> Which ones scale exponentially? How come a Fairphone isn't dramatically more expensive than an iPhone?
Nothing that they've chosen. The Fairphone isn't dramatically more expensive than other phones because it really isn't dramatically different. They don't have any modular parts that aren't already available on other phones in various combinations. They are just offering a slightly different box of COTS parts with a promise to support it. This is relatively cheap to do. It still suffers from some of the problems you mention and more: the logic board is all soldered, the modem is not modular, the software support promises are reliant on third parties, etc.
Meanwhile many PC laptops do make the upgradeable parts modular. Laptops with modular memory, CPU, wireless etc. are available for less than a Macbook or iPhone. So where is the exponential cost?
> the software support promises are reliant on third parties
This is a "current vendors are crap" problem rather than any kind of technological barrier.
And that can't be the reason that Apple can't do it.
It's a few cents, and it makes packaging more difficult. If Samsung shaves a single penny off of each phone they sell, they are more a million dollars more profitable per year.
But most mid-to-high end phones do have a significant number of modular parts. Battery, camera, speakers, displays, and sometimes IO are modular on many phones, and repair shops will fix these for customers.
> Meanwhile many PC laptops do make the upgradeable parts modular. Laptops with modular memory, CPU, wireless etc. are available for less than a Macbook or iPhone.
Modular CPUs are generally not available on laptops anymore, that's a thing of the past anymore. Modular memory and wireless are starting to disappear, in part due to latency requirements for the later memory standards, and packaging and power requirements for popular thin-and-light segment devices.
If you're asking why phones don't have modular memory, the answer is simple: they're optimized for small size and low power, and adding a connector would compromise on that. Also, it would require inventing a new memory standard since modular LPDDR does not exist.
> So where is the exponential cost?
You see exponential costs if you look at devices that actually have extended lifetimes. I'm not talking about one consumer phone compared to another. In the scheme of things, they really all have expected lifetimes around a handful of years. Compare the redundancy and expected lifetime of consumer products to industrial products to aerospace products if you want to see how making something last a long time can be very expensive. If you want to double a product's lifetime, you are often looking at 10x the cost, if not more.
> This is a "current vendors are crap" problem rather than any kind of technological barrier.
Well, it's not possible to make a phone without relying on vendors for some parts. There are too many specialized parts. Not even Apple has enough resources to in-house it all, and they have been trying.
Or they could just raise the price by five cents and make even more money per unit and get more sales because more people are attracted by the upgradeable device than are deterred by a 0.01% difference in price on a $500+ device.
> But most mid-to-high end phones do have a significant number of modular parts. Battery, camera, speakers, displays, and sometimes IO are modular on many phones, and repair shops will fix these for customers.
Modular in the sense that they're intrinsically separate components you have to go out of your way to stick together, not in the sense that they're attached with standard modular connectors and are easy to replace.
> Modular CPUs are generally not available on laptops anymore, that's a thing of the past anymore.
This is basically down to Intel changing their CPU socket every ten seconds, not any actual cost reason. Nobody cares if you can replace the CPU if the upgrade isn't compatible with the socket. The laptops that previously had this didn't have a significantly higher cost.
> Modular memory and wireless are starting to disappear, in part due to latency requirements for the later memory standards, and packaging and power requirements for popular thin-and-light segment devices.
Dell has sorted this and it's on track to become the standard:
https://www.storagereview.com/review/dell-camm-dram-the-new-...
> If you're asking why phones don't have modular memory, the answer is simple: they're optimized for small size and low power, and adding a connector would compromise on that. Also, it would require inventing a new memory standard since modular LPDDR does not exist.
So invent a new standard. All new standards are invented by somebody; see above. If Dell can do it, why not Samsung?
> You see exponential costs if you look at devices that actually have extended lifetimes. I'm not talking about one consumer phone compared to another. In the scheme of things, they really all have expected lifetimes around a handful of years.
You can run the latest version of the Linux kernel on commodity PCs from the 90s. Why can't you do that on a phone half that old? None of the reasons are good.
> Compare the redundancy and expected lifetime of consumer products to industrial products to aerospace products if you want to see how making something last a long time can be very expensive. If you want to double a product's lifetime, you are often looking at 10x the cost, if not more.
These products are expensive because they're purchased primarily by bureaucracies and everything purchased primarily by bureaucracies is cost inflated. Which part of extending the lifetime is the part which is supposed to make it so much more expensive?
> Well, it's not possible to make a phone without relying on vendors for some parts. There are too many specialized parts. Not even Apple has enough resources to in-house it all, and they have been trying.
You don't have to make your own DRAM or display because those were never the issue. The lack of support comes from the SoC whose vendor has published neither driver source code nor documentation sufficient to write one, so when the kernel they gave you a binary blob to go with falls out of support you're screwed.
Qualcomm has been a major offender here and is one of the only suppliers of performant SoCs for phones but they're supposedly getting better. Samsung has no excuse because they make their own chips.
There is a lot of excitement around RISC-V because it could potentially do something about this, if not by actually producing a competitive chip then by putting pressure on the incumbents to get it together and publish open source drivers.
You're arguing against the specialization of labor. I don't think this is a winning argument.
Labor specialization is supposed to make things more efficient. You have some expert who can do the job in five minutes that would take you an hour, and then you can justify paying them ten times your own hourly rate for those five minutes.
It doesn't do any good to cause something to take the high-paid expert an hour instead of five minutes, all it does send the device to the landfill because the repair becomes uneconomical.
Doing this raises the labor cost to the point of making it uneconomical in many cases that it wouldn't be otherwise.
It also raises the parts cost (and the amount of waste), because a failed logic board that has every chip in the device soldered to it is going to cost more to replace than a card with only the bit that failed on it.
> Some of the costs of building and/or repairing a device scale linearly, and some scale exponentially.
Which ones scale exponentially? How come a Fairphone isn't dramatically more expensive than an iPhone?
> And they tend to depreciate logarithmically.
Depreciation on existing devices is precisely because they can't be upgraded or repaired. If your phone has a 3G cellular modem in it, it stops being useful as a phone when they shut down the 3G towers -- unless you can replace the modem.
Compare this to PC components that don't depreciate quickly, like a standard ATX chassis or power supply or monitor. Or an AM4 system board that could have been purchased in 2016 yet supports 2022 CPUs that could viably still be in use for five or ten more years.
And a decline in value of a component only makes the repairs more affordable. A Ryzen 7 2700X has lost half its value over 5 years, which only means that if you should need to replace one it costs half as much -- or can be replaced with a faster model in the same system.