Ontopic The new car-seching thread

[HipHugHer]

Yes. That’s why the recall recommended parking outside (and later to park 50 ft away from anything) until they replace the batteries

Where in a typical city or suburb are you 50 feet away from "anything"?
You could be in the middle of rural Montana and not be 50 feet away from a few million blades of grass that'll burn real easy.

 

[Jehannum]

Out at the fair today, letting kids fiddle with the GTO.

 

[Immigrant]

LOVE that color on a GTO. It’s the perfect color. I don’t know why because I don’t really like it, or wouldn’t want it on anything else, but it seems to fit GTOs (and Tempests, etc I guess).

 

[fly]

So there ends my foray into EV driving (so far). I bought my 2020 Bolt last October for $25,200 (after lots of incentives and after TTL). I loved driving my car, but shortly after GM recalled the batteries that were made in Korea. They were starting fires from overcharging. My car’s battery, however, was build in MI so I was originally out of the recall. But last month, GM expanded the recall to all Bolts, even the brand new ones of the redesign they did. But because of the global manufacturing and electronics shortages, and how long it took them to identify the issues, they haven’t started making new batteries yet to replace, and it would be a full replace. The recommendations were to limit charging to 90% max, and drain only to 70 miles left (30%) making the effective range very short. They also recommend not to park in a garage and recent recommendations were to park 50 ft away from anything and at the top of car parks. A lot of parking lots are starting to ban the car actually.

Since I didn’t know when I would get a replacement battery, I went to go bo a buyback. That was going to take a month to get as well. Just on A whim I went to Texas Direct Auto and they offered $22k. Not bad, but I went to Carmax and they offered $24k so I took it.

Was I too stupid for taking that deal? Technically the car cost me $1200 to run it for almost a year. Should I have waited for GM’s price? But then there are reports that places like Carvana are becoming saturated with Bolts and they aren’t offering more than $18K because they can’t shift them. I think I got lucky getting $24k.

That sucks. Yeah, the car market is its own dumpster fire right now.

I was reading yesterday that the car chip shortage is because all the car manufacturers use OOOOOOLD chips. Think 90nm chips from the very first iPhone. Apparently none of the fabs want to build new lines for this shit, because they know that the manufacturers will eventually switch.

Oops

 

[gee]

90mm isn't old, and there's piles of advantages (leakage current, ESD/voltage spike tolerance, voltage range, blah blah blah) to building certain types of parts on bigger processes. Like a CANBUS transciever built on a 7nm process would probably last 5 seconds in an automotive environment.

Little transistors are delicate little things. Big 'ol transistors can take a beating.

 

[Jehannum]

LOVE that color on a GTO. It’s the perfect color. I don’t know why because I don’t really like it, or wouldn’t want it on anything else, but it seems to fit GTOs (and Tempests, etc I guess).

It's GM's "F" code blue, which Pontiac sold as Tyrol blue. It was pretty common on GMs of the time, but I agree, it suits this car pretty damn well.

 

[fly]

90mm isn't old, and there's piles of advantages (leakage current, ESD/voltage spike tolerance, voltage range, blah blah blah) to building certain types of parts on bigger processes. Like a CANBUS transciever built on a 7nm process would probably last 5 seconds in an automotive environment.

Little transistors are delicate little things. Big 'ol transistors can take a beating.

Maybe it wasn't 90nm then, but whatever was in the original iPhone. According to the article, it's because these chips "work".

edit: Found the article https://fortune.com/2021/09/17/chip-makers-carmakers-time-get-out-semiconductor-stone-age/

 

[Jehannum]

Maybe it wasn't 90nm then, but whatever was in the original iPhone. According to the article, it's because these chips "work".

edit: Found the article https://fortune.com/2021/09/17/chip-makers-carmakers-time-get-out-semiconductor-stone-age/

IDK, it seems like @gee's comment is still pretty valid.

That article read like the dinguses who used to say we ought to just "upgrade" the shuttle's systems with modern x86 hardware instead of the proven 8086 chips, y'know, "because".

 

[fly]

IDK, it seems like @gee's comment is still pretty valid.

That article read like the dinguses who used to say we ought to just "upgrade" the shuttle's systems with modern x86 hardware instead of the proven 8086 chips, y'know, "because".

I can't imagine that any of these problems are insurmountable.

 

[Jehannum]

I can't imagine that any of these problems are insurmountable.

I can't imagine introducing problems on purpose when you have a system that works.

 

[Mr. Argumentor]

I can't imagine introducing problems on purpose when you have a system that works.

This. Whats the advantage here, other than production lines?

 

[Domon]

Maybe it wasn't 90nm then, but whatever was in the original iPhone. According to the article, it's because these chips "work".

edit: Found the article https://fortune.com/2021/09/17/chip-makers-carmakers-time-get-out-semiconductor-stone-age/

Theres significant reliability advantage to older node chips for industrial and high reliability applications that dont require performance. The feature size being bigger alleviated a lot of reliability concerns in smaller nodes, although them running hotter cancels out some of those. For critical applications, theres also the cosmic ray aspect to consider. You dont want muons flipping key bits in the giant hunk o metal going 80mph.

That said, 90 really is truly ancient these days. I expect they could drop to 32, which is still pretty damn old and thats robust enough to not having any of the issues i mentioned, besides a slightly higher rate of particle effects. There are certain nodes that are inherently rad-hard, and certain 32 fabs are one of them. (90 is another)

 

[gee]

I can't imagine that any of these problems are insurmountable.

Economics. There's a pretty big sunk cost in redesigning something like an engine computer to do a major architectural change. That's high volume, high reliability shit that's done by large teams of people and tested to fuck and if you're amortizing that over only a year's worth of cars, you're wasting a fuckton of money.

Timeframe. It takes us years to bring shit to market where I work, and we're a small shop doing relatively simple shit in comparison. I'd imagine developing a new model ECU takes many years. By the time the design of a new temporary ECU is qualified and production ready, a chip shortage could be over and done with and the thing might never go into production anyway.

Plus there's scheduling. You're pulling together an engineering team to do that job, that would ordinarily be working on the engine computer for the next model, so you're potentially fucking over the release date of a new model car. Retiring one car, using up the remaining parts to build the last of them, changing over the production line, starting up the new model etc is all planned years in advance and "hey lets make the ECU team design this other thing that'll only get used for a year" could toss a wrench in that whole thing.

Generally auto manufacturers just used their corporate might and long term contracts to buy parts and shortages were never an issue. GM/Ford/whoever signs a big contract with Freescale or NXP or whoever, they set aside the fab capacity and they're guaranteed parts, and it's always worked. It's only in recent years when semiconductor companies have merged/gone fabless/whatever, and now you've got multiple chip manufacturers fighting for capacity at foundries, that this shit has even been an issue.

Part of my job is keeping old designs going. We buy thousands of chips when they go EOL so we can keep making the old board. Occasionally we'll redesign a board to change 1-2 parts but we're still keeping the engineering time at an absolute minimum - why? because we're busy designing that whole product's replacement and when we've got that done, the old product fucks off and its old chips and their availability issues are never a problem again.

 

[Jehannum]

Theres significant reliability advantage to older node chips for industrial and high reliability applications that dont require performance. The feature size being bigger alleviated a lot of reliability concerns in smaller nodes, although them running hotter cancels out some of those. For critical applications, theres also the cosmic ray aspect to consider. You dont want muons flipping key bits in the giant hunk o metal going 80mph.

YEAH, BUT NEWER CHIPS ARE SMALLER AND FASTER!

</Pakled>

 

[Domon]

YEAH, BUT NEWER CHIPS ARE SMALLER AND FASTER!

</Pakled>

I need access to your set of Lower Decks gifs. Theyre all amazing.

This is still my favorite:

 

[Domon]

Economics. There's a pretty big sunk cost in redesigning something like an engine computer to do a major architectural change. That's high volume, high reliability shit that's done by large teams of people and tested to fuck and if you're amortizing that over only a year's worth of cars, you're wasting a fuckton of money.

Timeframe. It takes us years to bring shit to market where I work, and we're a small shop doing relatively simple shit in comparison. I'd imagine developing a new model ECU takes many years. By the time the design of a new temporary ECU is qualified and production ready, a chip shortage could be over and done with and the thing might never go into production anyway.

Plus there's scheduling. You're pulling together an engineering team to do that job, that would ordinarily be working on the engine computer for the next model, so you're potentially fucking over the release date of a new model car. Retiring one car, using up the remaining parts to build the last of them, changing over the production line, starting up the new model etc is all planned years in advance and "hey lets make the ECU team design this other thing that'll only get used for a year" could toss a wrench in that whole thing.

Generally auto manufacturers just used their corporate might and long term contracts to buy parts and shortages were never an issue. GM/Ford/whoever signs a big contract with Freescale or NXP or whoever, they set aside the fab capacity and they're guaranteed parts, and it's always worked. It's only in recent years when semiconductor companies have merged/gone fabless/whatever, and now you've got multiple chip manufacturers fighting for capacity at foundries, that this shit has even been an issue.

Part of my job is keeping old designs going. We buy thousands of chips when they go EOL so we can keep making the old board. Occasionally we'll redesign a board to change 1-2 parts but we're still keeping the engineering time at an absolute minimum - why? because we're busy designing that whole product's replacement and when we've got that done, the old product fucks off and its old chips and their availability issues are never a problem again.

A counterargument would be Tesla. All their shit is from scratch, and really fucking new and so far so good from the chip reliability standpoint.

 

[Jehannum]

A counterargument would be Tesla. All their shit is from scratch, and really fucking new and so far so good from the chip reliability standpoint.

wait a minute, the same teslas that were bricking because they exceeded the maximum write cycles for some flash memory onboard?

 

[Domon]

Economics. There's a pretty big sunk cost in redesigning something like an engine computer to do a major architectural change. That's high volume, high reliability shit that's done by large teams of people and tested to fuck and if you're amortizing that over only a year's worth of cars, you're wasting a fuckton of money.

Timeframe. It takes us years to bring shit to market where I work, and we're a small shop doing relatively simple shit in comparison. I'd imagine developing a new model ECU takes many years. By the time the design of a new temporary ECU is qualified and production ready, a chip shortage could be over and done with and the thing might never go into production anyway.

Plus there's scheduling. You're pulling together an engineering team to do that job, that would ordinarily be working on the engine computer for the next model, so you're potentially fucking over the release date of a new model car. Retiring one car, using up the remaining parts to build the last of them, changing over the production line, starting up the new model etc is all planned years in advance and "hey lets make the ECU team design this other thing that'll only get used for a year" could toss a wrench in that whole thing.

Generally auto manufacturers just used their corporate might and long term contracts to buy parts and shortages were never an issue. GM/Ford/whoever signs a big contract with Freescale or NXP or whoever, they set aside the fab capacity and they're guaranteed parts, and it's always worked. It's only in recent years when semiconductor companies have merged/gone fabless/whatever, and now you've got multiple chip manufacturers fighting for capacity at foundries, that this shit has even been an issue.

Part of my job is keeping old designs going. We buy thousands of chips when they go EOL so we can keep making the old board. Occasionally we'll redesign a board to change 1-2 parts but we're still keeping the engineering time at an absolute minimum - why? because we're busy designing that whole product's replacement and when we've got that done, the old product fucks off and its old chips and their availability issues are never a problem again.

man, lifetime buys suck from a finance perspective though. Thats always hard to justify the budgetary need to people that dont understand.

 

[Domon]

wait a minute, the same teslas that were bricking because they exceeded the maximum write cycles for some flash memory onboard?

that one, i dont recall. But google says ... yes

Wow, and their solution to it was super-dumb too "In the summer of 2020, Tesla introduced an MCU with a 64GB eMMC chip instead of 8GB"

So same wear write problem, just way more slack to eat up before it dies, rather than switching to a better technology

double edit: wow, that is some spectaculary shitty NAND that only has a 24tb lifetime write.

 

[gee]

man, lifetime buys suck from a finance perspective though. Thats always hard to justify the budgetary need to people that dont understand.

Lifetime buys suck when you've got management who are sucking the lean manufacturing dick and are like "you seriously need to buy trays of 1000 USB controllers? that we're just going to keep in the parts cage for a couple of years? that's not very lean/kanban/whatever-other-buddhist-bullshit-term!"

We retired a product a short while ago that had an MCORE processor in it. Good riddance to that shit.