Update efficiency

[BillyBurke]

When will the computational power of Apple's M chips reach the point that a System Update takes seconds rather than minutes to update? Are there more components at play than raw computational power?

It seems to me we should be at that stage now, but we obviously aren't. I have an M3 iMac. What am I missing?

 

[Bigwaff]

What am I missing?

A lot... that the update package(s) needs to be written to storage, package(s) decompressed, installation scripts executed which write files, system files encrypted and sealed, the encryption verified, and so on.. etc... etc. Sometimes there are firmware updates included. It's not about computational power.

 

[BillyBurke]

A lot... that the update package(s) needs to be written to storage, package(s) decompressed, installation scripts executed which write files, system files encrypted and sealed, the encryption verified, and so on.. etc... etc. Sometimes there are firmware updates included. It's not about computational power.

I apologize for my ignorance here, but aren't all those tasks simply (!!) instructions to be carried out? Isn't everything related to computational power? I would think the latest Apple Silicone would do all those tasks incrementally faster with each chip improvement. But... that doesn't seem so.

 

[Bigwaff]

I apologize for my ignorance here, but aren't all those tasks simply (!!) instructions to be carried out? Isn't everything related to computational power? I would think the latest Apple Silicone would do all those tasks incrementally faster with each chip improvement. But... that doesn't seem so.

Automobiles have CPU's in them. Why doesn't a car go 1K miles an hour?

 

[BillyBurke]

Well, it seems obvious to me that the CPU in a computer is the engine of the primary function of the computer while the CPU in a car is to support some function other than the primary mission of the car , i.e., travel from A to B. Thanks for your thoughts anyway.

 

[throAU]

When will the computational power of Apple's M chips reach the point that a System Update takes seconds rather than minutes to update? Are there more components at play than raw computational power?

It seems to me we should be at that stage now, but we obviously aren't. I have an M3 iMac. What am I missing?

Never, software expands to consume available resources.

To be fair, I suspect one of the biggest time consumers of apple update is re-sealing the sealed system volume. Basically it needs to re-checksum every file on the system disk and record that so that files can be confirmed not tampered with by malware.

That's more a storage speed issue.

 

[BillyBurke]

Thank you. Storage Speed issue makes a bit more sense.

 

[throAU]

For those curious, a detailed explanation of volume sealing is here (under system data integrity):

GitHub - AsahiLinux/docs: Asahi Linux documentation

Asahi Linux documentation. Contribute to AsahiLinux/docs development by creating an account on GitHub.

github.com

short version in English - whenever system files change on the protected system volume (e.g., an update), the entire volume "seal" needs to be re-calculated.

Sounds like a lot of BS to prevent a machine from booting if it goes sideways, but it ensures that you can't accidentally boot a maliciously modified OS (e.g., one infected with malware).

 

[leman]

I think you might be underestimating the complexity of the task. While these newer machines are much, much faster at processing data, there is still quite a lot to do. Decompressing and verifying multiple gigabytes of patches is a lot of work, and system updates are much more involved than just copying some data. Not to mention that a lot of the install process does not depend on the CPU speed — there are hardware tests do be done, data written to the storage, internal encryption systems to engage... Over the years, Apple did a lot of foundational work to make updates faster and more secure, much of the improvements come from the system design (avoiding work) rather than making the hardware faster.

Is it potentially possible to do a system update in seconds? I do believe so. It would require a system that does the update in the background, while you are working with the computer, and then quickly "activates" the new system on restart. Internal tests are still a challenge though.

 

[Bigwaff]

Well, it seems obvious to me that the CPU in a computer is the engine of the primary function of the computer while the CPU in a car is to support some function other than the primary mission of the car , i.e., travel from A to B. Thanks for your thoughts anyway.

Storage IO is still magnitudes slower than RAM or CPU data caches. CPU gots to get the data it processes from somewhere. Your CPU cores live a long, starved, and exceedingly idle life. Nothing can feed them data fast enough. Everything they depend on fails far sooner than they do.

 

[MRMSFC]

When will the computational power of Apple's M chips reach the point that a System Update takes seconds rather than minutes to update? Are there more components at play than raw computational power?

It seems to me we should be at that stage now, but we obviously aren't. I have an M3 iMac. What am I missing?

One of the fundamental laws of computing is that for every advancement in processing power there is an equal or greater advancement in software bloat.

 

[haralds]

Updates are designed to be staged in the background.

 

[BillyBurke]

I think you might be underestimating the complexity of the task. While these newer machines are much, much faster at processing data, there is still quite a lot to do. Decompressing and verifying multiple gigabytes of patches is a lot of work, and system updates are much more involved than just copying some data. Not to mention that a lot of the install process does not depend on the CPU speed — there are hardware tests do be done, data written to the storage, internal encryption systems to engage... Over the years, Apple did a lot of foundational work to make updates faster and more secure, much of the improvements come from the system design (avoiding work) rather than making the hardware faster.

Is it potentially possible to do a system update in seconds? I do believe so. It would require a system that does the update in the background, while you are working with the computer, and then quickly "activates" the new system on restart. Internal tests are still a challenge though.

Thank you for this very succinct explanation. I like the "background" update concept. Maybe one of these days.