Rivian Unveils Custom Silicon, R2 Lidar Roadmap, and Universal Hands Free

is everyone designing their own silicon getting so much additional them-specific utility out of it that it's actually worth it?

Rivian has a huge interest in being the outsourcer for legacy automakers. They’re not able to sell $100k cars enough and even with the promised R2, they probably will only be a small-ish player in the EV market. Their CEO recognizes how crazy good Chinese EVs are and currently they’re not even a competitor for Tesla.

But, VW is willing to pay $5B for their software platform. I think they want to extend that to being able to sell custom chips and “AI” capabilities, whatever that means.

Which honestly is crazy to me. I have a Rivian, and to say the software is disappointing would be an understatement. There are heisenbugs galore; some examples:

* Doors refuse to open

* Lose the ability to control media playback using any controls

* Any button in the UI just opens and closes the windows

Granted, I'm a server side/backend engineer mostly, and I don't know much about writing software/firmware for a very hostile emf environment. But if any project I worked on had bugs like this, fixed at the rate they're fixed on Rivian, I would assume a badly flawed architecture or non existent technical leadership

Yet VW paid billions for this very software. I can't imagine how bad it must've been on their own stack that they gave up and bought this other seemingly broken stack

I have the same question. It makes sense that they might need bespoke software, but how could they possibly be more efficient at creating chips than an AMD/Nvidia?

This sounds nothing like my experience, you should get that vehicle serviced.

Well, if AMD and/or Nvidia were to invest on a chip for an auto, maybe you might have a point.

I share your skepticism. This feels like an attempt to tap the trainloads of money piling into "AI", for a company that is in pretty desperate need of more cash to stay alive.

In a vacuum there are potentially some advantages to doing your own silicon, especially if your goal is to sell the platform to other automakers as an OEM.

But custom silicon is pricey as hell (if you're doing anything non-trivial, at least), and the payoffs have a long lead time. For a company that's bleeding cash aggressively, with a short runway, to engage in this seems iffy. This sort of move makes a lot more sense if Rivian was an established maker that's cash-flow positive and is looking to cement their long-term lead with free cash flow. Buuuuut they aren't that.

NVIDIA has been selling automotive-specific silicon for a decade.

Both AMD and Nvidia have strong automotive offerings.

Possibly. Realistically this is replacing the expensive category of FPGA (Zynqs or similar with strong hardware CPU cores), this means they get all the peripherals they desire in hardware, and they can pick the core variant in order to optimise for their workloads (all the different vector extensions for example). There's an interesting market for that kind of thing, either full FPGA to ASIC replacement, or drop in replacement FPGAs of lower cost (The Rigol MHO98 replaced the Xilinx FPGA of the previous generation with a substitute from Fudan). If you're shipping a lot of hardware, that sort of thing becomes worthwhile.

Same, I've had mine for a couple of years now with no notable software issues at all.

I would wager that's because there isn't a lot of existing silicon that fits the bill. What COTS equipment is there that has all the CPU/Tensor horsepower these systems need... AND is reasonably power efficient AND is rated for a vehicle (wild temp extremes like -20F to 150F+, constant vibration, slams and impacts... and will keep working for 15 years).

Yea, Tesla has some. But they aren't sharing their secret sauce. You can't just throw a desktop computer in a car and expect it to survive for the duration. Ford et all aren't anywhere close to having "premium silicon".

So you're only option right now is to build your own. And hope maybe that you can sell/license your designs to others later and make bucks.

NVIDIA orin series is the big one for tensor horsepower. Horizon robotics and Qualcomm also have competitive automotive packages.

They are all expensive, but less than the risk adjusted cost of developing a chip.

I was just in a discussion on this very topic. It's the build vs buy equation applied to silicon. Early in the tech boom the entire silicon stack was proprietary and required a lot of time and investment to train up people who could design the circuitry, we got our first "ASICS" which was basically a bunch of circuitry on a die and you then added your own metal layer so it was like having a bunch of components glued to a board and you could "customize" it by putting wires between the parts. Then we had fabs that needed more wafer starts so they started doing other peoples designs which required they standardize their cells and provide integration services (you brought a design and they mapped it to their standard cells and process). And as the density kept going up they kept having loots of free space they needed to fill up. The 'fabless' chip companies continued to invest in making new parts until the pipeline was pretty smooth. And at that point the level of training you needed a the origin to get it into silicon dropped to nearly zero, you just needed the designs. And into that space people who were neither 'chip' companies, nor were they 'fabless' OEMs, realized they could get their integration needs met by asking a company to make them a chip that did exactly what they wanted.

One the business side, the economics are fabulous, your competitors can't "clone" your product if they don't have your special sauce components. So in many ways it becomes a strategic advantage to maintaining your market position.

But all of that because the all up cost to go from specification to parts meeting the specification dropped into the range where you could build special parts and still price at the market for your finished product.

A really interesting illustration is to look at disk drive controller boards from the Shugart Associates ST-506 (5MB) drive, to Seagate's current offerings. It is illustrative because disk drives are a product that has been ruthlessly economized because of low margins. The ST-506 is all TTL logic and standard analog parts, and yet current products have semiconductor parts that are made exactly to Seagate's design specs and aren't sold to anyone else.

So to answer your question; apparently the economics work out. The costs associated with designing, testing, and packaging your own silicon appears to be cost effective even on products with exceptionally tight margins, it is likely a clear winner on a product that enjoys the margins that electric vehicles offer.

Isn't that risk balanced by a healthy reward of controlling their verticals and possible secret sauce?

And their chips give "1600 sparse INT8 TOPS" vs the Orin's "more than 1,000 INT8 TOPS" -- so comparable enough? And going forward they can tailor it to exactly what they want?

Mostly it costs hundreds of millions to develop a chip; it relies on volume to recover the cost.

NVIDIA also tailor their chips to customers. It's a more scalable platform than their marketing hints at... Not to mention that they also iterate fairly quickly.

So far anyway, being on a specialised architecture is a disadvantage; it's much easier to use the advances that come from research and competitors. Unless you really think that you are ahead of the completion, and can sell some fairly inflexible solution for a while.

Service can’t do anything about the state machine being wrong.

The Rivian app does not permit you to send a command to the car while the app thinks the car is processing a command. Trunk opening? You can’t unlock the door. On top of this, if you try to open the trunk while outside Bluetooth range and then Bluetooth connects, you are still stuck waiting for the pending command to complete.

Oh, and the ridiculous “hey let’s always remind you that you own a Rivian” Live Activity seems to synchronize on a schedule that involves being hours and hours out of date.

The Rivian app sucks.

Orin is Nvidia's last generation. Current gen is Thor at 1k TOPS. Rivian's announcement specifies TOPS at the module level. The actual chip is more like 800 and probably doubled. Throw two Thors on a similar board and you're looking at 2000 sparse int8 TOPS.

I've been involved with similar efforts on both sides before. Making your own hardware is not a clear cut win even if you hit timelines and performance. I wish them luck (not least because I might need a new job someday), but this is an incredibly difficult space.

I own a Rivian too, and previously owned a Tesla. While I too have my gripes about the UX on the Rivian, it still beats the cr*p out of a Tesla.

Having to work with Qualcomm is enough reason to not buy Qualcomm

I agree that the app leaves something to be desired - my personal pet peeve is that it shows stale or cached data while waiting to do some async update, leading to just outright fabricated charge or lock state. Never had those kinds of problems with the truck's software proper though

Nope, it's replacing Orin.

very interesting. thanks for sharing!

stuff like that happens with tesla.

one funny one is that periodically you can trigger the "more cowbell" rainbow road easter egg. You can cancel the road animation, but you can't cancel the easter egg music or control the volume.