Why Chip-Constrained Carmakers Can't Just Transition To Newer Chips (jalopnik.com) 256
Car buyers are discovering that supply chain constraints "have thrusted prices upwards considerably for new and used vehicles alike," notes Jalopnik.
But while last month Fortune ran an article headlined "Chipmakers to carmakers: Time to get out of the semiconductor Stone Age," Jalopnik argues it's not that simple. The implication here is that the auto industry is far too reliant on archaic tech that isn't applicable to other consumer tech fields. It's now finally reckoning with its reluctance to change, and only a fool would invest in shops to pump out the outdated silicon cars require. But is that a fair assessment? As Fortune notes in its own piece, there are reasons why carmakers — some of the largest corporations in the world — choose the chips they do. The comparison to smartphones is moot... The potential ramifications of a glitch in a metal box traveling at many miles per hour are a little more severe. That's especially true if you're talking about modern vehicles with driver-assist functions...
I asked some auto industry veterans to weigh in... What automakers require is somewhat at odds with what chipmakers prefer and are tooled to produce: smaller, more densely packed chips, that can be manufactured at lower cost and yield more units.... However, to suggest as [Intel CEO] Gelsinger did that the burden to adapt should fall squarely on automakers simplifies the issue. General purpose chipmakers don't seem to grasp the unique challenges of the automotive sector — something that became clear to me after chatting with Jon M. Quigley, Society of Automotive Engineers member and columnist at Automotive Industries. "Qualifying a product, specifically testing activities, are costly and requires time, talent, and equipment," Quigley said. "Some of the test equipment requirements are expensive and often not on hand at the OEM but will require an external lab, and booking time at this lab can be a long lead time activity, and is necessary for certain product certifications. Depending upon the vehicle system commonality, this testing might have to be performed on multiple vehicle platforms. Making changes to an existing product, changing an integrated circuit that only has the difference in the manufacturing processes would still require this sort of testing. Unless there are some compelling associated cost improvements to recoup the investment, this is not very plausible."
It's easy for those of us on the outside to miss the many steps of validation automotive components are required to go through before they end up in what we drive. Ultimately, carmakers don't care how small or new a chip is; all that matters is that it works for its intended purpose and is properly vetted... Chipmakers want as much miniaturization as possible to maximize production efficiency, automakers need significant lead time to make sure a chip will work for them. Each industry has reasons for operating the way it does. That doesn't change the fact that someone's going to have to budge to address this shortfall....
Over time, the transition to newer technology may naturally happen, but certainly not quickly enough to Band-Aid the snags of the present moment. That doesn't give anyone a single, solitary scapegoat, and it's not the easy answer anyone likely wants to hear — not prospective shoppers, not automakers and not the CEO of Intel. But it's the most realistic answer nonetheless.
In the meantime, one analyst that Jalopnik spoke to predicted automakers will try strategic partnerships with chipmakers — that is, "find ways to own or control more of the chip supply base going forward by partnering with ASIC design companies who do similar design service for networking companies."
But while last month Fortune ran an article headlined "Chipmakers to carmakers: Time to get out of the semiconductor Stone Age," Jalopnik argues it's not that simple. The implication here is that the auto industry is far too reliant on archaic tech that isn't applicable to other consumer tech fields. It's now finally reckoning with its reluctance to change, and only a fool would invest in shops to pump out the outdated silicon cars require. But is that a fair assessment? As Fortune notes in its own piece, there are reasons why carmakers — some of the largest corporations in the world — choose the chips they do. The comparison to smartphones is moot... The potential ramifications of a glitch in a metal box traveling at many miles per hour are a little more severe. That's especially true if you're talking about modern vehicles with driver-assist functions...
I asked some auto industry veterans to weigh in... What automakers require is somewhat at odds with what chipmakers prefer and are tooled to produce: smaller, more densely packed chips, that can be manufactured at lower cost and yield more units.... However, to suggest as [Intel CEO] Gelsinger did that the burden to adapt should fall squarely on automakers simplifies the issue. General purpose chipmakers don't seem to grasp the unique challenges of the automotive sector — something that became clear to me after chatting with Jon M. Quigley, Society of Automotive Engineers member and columnist at Automotive Industries. "Qualifying a product, specifically testing activities, are costly and requires time, talent, and equipment," Quigley said. "Some of the test equipment requirements are expensive and often not on hand at the OEM but will require an external lab, and booking time at this lab can be a long lead time activity, and is necessary for certain product certifications. Depending upon the vehicle system commonality, this testing might have to be performed on multiple vehicle platforms. Making changes to an existing product, changing an integrated circuit that only has the difference in the manufacturing processes would still require this sort of testing. Unless there are some compelling associated cost improvements to recoup the investment, this is not very plausible."
It's easy for those of us on the outside to miss the many steps of validation automotive components are required to go through before they end up in what we drive. Ultimately, carmakers don't care how small or new a chip is; all that matters is that it works for its intended purpose and is properly vetted... Chipmakers want as much miniaturization as possible to maximize production efficiency, automakers need significant lead time to make sure a chip will work for them. Each industry has reasons for operating the way it does. That doesn't change the fact that someone's going to have to budge to address this shortfall....
Over time, the transition to newer technology may naturally happen, but certainly not quickly enough to Band-Aid the snags of the present moment. That doesn't give anyone a single, solitary scapegoat, and it's not the easy answer anyone likely wants to hear — not prospective shoppers, not automakers and not the CEO of Intel. But it's the most realistic answer nonetheless.
In the meantime, one analyst that Jalopnik spoke to predicted automakers will try strategic partnerships with chipmakers — that is, "find ways to own or control more of the chip supply base going forward by partnering with ASIC design companies who do similar design service for networking companies."
The only disagreement is who gets to pay (Score:5, Insightful)
Guess what, the chip-makers don't seem to care about much your business so the compelling reason is that you can't make your products.
Re:The only disagreement is who gets to pay (Score:4, Insightful)
"We don't sell vanilla anymore, you need to learn to like chocolate."
Re: (Score:3)
Re:The only disagreement is who gets to pay (Score:5, Interesting)
The issue is more of leaving money on the table. There is demand, but no desire to fill that demand. Almost all embedded systems have similar requirements, it's not just the auto industry. Industrial grade chips, reliable chips, chips that have to work in poor conditions, etc.
The issue is also one of more of consolidating fabs into relatively few these days, and if there's high demand for smart phone chips then the capability for other chips goes down. Manufacturing capabilities are tight, difficult to ramp up and down, tuned to work with little wiggle room, so one tiny problem and it disrupts the pipeline. Many chip makers DO want to sell the industrial stuff but they don't have their own fabs anymore.
Re: (Score:3)
Re: (Score:2)
So a carbureted car from the 70s? And most likely pre 1975?
Re: (Score:2)
I had a '86 Nissan pickup with no computer and no engine electronics besides the ignition, choke and anti-dieseling solenoid. Before that a '84 Datsun Diesel pickup that, given a hill to jump start it, would run without any electricity, mechanical fuel injection. So mid '80's for the last non-computerized vehicles.
Re:The only disagreement is who gets to pay (Score:4, Informative)
It really isn't. The issue is not being willing to spend $100 in order to make a $50 sale. The ancient chips the automakers want were made by ancient FABs that are rapidly approaching the point where they cost more to maintain than they earn in profits. Meanwhile the auto industry wants higher volumes than were produced in the past.
Building a new fab is expensive to say the least. If you were laying out a couple billion, would you build a fab that produces outdated parts only, or one that produces the modern parts everyone but the auto industry wants? Keep in mind, there's no guarantee that the auto industry won't update eventually and leave your multi-billion dollar investment idle if you go with the old tech.
Re:The only disagreement is who gets to pay (Score:5, Insightful)
It's not about who pays. It's about an ancient rigid process used by automotive engineers that is hallowed and defies innovation.
Did, for an example, a decade ago, automotive engineers ramp up their testing methodology to find new labs that could vet chips through the cyclic regimen necessary to foresee what might happen in ten years?
No.
They were just getting out of bankruptcy. Read the history.
Did they look to Tesla's innovation, secure supply chain, and do the model restructuring needed to limit their insane offerings? Only recently did Ford kill cars they couldn't make money on because product line support costs have become so high.
Did they capture the market through unrelenting yet safe innovation, ensuring product roadmaps that were realistic, and not last-minute reactions to the EV-ing of the world?
No.
Boo freaking hoo. This is the automaker's fault. Supply chain is key, and semiconductor research should've been an organizational discipline twenty years ago. Now, a single iddy biddy part, one of those darn chips things, IS KILLING THEM.
Re:The only disagreement is who gets to pay (Score:5, Interesting)
I have run across the problem, in this last decade, where the chips were behind of what we wanted to do. Lots of cheap Arm system-on-a-chip, some industrial which is handy, but not doing the exactly right thing. Security was generally weak - they did whatever wifi wanted, which means low security - AES accelerators, which is not that slow an algorithm compared to say elliptic curve. But that's ok, we can work around it, we do elliptic curve rarely (validating certs) but AES once a key is negotiated. But then they don't do secure key storage with the algorithms you want and you're forced to have keys in RAM which kills your security story to customers. Bleh. Anyway, the chip makers did seem to be slowly catching up, realizing that not everyone wants wifi or bluetooth, and that they're not all consumer stuff that is recharged every night.
It's a bit frustrating because you're trying to find a chip that will be suitable in 5 years when the product is finally ready and not be pounding yourself on the head because the perfect chip came out halfway into the development. So you end up believing the crap that the chip's marketing department tells you but they cannot actually achieve it in the end. So yes, you're always looking 5+ years in the future, everyone doing anything remotely industrial or with constrained resources.
Re: (Score:2)
>It's a bit frustrating because you're trying to find a chip that will be suitable in 5 years when the product is finally ready
So you design a chip, get it ready in time for the product. 5 years is plenty of time. In the mean time, implement the same functionality with FPGAs and microcontrollers since it's low volume and you want to test your algorithms and RTL.
Re:The only disagreement is who gets to pay (Score:5, Informative)
Designing a chip is horribly expensive. Even producing one chip design is at least $2M just to commit the design to silicon - that's not even before the chip is made. (A mask for photolithography costs arouns $100K each, and modern chips require 20+ masks, so that's a $2M investment off the bat).
FPGAs are good for prototyping, but that's it. When I worked on the systems, the design couldn't fit on one 9 FPGAs - we had various blocks mapped out, some single core, some dual core and not everything was there. It ran at a whopping 3MHz.
Oh yeah, each FPGA station was a quarter million bucks. And it was only good for one design - the next iteration needed brand new FPGA stations, this time they contained a quarter million bucks in FPGAs, so they probably were a half million dollars.
We did have access to multi-million dollar boolean logic accelerator systems (many millions of dollars each, clusterable) which could simulate the entire design at around 100kHz or so. Of course, those machines were big and heavy they even had installation requirements for building foundation strength.
The big thing is, automotive parts are long lived. Because of requirements that mean brand new parts have to be available for 10 years, they're also often used in industry - if your product won't launch for 5 years, it's probably OK as the chip won't be obsoleted for another 15 years.
They're also the reason why you could find 30GB hard drives and other tiny storage devices still hanging around.
Safety critical MCUs, they're also another beast altogether. Many automotive parts are meant for less critical applications, like say an infotainment system. But there often needs to be a critical safety part like things like airbags and such. Those parts have extra self-checking systems built in - things that check the system clocks for example (yes, there are blocks that monitor the voltage rails, blocks that mother the clocks, and blocks that monitor other aspects of the chip. If you get that airbag light, any one of those things could've failed. Or any one of many other things are constantly being checked. Most of the time while the airbags are sitting idle, the processor is running tons of diagnostics making sure the sensors are all OK and reporting valid data, that the power rails are functioning, that the clocks are working, that the airbag itself is connected and that it's in good repair (no bad connections have cropped up, etc). It's also double-checking seat belts and seat occupancy sensors, verifying if the switch you push ti disable airbags is set, that the light indicating airbags are disabled is lit, etc. Heck, a modern airbag computer also has a backup power supply to ensure that even if the electrical system is disabled, it has reserve power to fire the airbags log the events and shut down. It's also got to monitor that these power reserves are adequate for the job as they degrade.
Re:The only disagreement is who gets to pay (Score:5, Insightful)
To be fair, not many businesses put in place measures to mitigate a global pandemic.
The reason they don't want to pay now is that they think supply will back eventually. Well, they thought it would be back by now, they took all the political leaders telling them that it would be over in 12 weeks (Boris Johnson) or that it would just go away soon (Donald Trump) at their word, and didn't want to gamble that they would be wrong.
Everyone is in the same boat now, nobody can get parts. Not just automotive grade ancient stuff, even newer parts are unavailable.
Re: (Score:3)
It really isn't. The issue is not being willing to spend $100 in order to make a $50 sale. The ancient chips the automakers want were made by ancient FABs that are rapidly approaching the point where they cost more to maintain than they earn in profits.
My understanding is that it's the opposite. Those fabs have historically been really cheap, because relatively few companies still wanted large quantities of parts that can be built using those older processes, so they had a lot of capacity supply and relatively low demand. But then, when the automobile manufacturers realized that they didn't expect to sell as many cars because of the pandemic, they cancelled their contracts, and as a result, demand for parts built by many of those plants fell to the poin
Re: (Score:3)
Those fabs were cheap only in the sense that they had long ago paid for themselves so running them into the ground made fiscal sense. There is something to the fact that they shut down during a sharp dip in demand. Re-starting a shut down fab is a non-trivial task. Had the auto makers not been so allergic to having even a single crate of inventory in storage, they could have averted some of this by keeping the orders and stockpiling a bit.
The new processes available in newer fabs can AFAIK produce chips sim
Re: (Score:3)
But then, when the automobile manufacturers realized that they didn't expect to sell as many cars because of the pandemic, they cancelled their contracts, and as a result, demand for parts built by many of those plants fell to the point that they couldn't keep the lights on, so they shut the plants down.
That is where the corporate culture of just-in-time and pinching pennies to maximize profits bit them on the arse, hard.
They should't have cancelled those orders. They should have taken delivery even though the did not need the chips for six months or -- horrors! -- a year.
Seriously, how much precious capital was that going to tie up? Sure sales, and thus revenue was down, but you should not let bean counters run your production operation. Set yourself up for the sales surge when unmet demand returns. Any
Re: (Score:3)
The issue is more of leaving money on the table.
There is no money left on the table. The cost to run fabs for old designs outweighs the financial gain from the sales. This is precisely why the fab industry is moving on and trying to force the issue.
The issue is also one of more of consolidating fabs into relatively few these days
No it's not. We're not talking about hyper modern 7nm nodes here. There are literal hundreds of fabs spread out in countries all over the world producing chips that would suit the automotive industry. Hell even Intel still run 65nm fabs. The car industry is only competing with "smartphones" (I use that word in
Re: (Score:2)
I guarantee you there is a shop across town who found a way to make Vanilla-flavored.
Re: (Score:2)
Or buy the old fab (Score:3)
Or, since they want to make old chip designs using the old process, buy the old fab. Or the old fab equipment, anyway.
Re: (Score:2)
This is the problem. Chips aren't ice cream--manufacturing them involves massive fixed costs. When the chips automakers were using came out of the same plants as the chips everybody else used, there was no problem. But now everybody else is using much more capable chips, and the plants are built to make them--but they can't make the chips the automakers still want. Making a plant just for the automaker would make chips for the automaker exorbitantly expensive. Nobody's going to do that if they can't get
Re: (Score:2)
Re: (Score:2)
At the end of the summary it has the prediction that the automakers are going to invest in chip makers that will give them what they want in a reliable manner. If the auto company owns half the stock in a company, it can probably tell them what to make.
Re: (Score:3)
Re: (Score:3)
A lot of people "chase the conclusion" and in this case the conclusion was that automakers should produce their own chips.
However, another chase is that automakers should have stockpiled enough inventory so that it still couldnt be a problem yet. They are trying to make cars now, using chips that they could have purchased years ago, yes?
Just because the conclusion got chased, in no way supports said conclusion.
Consider that the optimal solution is probably for a consortium of manufac
Re:The only disagreement is who gets to pay (Score:5, Insightful)
The funny part is that that is more expensive than just deciding how many you need in advance, and buying that many.
The real problem is that they insist on doing JIT in a way that saves them money by their not needing to make that decision. So then, somebody else who did make that decision gets 100% of the product in a shortage, because they'd already made guaranteed purchases.
The automakers need to hire some accountants to make a spreadsheet with multiple scenarios, and then hire somebody to convert that into a PowerPoint presentation for them, so they can understand that if they're not committing to orders, they're not even a customer yet, they have no power at all. Their importance is how much product they can promise to buy, not how important they think their product is.
But the idea that shareholders can get a big advantage through self-dealing is dubious. Other shareholders might not feel that they get their best value otu of the investment if they act as a subsidiary of junior shareholders.
Re: (Score:3)
What on earth makes you think some of the biggest manufacturing concerns on the planet don't already do that. It's an extremely safe bet that they do.
Oh I don't know, just spit-balling here, may because of the the fact that they didn't do it, hence the topic of TFA, and widespread reporting on the failure?
Of course your point may be that they did those things, but the execs did not effing listen. Yeah, that sounds about right.
Re: (Score:2)
Re-designing for modern parts would be a lot cheaper.
Re: (Score:2)
People think its just a fab.
If you will be fabbing upon modern-day commodity silicon wafers purchased from a 3rd party manufacturer, then you havent solved your problem at all.
Re: The only disagreement is who gets to pay (Score:2)
The arrogant auto makers have two choices. Sell no cars and wait an unknown amount of time for backlogs to clear, or start upgrading with a defined cost and schedule to plan from.
Which would you choose?
I believe all of this media coverage is designed to gain political support to force the chipmakers to lose money and bail out the automakers.
Re: (Score:2)
The arrogant auto makers have two choices. Sell no cars and wait an unknown amount of time for backlogs to clear, or start upgrading with a defined cost and schedule to plan from.
Which would you choose?
I believe all of this media coverage is designed to gain political support to force the chipmakers to lose money and bail out the automakers.
Indeed. It costs less than a million dollars to set up a really nice electronics lab with environmental test fixtures for qualifying designs. Stick some good engineers in it and don't micromanage them and you might drag your sorry car company into the 1990s.
If halting your production line costs many millions of dollars, the automakers (excepting the saner ones like Tesla) are clearly making bad tech decisions.
Re: (Score:2)
Its not just chip fabrication though. Wafer production too.
Re: The only disagreement is who gets to pay (Score:2)
We arenâ(TM)t talking about a car design. We are taking about an onboard computer that needs special packaging. The car companies have to push a new design through their bloated and inefficient bureaucracy and are throwing a tantrum because they would rather lose hundreds of billions than change their outdated thinking. That tells us way more about them car companies than it does about the chip industry.
Though not explicit, also not trivialized.. (Score:2)
The implication was not that automakers could just trivially do it, just that transitioning would be less costly than expanding fab capacity.
Unless there are some compelling associated cost improvements to recoup the investment, this is not very plausible.
Well, that's essentially the situation, the parts being used are unavailable, and the requested remedy to build out is massively expensive so the chip vendors would have to demand more money from the automotive companies to make it worthwhile. Essentially the same reason given by a chipmaker, that it's not plausible to expand with no path to recoup investment is being
Re: (Score:2)
Not clear. Smaller, denser, chips tend to have a higher failure rate and be more sensitive to environmental stresses. Cars tend to have lots of environmental stresses, and people are already complaining about the failure rate.
It may well be that the automakers literally can't budge on this disagreement. They didn't make that argument, and said it was just about costs, but that may not be quite correct.
Re:Though not explicit, also not trivialized.. (Score:4, Informative)
Re: (Score:3)
How To Build An Automotive Chip [semiengineering.com]
Re: Though not explicit, also not trivialized.. (Score:2)
Re: (Score:2)
Not clear. Smaller, denser, chips tend to have a higher failure rate and be more sensitive to environmental stresses. Cars tend to have lots of environmental stresses, and people are already complaining about the failure rate.
It may well be that the automakers literally can't budge on this disagreement. They didn't make that argument, and said it was just about costs, but that may not be quite correct.
It's easy to make modern processes more reliable than old ones. Use thick gates for higher voltage tolerance, differential signaling for noise tolerance and redundant logic for failure tolerance. All automakers need to do is pay enough for decent chip engineers that can put such things together.
Re: (Score:2)
Are those things economical for replacing a $0.50 chip with a $1 one?
Re: (Score:2)
Are those things economical for replacing a $0.50 chip with a $1 one?
If you are doing small functions in modern processes, the chips are going to be pad bound and so it costs exactly $0 to duplicate circuits for reliability. The costs will be lower with a current process compared to a bleeding edge or a 10-year-old process. There's a moving sweet spot for semiconductor cost when you aren't pushing the high end.
Re: (Score:2)
There also *are* compelling reasons to move to more modern chips. If you want to integrate fancy pants AI functionality into your car, you're going to need some fancy pants neural network chips. There, power consumption, and size is *really* important, because you need it to stay cool, and you need to be able to fit thousands of copies of your madd circuitry onto the chip. The car makers *are* going to have to update their chip designs. Intel is 100% right. The car manufacturers are just being stick in
Re: (Score:2)
Tesla isn't suffering because they're able to frantically redesign their hardware around whatever chips they can get their hands on. They probably pre-buy a lot of stock as part of this process.
Of course the entire embedded industry is doing this a lot right now, so its eventually going to catch up with everyone and they'll all be SOL again. (Or the backlog will clear by then, and things will settle down.)
We're not talking about fancy-pants infotainment or AI chips here. We're talking about the $2 (or less)
What did you expect, you fools. (Score:2)
Since the 1980s people stupidly buy cars with no technological advancements over the previous generation, so who could blame auto manufacturers for this? It's the consumer's fault 100% for not wanting more advanced vehicles. The consumer is quite happy with 40,000 people, just in the USA (one million worldwide), being murdered in automobile accidents. Most people don't know anyone killed or severely hurt in automobile accidents, so why should they care? Therefore we haven't had a single non-incremental adva
Re: (Score:3)
Re: (Score:2)
I've had a car computer crash on me before. It's annoying because you lose all the electronics, so minor things like the speedo and turn signals stop functioning. But things like steering, acceleration, and brakes keep working because they use physical connections. Well, at least, they did in that car.
So I had to pull over and turn the car off and back on again to get things working again. It happened at low speed, thankfully, I wouldn't want it to happen on the highway. But the idea that it can't happen is
Re: (Score:2)
Having the "display stuff" computer crash is much different than having the computer that controls ABS/STC, fuel injection, etc crash.
Re: (Score:2)
Brakes do have a backup physical connection, yes, as does steering, even for electric assist. But the accelerator pedal no. It's completely digital on all cars in the last 10 years. So your pedal is simply driving a dedicated ECU. Fortunately that one rarely crashes, or you'd lose power completely. Fortunately most cars have several ECUs and layers of ECUs, with the critical ones being engine and throttle control, and those are super hardened and quite reliable. They can fail completely of course, leavin
Re: (Score:2)
Re: (Score:2)
Meanwhile Hyundai has recalled 550 000 vehicles because the turn signals might flash the wrong direction. [usatoday.com]
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
8 or 9 airbags.
anti-lock brakes.
stability and traction control.
crumple zones.
reinforced passenger compartment.
Federally mandated back up camera.
blind side alert lights on side mirrors.
pre collision warning, followed by:
pre collision braking.
lane departure warning (even has assist to nudge you back if you like that sort of thing).
-Maybe you are retired and don't drive or sho
Re: (Score:2)
Therefore we haven't had a single non-incremental advance in automobile safety in since 1970something. We have technology to avoid accidents, yet they are not implemented .
That is so factually untrue that I will deem it a lie, Since the 1970s, airbags, antilock braking, and traction control have not only been implemented but sometimes mandated. Why would you post something so blatantly untrue?
This article is bullshit (Score:2)
All you do is port old designs to new nodes that are suitable for milspec chips, done deal. Can't port the old design? Retool and do new masks so that the new chip can handle the same inputs/outputs as the old one.
In later models, maybe then you can work on more-sophisticated ICs that can do the job of older 40-90nm chips better, or consolidate functions into a smaller number of ICs to reduce the need for now-redundant modules in cars.
Just one option is Globalfoundires. Their FDX nodes are already in use
Re:This article is bullshit (Score:5, Informative)
Automotive spec is more stringent than milspec.
A lot of slashdot readers do not spend much time reading datasheets, so they don't know this, but new chips made with smaller internal circuitry have the same packaging as older components, and are generally more durable, and more resistant to noise and overvoltage than older chips were. These days you only pay a few cents extra for automotive grade over consumer grade for the new parts. It is only the obsolete parts where there is a big difference in capabilities.
People just can't get over the wrong idea that smaller is more fragile.
In almost every case, there is already a replacement offered by the IC vendors for the chips that the automakers want to buy. A huge amount of R&D goes into products for this sector. It is a major part of the R&D cost of the whole industry. And thankfully, because my robotics work benefits by all the cheap, automotive-grade parts! What's hilarious is that these are better parts than are in most of the cars...
Re: This article is bullshit (Score:2)
If auto spec is more stringent than mil spec I would say that auto tied itself up in a web of stupidity because mil spec is more demanding in practice. Auto deserves to get sodomized with a pitch fork
Re: (Score:2)
When the auto people talk about validation, they are talking about lives and are recalls and lawsuits. In 2021 for example, I believe every GM SUV was recalled for a controller issue.
Electronics manufacturers Used to understand this (Score:4, Interesting)
Equipment reliability is paramount.
"we'll fix it in the next release" has come from the idea that hardware is like software.
All we have to do is look at the Pentium divide bug and all the idiot processor bugs that have crept in and exposed user systems to horrifying vulnerabilities (can we say meltdown and spectre? I knew ya could!).
While capitalists generally like the returns of "Move fast, break things... Fix it later", hammering ones customers to accept it...
Well, as investors in the semiconductor industry are about to find out, that attitude will be costly in the long run in the C suite.
Re: (Score:2)
Exactly
Re: (Score:2)
Tesla (Score:5, Informative)
Funny that Tesla doesn't have that same problem.
They pivoted to new chips with certification and doubled there sales in the last quarter.
Re: (Score:2)
Re: (Score:2)
Tesla is not in business to save other car manufacturers.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Assuming you meant "valuation," and that you mean in comparison to their capitalization, but what armchair pundits don't know when they repeat this is that Tesla sells shares to raise money, and the big automakers sell bonds to raise money. You have to include the bond value in the capitalization equation for GM or Ford. When you do that, you find out that Tesla's valuation isn't really that high given their growth rate.
Ford has a P/E of 16.66 and Tesla's is 403. One of those numbers is real. The other is m
Re: (Score:2)
Re: (Score:2)
Re:Tesla (Score:5, Informative)
Tesla sold 241,300 cars last quarter.
VW sold 110,000 cars last quarter.
GM *the largest auto maker* sold 446,997 cars last quarter.
Do you want to re-word that statement ?
Maybe a better explanation... (Score:3)
It's called Quality Assurance.
Most things are produced in areas with the cheapest labor and so workers have no personal pride (and bosses that just say work faster and also don't give a darn) and no reason to care about quality of the work they do. So you contract with them to produce quality parts and let them know the testing standard. You process everything in batches according to statistics. If it fails the testing you reject the batch and company you ordered from has to redo the order at their expense. Therefore they can't cheat and are motivated to produce at a quality that will always pass the quality specified. If you don't do this companies that manufacture are economically motivated to cut corners.
The Japanese use to be know for making junk. They learned quickly enough about Quality Assurance (20-30 years; a generation and a half) and are now mostly acknowledged for their quality.
Re: (Score:2)
I don't think car makers should get off the hook so easily, in the name of "quality assurance." They knew chip technology wasn't standing still, but kept relying on old chips anyway. It's ore like a common problem in the software world, where companies get caught running antiquated software on antiquated mainframes because they are "tried and true." The reality is they just didn't do the work to keep up with change.
Problems may have multiple solutions (Score:5, Interesting)
I see a lot of crying on automotive industry falling apart, but none of them have done anything about it other than take features away from customers or shut down production.
Less QA testing is required if you use fewer chips. Why do I read stories that say 'modern cards may have 3000 chips'? Why? Just Why?
I guarantee you that my desktop, both screens, laptop, 2 smart TVs, and my chromebook, all sitting in this room with me, do not have 3000 chips between them.
Unfuck your designs. Use fewer chips. Make the chips you do have DO more. In fact, you could make 1 triplicate of super chip, and have that trio do almost everything. Then you just need maybe another 50 microcontrollers and actuators to do the rest...and you'd be able to meet reliability and redundancy requirements on the logic head.
Re: (Score:3)
Yeah but if they just have a microcontroller and an actuator, plus a mosfet and capacitor, then when the mosfet or cap predictably go bad, they only sell a tiny little cheap part. If the little controller board for the actuator has 20 chips on it, they can charge a lot more when that same cap goes bad. And it is the same cost for them to refurb it either way. Even the refurb has more value this way.
I'm so happy it is biting them in the ass for once.
Educate Yourself before pontificating (Score:4, Informative)
Less QA testing is required if you use fewer chips. Why do I read stories that say 'modern cards may have 3000 chips'? Why? Just Why?
Why would you assume less "QA testing is required if you use fewer chips"? Do you have any idea what's involved in automotive testing? If you're qualifying a box for use in the car, why would you assume that it would take less work Qualifying it for use in the vehicle if the box has one chip in it rather than 50? The same number of production representative prototypes will have to go through the same environmental tests, for the same amount of time, with the same variations in temperature, humidity, moisture (with and without salt), changing power levels, vibration and possibly G-Levels. The automotive environment is really challenging and testing electronics to work in it is not trivial nor can it be minimized by sweeping everything up in one chip.
I guarantee you that my desktop, both screens, laptop, 2 smart TVs, and my chromebook, all sitting in this room with me, do not have 3000 chips between them.
Your "guarantee" isn't worth the bits used to store the statement. Why don't you count up how many chips are in the seven devices you've quoted? You'll find that each device has at least 500. Each one has a processor and a support chip, but there are a myriad of power management, driver (especially for LCD displays), communications, buffer and logic on the PCBs in the different devices. You probably don't think of them as "chips" but they're silicon and they're using the same technology as used in cars (although the versions used in cars are built to higher standards).
Unfuck your designs.
I'll stop here because you're clearly making demands towards multiple industries and disciplines you don't know anything about.
Re:Problems may have multiple solutions (Score:5, Insightful)
Less QA testing is required if you use fewer chips. Why do I read stories that say 'modern cards may have 3000 chips'? Why? Just Why?
In short it's about reducing wire counts. Think about the difference between a dumb climate control flap servo and a smart one. The dumb one needs no micro but 4 wires, the smart one needs its own micro, but only 3 wires. And one of those is a ground and another one is a power bus, and only one is a data lead, so most of those wires can fan out. If you've got six or eight climate control servos then you're saving a boatload of wiring by using smart servos.
Re:Problems may have multiple solutions (Score:4, Insightful)
Why do I read stories that say 'modern cards may have 3000 chips'? Why? Just Why?
I assume you mean "cars" instead of "cards." The answer is to save on weight, thus increasing fuel efficiency. Before in-vehicle busses you used to have a thick bundle of wires going to each door. Two to run the lock/unlock actuator, three for the door lock switch, four for the rear view mirror motors, two for the window actuator, a few more for any lights in the door, etc...
Now you have three wires, one for power and two for data, and a single door controller module controls all the devices in the door. Now you have fewer wires that have to flex when the door opens and closes, which makes it more reliable. It also weighs less
Now spread that out over the whole car. Light clusters, fuel sensors, collision radar, backup cameras, impact sensors for airbags, motorized seat controllers, seat occupancy sensors... Weight savings add up pretty quick when you can put everything on a three-wire bus. It's also cheaper to install and build.
If you don't believe me look up wiring harnesses in eBay motors, and look at how complicated a *modern* car harness looks - keeping in mind it's a fraction of what would be involved if there weren't controllers all over the vehicle. /Source: I work for a company that writes software for automotive suppliers, and most people I know work for car companies.
Absolute lag vs slower pace of innovation (Score:5, Interesting)
It also has to do with the obscene amount of "process" that goes on behind the scenes. I catch little glimpses of it at work now and then, and if the efficiency of software development of their mobile apps is even remotely indicative of automaker's general pace of development, they're screwed. There's absolutely no way they can change fast enough to even *match* the pace of chip development, let alone catch up to where they need to be.
Tesla, maybe Rivian, and the newer companies are the only ones that have a chance, because they were built to keep up and even in some cases lead various technical advancements.
Re: (Score:2)
That's an interesting comparison but I'm not convinced it's true. Much has been said about intel sticking with the 14nm process for 7 years, but even average process changes are only happening every 3 years or so.
The whole problem seems so absurd to me. Intel is leaving old fabs on forever, and Global Foundries keeps cancelling plans to update their 12nm process. Most companies leave old fabs running even after they've lost their edge to get back as much of the initial investment cost as possible. This
so in a few words (Score:2)
Re: (Score:2)
It's total amateur hour for anything developed in-house.
Seems to me that the (Score:2)
Also, a very good point showing that to much consolidation is not always a good thing.
Re: (Score:2)
Or, the chip makers don't terribly much care about this bit of the market, they can just serve them on the side while catering to the real money in computers. If the vehicle manufacturers don't want technology that's reasonably similar to the stuff used in computers (or the stuff used in computers in the last 5 years), then it's just not worth serving them specially.
Re: (Score:2)
Intel is correct (Score:2)
The automakers have constructed a hyper bureaucratic process for updating their technology and believe that as customers they can impose their will on the suppliers. They are right, but the suppliers are under no obligation to meet their needs and idiocy. The automakers need to take a long hard look at their processes and make them ore efficient, not cling to bloated management processes because "its the way things have always been". To not adapt is to die.
You made your bed (Score:2)
You made your bed, now you lie in it. Car manufacturers cancelled chip orders, screwing their suppliers. And now, they want to have special treatment, because?.. reasons... Why?
Normally I would say this could be a good lesson, and they would learn from it. But who am I kidding? They don't want to invest in anything, unless absolutely necessary.
They will sell 1990 style UIs with embedded receivers, which can no longer be replaced (thanks to everything including A/C being controlled by it). They will install
It's not just testing either... (Score:5, Insightful)
The issue isn't just testing and certification, either. That's expensive and time-consuming enough, and I doubt the chipmakers are willing to shoulder any of the costs. There's also questions of reliability and durability. The chips in a car are subject to conditions that would appall any electronics technician. Will those modern 7nm chips operate at an ambient temperature of 180F while being subjected to continuous 3g vibrations? Not to mention the corrosive mix of oil, dirt, coolant and road chemicals some can be subjected to and the lack of cooling airflow almost all will have to deal with. And will they continue to operate reliably for the 10-15 year lifespan of typical automotive parts? Think back to the last time you had to have the ECU or one of the other major control units in your car replated. Have you ever needed to have one replaced? On top of that, auto makers are required to insure that there's a supply of spare parts available for the expected lifetime of that model, and that lifetime is set by consumer expectations not by manufacturer preferences and enforced by the government not just contracts. Are the chipmakers willing to sign binding contracts guaranteeing to continue making those chips for at least the next 20 years subject to having to buy new cars for anyone whose car can't be repaired 20 years from now because the parts those chipmakers made are no longer available?
And the final bit: guarantees and liability. Notice how all electronics these days disclaims all liability if it malfunctions? You can't get away with that with automotive parts. They have to perform as advertised for their entire lifespan and the manufacturer is legally required to stand behind that guarantee.
Re: (Score:3)
7nm refers (very approximately) to the dimensions of transistors and metal traces *within* the chip. These components are built in the fab, then sealed into place with thick layers of silicon dioxide, followed by other forms of airproof packaging. They are insensitive to vibrations due to being sealed on all sides, and because of the relative force strengths at that size. They are insensitive to oil and dirt because the packaging keeps oil and dirt out. They are sensitive to extreme temperatures, but are te
Blast from the past (Score:2)
1993 Sumitomo resin plant explosion [apnews.com]
Above was a single event that disrupted chip production and brought a price spike. We are better at repeating the past than at learning from it.
Re: (Score:2)
Well there was the Renesas [autoblog.com] fire march 2021.
"Move Fast and Break Things" is really awful ... (Score:2)
... when the thing being broken is several thousand pounds of metal, glass and plastic moving at 65 miles per hour.
In 1988, some idiot got drunk, climbed behind the wheel of his pick up, drove on the wrong side of the interstate at high speed and had a head-on collision with a school bus, which killed 27. (cite: Carrollton Bus Crash, 1988)
The manufacturer of the school bus was sued over the deaths, mainly due to deficiencies in the design. That bus has been built to an older (but legal at the time it was
Re: (Score:2)
People who write code just don't get it.... (Score:2)
That waits for minutes if not hours to do anything just don't understand
If they did the next few sentences would make sense.
Let's say that we build every system in a car, to talk to one (1) chip. Let's say it's Apple's M1 chip, or Intel's Core H series. Either one, it doesn't matter.
How would you connect all of the inputs? I doubt there enough I/O ports on either chip. Not logical, but physical I/O ports.
So even if you could magically get all these different systems to connect directly to either CPU
Aren't they big enough? (Score:2)
Why can't carmakers build/fund their own fabs? Takes time, of course, but if the open market is shoving a hot spike into their $trillion business, why dont they just do it themselves? Share one among themselves (no, of course they're not all the same)?
Asked like I asked why couldn't taxi companies just build and share an app that lets you call a yellowcab to your location?
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Following the issue very closely, I see... lololol
Re: (Score:2)