Hacker News Reader

Comment by twtw | Hacker News Reader

The most interesting question about this is what their motivation was to justify building an in-house ASIC design team. That isn't a super hard thing to do for a company of Tesla's size, but it is nontrivial and seems like a significant investment. It is probable that they got massive performance improvements by specializing their hardware, but it isn't clear that they couldn't have gotten the same perf improvement by upgrading to the next gen nvidia system (drive Xavier or something like that, the one with Volta + NVDLA).

View on HN

It's scary when a company with chronic implementation issues decides to take on a hard problem that is way outside its core competency.

If they succeed, it means slightly improved efficiency and cutting NVidia's profit margin out of their cost structure. If they fail, it would cost billions, and obscure chip bugs could put lives at risk. That risk/reward doesn't seem worth it unless you really know what you're doing and are confident you will succeed.

Of course, it fuels the hype machine in the short term.

"It's scary when a company with chronic implementation issues decides to take on a hard problem that is way outside its core competency."

This seems like the biography of Elon Musk's companies. He gets a crazy hard problem that people say he and his companies can not do, then he tries to implement it, often falling well behind and well short of the goalposts due to lack of foresight and overambitious scheduling. There were naysayers the whole way, including just as loud as there are now if not louder, and yet his companies' achievements and networth continue to climb.

What crazy hard problems have his companies solved?

Spaceflight was solved decades before Elon was born. Reusable rockets weren't a hard problem--they were simply a problem the incumbents were unwilling to address because it would have massively cut into their revenues and profits.

Electric cars actually predated ICE cars. The issue with EVs was always the charging infrastructure, which Tesla solved...by simply throwing a lot of money at it. (Not hard, just resource intensive.) Their batteries are built by Panasonic.

Boring Co literally is just a used tunneling machine. They have literally not done anything with it except test it out below the SpaceX parking lot (and the innovation in boring would come not from the tunneling but with the post-tunneling construction of the tunnel walls, stations, dirt removal, and ultimate extraction of the boring machine).

Even at Paypal, their biggest innovation was developed by Musk's biggest pre-Paypal competitor (Thiel's company) before they merged to form PayPal.

  Reusable rockets weren't a hard problem

Armchair engineer over here

I'm willing to back up my statements with my existing login, and I don't need to be an engineer to read the history of resuable launch systems. They've been around since the 1960s, and space-capable launch systems were a thing in the 1980s. They were killed the first time around by the collapse of the USSR, and then again when the Iridium went bankrupt, eliminating the demand.

You have nothing to back up the statement that "reusable rockets are not a hard problem", and let me back up this claim of mine by pointing out that only partial reusability exists today. In addition I invite you to build reusable rocket systems, sir, since you have already claimed that the problem is "easy". You could make some good money solving that problem, and you would once and for all put this argument to bed.

Everybody including every other space CEO, people from all levels of NASA and pretty much everybody that is innvolved in Space is giving SpaceX top marks on innovation on almost every single level, production, operation, technology and so on.

But of course you know much better how simple all these things are. For you to disagree with that is just emberecing yourself.

The same applies in a lesser degree to your other comments. Do you seriously belive that the origonal EV in 1900 was the end of development and all Tesla did was 'throw money at it'? That is the hight of stupidity and you again wont find a single expert on the subject agreeing with you.

Also the battery chemstry is co-licenced by Tesla and Panasonic with the design of the cells and even part of the chemistry being inhouse at Tesla. Panasonic can not sell this technology.

So when you are spreading baseless FUD, at least get your information correct.

You seem to have lost all rationality in context to of this question.

>It's scary when a company with chronic implementation issues decides to take on a hard problem that is way outside its core competency.

Not quite... The chip was designed by Jim Keller, who was a key player in designing the highly successful AMD Zen architecture and Apple's A4 and A5 SoCs (pretty much a legend).

Jim is known in the industry as having an unquenchable thirst for really hard problems. Solves them, leave the company to recharge and goes on to the next company/project.

Because it's expensive.

If you have your own ASIC it's a differentiator that's hard for other players to compete with. They'll take this to potential investors (or stockholders the "market") and say this is a big expensive thing we have that nobody else has. It's expensive and time-consuming for them to replicate and we also own a lot of IP that they would need.

You and I might know it's bullshit. That's not true of the investment market in general.

It has become easier to make ASICs in-house than it used to be, particularly if you're trying to do something relatively straightforward (versus trying to build a fully featured, backwards-compatible x86-based cpu architecture). The teams which developed Bitcoin mining ASICs were pretty small, for instance (although that's an extreme case of simplicity). You don't need your own foundry.

Reply on news.ycombinator.com