NVIDIA disputes Tesla’s TOPSApril 23rd, 2019
A good article from Electric NVIDIA disputes some of Tesla’s FSD computer claims, but they miss the main point nicely points out the importance of power consumption in this debate. I have some more opinions on the matter.
What makes a fair comparison? Is it fair to compare a product in mass production to a product that is in development? Not really.
In regards to Nvidia fires back at Tesla's claim that it created the world's best chip for self-driving cars, the title sums up the article very well.
But I cannot buy this chip today. I am only able to gain access to it by signing legal documents and joining an Early Access Development program.
If we are comparing chips that are in development, then Tesla described how their next generation would be better by a magnitude of 3.
But we are not. We are comparing chips in production. So Elon was right to claim anything he wanted about being the best as long as it pertains to currently mass produced.
Technology doesn't win because it is the best necessarily. Look at VHS and Betamax. Technological races are won by mass adoption. With Tesla, we know exactly how many cars will have this. All future installations + past production/FSD.
The research and development costs need to be recovered. Tesla's provided a clear plan to incorporate them into the hundreds of thousands of vehicles he has sold and is currently selling.
How much is the FSD computer for Tesla? Well Tesla is incorporating them into all vehciles now. Which means they are saving the cost of the Drive Xavier. How much are they saving?
"In terms of cost, the silicon cost of this solution is about 80% of what we were paying before. So we are saving money from using this soluton."
Not only is Tesla improving their hardware, they are saving money doing that vs what they were installing before with HW 2.5. Their cars are getting even cheaper.
Nvidia is not in a good position here. Not only is Tesla no longer incorporating their Drive Xavier chips into their vehicles, they have proven that it is more economically viable to not. As long as you can get past the R&D phase which they have.
The competition has options:
- charge < $6,600 for FSD (less profit margin)
- charge $6,600 for FSD (not first mover)
- charge > $6,600 for FSD (not as appealing)
Of these 3 options, there is no outcome better than Tesla. No matter what, they are at best a second mover offering equal quality FSD, or (more likely) they have a less appealing offer.
In terms of being the first mover - at least in North America to achieve volume production of FSD vehicles - having a lead here with demand, hardware, software and data is a tremendous advantage. If competitors had 3/4 of those things, it would be as good as having none at all. You need all those pieces to complete the puzzle.
Congratulations to Tesla on this great achievement. I can't wait to get my FSD computer.
Mediocre Developer always striving to be better. Learn from others' mistakes, learn by doing, fail fast, maximize productivity, and really think hard about good defaults. Computer developers have the power to add an entire infinite dimension with a single Int (or maybe BigInt). The least we can do with that power is be creative.