4 ms·

50 years? Isn't it a little too much? Maybe 10? 20 even. But 50?

Aromasin 5y ago
Depends on the use case but yes I'd agree, 50 might be a tad too long. I work in IoT and industrial computing and the specifications can be for 25+ years however, so it's not unheard of.
- tw04 5y ago
  >I work in IoT and industrial computing and the specifications can be for 25+ years however, so it's not unheard of. 25+ years isn't unheard of, but "in any capacity" absolutely is. There's no foundry on this planet that is continuing to produce you CPUs 25 years from now if your batch size is an order of 50 units.
  - zozbot234 5y ago
    There's plenty of foundries cheaply fabbing chips in "obsolete" process nodes. If the original design from 25+ years ago is available, there's no reason why it couldn't be fabbed today in those quantities.
    - rbanffy 5y ago
      Also, while I wouldn't want to redesign (and, Heaven forbid, recertify it with the authorities) the board and software, I wouldn't mind having a part done in a more power efficient way, as long as it's pin-compatible with the old one and done in a way similar enough no recertification is needed.
      - sitkack 5y ago
        This is doable, it is impossible right now, but given enough time, quite probable that one could take an old design, and move it to a new PDK and have it automatically adjusted to meet the same timing and power constraints. If only one could take the GDSII for an old process node and just make it smaller. Maybe someone will make a model that does that, just contracts images of polygons. :) It would require not only the tooling but also past test runs of all of the verification passes to confirm that the new design doesn't have a flaw.
  - reportingsjr 5y ago
    Rochester Electronics disagrees with you! You just have to have some deeeeep pockets for it.
smolder 5y ago
I like this comment. It's an interesting question. It's understandable why we wouldn't want to use the same CPUs for 50 years or more given their history of rapid improvement, but on the other hand, engineering things to "heirloom quality" is important for sustained growth and for sustainability in general.
tablespoon 5y ago
> 50 years? Isn't it a little too much? Maybe 10? 20 even. But 50? There are a lot of products that don't change on tech industry timescales (e.g. all those stories of organizations scavenging replacement chips from eBay). It could be really compelling to have the option to use components that are likely be available for 50 years.
jecel 5y ago
The 6502 will be 50 in 2025, so unless they stop making them between now and then we will have at least one example of a processor lasting that long commercially. If not, then the Z80 will be 50 in the following year.
- johnwalkr 5y ago
  But those are running lightweight embedded stuff with no or minimal OS. A modern intel x86 CPU is never used that way and a) is broadly compatible with other x86 cpus and b) is used in products that rely on a million software dependencies and a supported OS, along with all of the updates along the way to keep it working online which is a standard requirement these days. So unless the system is totally offline, all the mentioned stuff, plus the other spare parts are also supported for 50 years, there’s not much of a market there. But by all means the embedded stuff, motor controllers, instruments and so on should have simple interfaces and either last 50 years or be easily replaced 50 years in the future. This is largely true for industrial stuff. In industrial plants you rarely even have a serial connection to instruments. Instead you will have something like a 4-20mA analogue signal.
  - filereaper 5y ago
    Maybe the OS and everything we run today will be considered "lightweight" 50 years from now...
    - johnwalkr 5y ago
      Maybe, but in that case it will be lightweight on current hardware, not on 50 year old hardware. I wouldn't be surprised if a midrange microcontroller used for a one-off application, in 50 years has the equivalent of docker (x2, one for the development environment and one for the microcontroller). That would hopefully work for another 50 years. On the other hand, look at any small microcontroller vs a complex CPU today. In laymen's terms, arduino vs raspberry pi. For controlling some devices, a small microcontroller is faster booting, faster to react, has fewer unnecessary states, is deterministic, testable, more reliable, uses less power. That will probably never change and in 50 years 8 bit micros will still be used for a lot of applications. Even after 50 years of development you probably don't want the brake controller in your self-driving car to run something in python 2.7.19 in docker in windows 39. I also wouldn't be surprised (and hope) it goes the other way and in 50 years development using FPGAs is as easy banging stuff out in an IDE and making ASICs is as easy compiling and dollars per chip in small quantities.

Depends on the use case but yes I'd agree, 50 might be a tad too long. I work in IoT and industrial computing and the specifications can be for 25+ years however, so it's not unheard of.

tw04 5y ago
>I work in IoT and industrial computing and the specifications can be for 25+ years however, so it's not unheard of. 25+ years isn't unheard of, but "in any capacity" absolutely is. There's no foundry on this planet that is continuing to produce you CPUs 25 years from now if your batch size is an order of 50 units.
- zozbot234 5y ago
  There's plenty of foundries cheaply fabbing chips in "obsolete" process nodes. If the original design from 25+ years ago is available, there's no reason why it couldn't be fabbed today in those quantities.
  - rbanffy 5y ago
    Also, while I wouldn't want to redesign (and, Heaven forbid, recertify it with the authorities) the board and software, I wouldn't mind having a part done in a more power efficient way, as long as it's pin-compatible with the old one and done in a way similar enough no recertification is needed.
    - sitkack 5y ago
      This is doable, it is impossible right now, but given enough time, quite probable that one could take an old design, and move it to a new PDK and have it automatically adjusted to meet the same timing and power constraints. If only one could take the GDSII for an old process node and just make it smaller. Maybe someone will make a model that does that, just contracts images of polygons. :) It would require not only the tooling but also past test runs of all of the verification passes to confirm that the new design doesn't have a flaw.
- reportingsjr 5y ago
  Rochester Electronics disagrees with you! You just have to have some deeeeep pockets for it.

smolder 5y ago

I like this comment. It's an interesting question. It's understandable why we wouldn't want to use the same CPUs for 50 years or more given their history of rapid improvement, but on the other hand, engineering things to "heirloom quality" is important for sustained growth and for sustainability in general.

tablespoon 5y ago

> 50 years? Isn't it a little too much? Maybe 10? 20 even. But 50? There are a lot of products that don't change on tech industry timescales (e.g. all those stories of organizations scavenging replacement chips from eBay). It could be really compelling to have the option to use components that are likely be available for 50 years.

jecel 5y ago

The 6502 will be 50 in 2025, so unless they stop making them between now and then we will have at least one example of a processor lasting that long commercially. If not, then the Z80 will be 50 in the following year.

johnwalkr 5y ago
But those are running lightweight embedded stuff with no or minimal OS. A modern intel x86 CPU is never used that way and a) is broadly compatible with other x86 cpus and b) is used in products that rely on a million software dependencies and a supported OS, along with all of the updates along the way to keep it working online which is a standard requirement these days. So unless the system is totally offline, all the mentioned stuff, plus the other spare parts are also supported for 50 years, there’s not much of a market there. But by all means the embedded stuff, motor controllers, instruments and so on should have simple interfaces and either last 50 years or be easily replaced 50 years in the future. This is largely true for industrial stuff. In industrial plants you rarely even have a serial connection to instruments. Instead you will have something like a 4-20mA analogue signal.
- filereaper 5y ago
  Maybe the OS and everything we run today will be considered "lightweight" 50 years from now...
  - johnwalkr 5y ago
    Maybe, but in that case it will be lightweight on current hardware, not on 50 year old hardware. I wouldn't be surprised if a midrange microcontroller used for a one-off application, in 50 years has the equivalent of docker (x2, one for the development environment and one for the microcontroller). That would hopefully work for another 50 years. On the other hand, look at any small microcontroller vs a complex CPU today. In laymen's terms, arduino vs raspberry pi. For controlling some devices, a small microcontroller is faster booting, faster to react, has fewer unnecessary states, is deterministic, testable, more reliable, uses less power. That will probably never change and in 50 years 8 bit micros will still be used for a lot of applications. Even after 50 years of development you probably don't want the brake controller in your self-driving car to run something in python 2.7.19 in docker in windows 39. I also wouldn't be surprised (and hope) it goes the other way and in 50 years development using FPGAs is as easy banging stuff out in an IDE and making ASICs is as easy compiling and dollars per chip in small quantities.