug in stuff that can’t take 220 volts.
If you plan on living in that house for any length of time then you’re going to be making the upgrade whether it’s for this car or another EV. Given the price of the car it’s not like it’s any kind of financial burden.
That part actually makes sense to me. They and the Gulf states need to figure out a way to remain wealthy as oil becomes less important, so leveraging their vast resources to be like sugar daddies for new industries seems like a smart move. It’s a better strategy than riding the horse until it’s totally dead.
What baffles me is the “funding secured” line. It sounds to me like if someone approached you about selling your house, saying they really want it… but there isn’t even a formal offer made, much less a title search, and claiming you claim that have a buyer lined up?
I agree that it seems like jumping the gun a bit, but to stretch your analogy a bit further–if you’re the seller, then you know the title search is going to come back clean. So if that’s the main obstacle, then it might not be too much to say you have a secure deal.
The Saudis are serious at any rate, given their existing 5% ownership in Tesla. And if they’ve been talking for years now, that probably lends some credibility.
It wouldn’t surprise me if the SEC does something about the statement, though it would surprise me if it’s more than a slap-on-the-wrist fine.
The Middle Eastern understanding that the petroleum age will eventually end is almost proverbial at this point. They’ve been saying things like “the stone age didn’t end for lack of stone, and the oil age won’t end for lack of oil” and “my grandfather rode a camel, my father drove a Range Rover, I drive a Ferrari, my son drives a Lamborghini, and my grandson will ride a camel” for many years.