[silverbyte]

Brilliant article i came across:
https://jalopnik.com/i-asked-experts...o-n-1847739665 - "I Asked Experts Why Carmakers Can't Just Transition To Newer Chips In Stock. Here's What They Told Me"

“I’ll make them as many Intel 16 [nanometer] chips as they want,” Intel chief executive Pat Gelsinger told Fortune last week during his visit to an auto industry trade show in Germany.
Carmakers have bombarded him with requests to invest in brand-new production capacity for semiconductors featuring designs that, at best, were state of the art when the first Apple iPhone launched.

[adroit_91]

As someone from a semiconductor background, and having worked a couple of years for automotive chipsets (Advanced Driver Assistance & Safety — ADAS, as well as entertainment), here’s my 2 cents:
1. Automotive chips need a lot more qualification for long-term reliability and robustness than conventional smartphone/workstation grade chipsets, as failure can be catastrophic when one is driving on the highway at 100 mph;
2. Intel isn’t exactly known for their embedded systems chops, having had multiple false starts in that segment — with ARM chips as well as their own architecture;
3. Intel is known for killing off projects suddenly, which would leave the automobile manufacturers high and dry (who like to retain similar systems for 6-10 years, with mid-life facelifts being very much like service packs in software parlance rather than a new design);
4. “Newer tech” is being redefined every 6 months if we follow TSMS’s cadence of coming up with new transistor technology. We certainly cannot expect the same to be qualified for high automotive grade reliability before the next generation hits the fabs.

Although, as things start moving faster (earlier smartphone chips would take 3-4 years for a new version, and are now taped out every 3-4 months), and more reliable methods come up for quickly qualifying processes and chips, we might see shortened cycles even for automotive chips.

[Aroy]

Chips used in automotive industry come between Military Grade and Industrial chips in both reliability as well as temperature range. Though state-of-art is desirable, reliability in the field is the name of the game.

In general both automotive and military chips a quite a few generations behind the consumer chips. Further these chips are packaged in more robust materials than consumer chips, hence the fabrication centres use older generation equipment (though not necessarily older equipment).

So the chip famine cannot be solved that easily by just going to newer fabs.

[wpravin]

Semiconductor companies wouldn't budge much to this surge in demand because they will be aware that majority of automotive chip orders today are a result of panic buying by car makers and few years down the line there wouldn't be much demand so it won't make sense for them to scale-up automotive chip production (in lower technology node) at this point and suffer later when demand takes a nosedive.
Reason for Nosedive - There could be case that a single car maker would place same order to different different semiconductor companies hoping at least one of them would be able to supply at this point and finally when they stock up from one of the semiconductor supplier, cancel the orders places in other semiconductor companies.

[lsjey]

Semiconductor Fab companies are facing huge demand on the higher end of the tech spectrum from Intel, AMD, nVidia, Apple and Samsung. Try getting the latest and greatest CPU/GPUs from AMD or nVidia or Intel at MSRP, and you'll know the demand. Once this demand is met, and only then, the car manufacturer's demand will be addressed. Anyway, car chips will not be generating that much of margin, compared to the higher end chips of Intel or AMD. Also, the demand of car manufacturers will be comparatively very less compared to the demand from the tech companies.

Regards,
lsjey

[GTO]

“I’ll make them as many Intel 16 [nanometer] chips as they want,” Intel chief executive Pat Gelsinger told Fortune last week during his visit to an auto industry trade show in Germany.

Never mind that no carmaker (or even smartphone maker) will consider Intel chips. Everyone saw what happened with "WinTel" and the games Intel played to ensure dominance in the PC segment. No one wants that kind of pricing, and no one wants that kind of lock-in. I'm glad Intel isn't in my smartphone or in my car. I don't want that kind of monopoly in an industry I'm so passionate about.

[banana]

Originally Posted by wpravin

Semiconductor companies wouldn't budge much to this surge in demand because they will be aware that majority of automotive chip orders today are a result of panic buying by car makers and few years down the line there wouldn't be much demand so it won't make sense for them to scale-up automotive chip production (in lower technology node) at this point and suffer later when demand takes a nosedive.
Reason for Nosedive - There could be case that a single car maker would place same order to different different semiconductor companies hoping at least one of them would be able to supply at this point and finally when they stock up from one of the semiconductor supplier, cancel the orders places in other semiconductor companies.

I disagree a little,
iPhone suffered a serious issue when they switched their suppliers. They asked app developers to patch applications with some provided code fragment to allow those apps to run fine with both sensor types.

https://gizmodo.com/heres-why-the-ip...-up-1445966306

Just going with the previous post.

Originally Posted by Aroy

Chips used in automotive industry come between Military Grade and Industrial chips in both reliability as well as temperature range. Though state-of-art is desirable, reliability in the field is the name of the game.

In general both automotive and military chips a quite a few generations behind the consumer chips. Further these chips are packaged in more robust materials than consumer chips, hence the fabrication centres use older generation equipment (though not necessarily older equipment).

So the chip famine cannot be solved that easily by just going to newer fabs.

It became a big problem on a consumer grade phone.
My addition will be, auto manufacturers will either swap out a complete system or will stick to the tried and tested. If changing one chip requires a whole lot of real world testing, they'll mostly avoid it.

Just my opinion

[wpravin]

Originally Posted by banana

I disagree a little,
iPhone suffered a serious issue when they switched their suppliers. They asked app developers to patch applications with some provided code fragment to allow those apps to run fine with both sensor types.

https://gizmodo.com/heres-why-the-ip...-up-1445966306

Just going with the previous post.

iPhone's software is tightly coupled with the hardware. That meaning they write software/firmware specific to the underlying hardware because of which they could extract maximum performance out of the hardware which is not the case with other phone makers. End of the day Apple started making their own chips.
For car makers, most of semiconductor requirement would be generic functional modules for which they can switch suppliers without major hiccups.

[ecenandu]

Originally Posted by adroit_91

As someone from a semiconductor background, and having worked a couple of years for automotive chipsets (Advanced Driver Assistance & Safety — ADAS, as well as entertainment), here’s my 2 cents:

And also the software porting and testing overheads if they wanted to move to newer chips.

Originally Posted by wpravin

iPhone's software is tightly coupled with the hardware. That meaning they write software/firmware specific to the underlying hardware because of which they could extract maximum performance out of the hardware which is not the case with other phone makers. End of the day Apple started making their own chips.
For car makers, most of semiconductor requirement would be generic functional modules for which they can switch suppliers without major hiccups.

The vertical integration that you are trying to highlight of Apple is correct, but the quoted article is saying something different, it is about the accelerometers. And the catch in this particular situation was that Apple didn't write specific software for the new accelerometer.

[kedar_GT]

As some members have pointed out, automotive-grade chips undergo a lot more stringent testing and qualification. Most OEM test these @ extreme temperatures in the polar region as well as deserts.
Due to this, the chip manufacturing process for car electronics is in the 40nm and 28nm technology nodes as reliability is of utmost importance.
A fab is designed for a specific technology node and the supply is also coupled with the demand - so repurposing one just to take care of this demand uptick may not be possible without huge additional investments.

To add, the semiconductor chips used in today's cellphones range from 16nm to 7 & 5nm processes as area and power consumption are of utmost importance as opposed to long term reliability per se.

[poloman]

There are no Intel chips which are going to be developed for automobile industry. Pat is talking about the usage and repurposing of Intel 14nm fabs for manufacturing automobile chips.

There are only 3 large chip manufacturers or fabs in the world currently. Intel, Samsung and TSMC. TSMC is overloaded with orders for advanced computer and smartphone chips < =7nm process nodes mostly from Apple, AMD, Nvidia Qualcomm etc. Same with Samsung. Intel so far was using its fabs mostly for inhouse processors. But since it has moved on to sub 10nm process nodes, idle capacity is available in 14nm fabs which Intel is asking automobile industry to make use of.

But as mentioned in the article, automakers mostly used mixed signal chips and can easily work with 28nm or above process nodes. So Intel is asking them to move to more advanced process so that the existing 14nm fabs can be repurposed. But automakers want new fabs with older process nodes which Intel is not willing to put capex on.

[aashishnb]

I think there is a general tendency in the industrial type applications world to stick to tried & tested systems and not bother with upgrading just because an upgrade is available - 'why fix if something is not broken'. The preference is for systems to be backward/forward compatible and reduce the number of parts/SKUs.

Here are a few examples:

1. There was some news a few years ago that said NASA still uses the old 486/pentium chips and was looking to buy a lot of obsolete/outdated computers.
2. When Hyundai launched Santro, if I remember correctly, they marketed it with '16-bit computer onboard'. I'm sure, even at that time, there were much more advanced chips available, but maybe they did not need any of the fancy hardware for ECU,etc.
3. A similar thing occurs in the POS terminals that you would see in retail outlets (the one that is used for billing). My experience, from discussions for an F&B chain a few years ago - they were comparatively outdated configurations for that time as compared to the normal computers available in the market, had lesser functionality, cost almost 2x. When I asked the vendor for reason, they said the machines are built for robust (abuse-friendly) use. Moreover, they will support the systems for many years to come, which the full fledged computers will not be able to do. Also, old/new machines will all be compatible with the system.

[gauravanekar]

I would say it is the lack of foresight and planning on part of automotive manufacturers, tesla has been showing them how to do it since a long time now yet we have so many different computational centres within the same car, the least they could have done is consolidate all computational power in a car into safety critical and non critical centres. Also consolidate chips among all the models product lines. With such consolidation of part numbers they could have had a better hand while negotiating with chip makers.
OT: similar confused strategy is being seen on part of traditional auto cos when it comes to EV cars as well.

[Aroy]

Originally Posted by gauravanekar

I would say it is the lack of foresight and planning on part of automotive manufacturers, tesla has been showing them how to do it since a long time now yet we have so many different computational centres within the same car, the least they could have done is consolidate all computational power in a car into safety critical and non critical centres. Also consolidate chips among all the models product lines. With such consolidation of part numbers they could have had a better hand while negotiating with chip makers.
OT: similar confused strategy is being seen on part of traditional auto cos when it comes to EV cars as well.

Blindly introducing new technology has its own pitfalls. How many users have faced European cars "freezing" at the most inappropriate time? or the many problems faced with "Mechatronics" of advanced automatic gear box?

This is the reason why automakers stick to tried and tested technology. After all the computation power required is minuscule compared to phones, let alone computers. What they need is reliability, reliability and reliability under the wide range of temperatures, vibration and dirty environment. In short the electronics needs to be as reliable as the mechanical components after all when the vehicle stops it does not matter to the user why it did, just that it stopped.

[banana]

Not sure if this is the right thread [please move to a more suitable thread if applicable]

Auto chip shortage shows sign of easing
Top suppliers see first inventory uptick in nine months, but demand remains high

source: https://twitter.com/i/events/1465400038125756418