[WheelsAtLarge]

Right now, Intel produces multiple types of CPUs by having one die and modifying it at will to produce different capabilities or using chips that have some kind of fault that reduces its capabilities but are good enough to function at lower speeds or functions.

The subscription model is better since customers don't have to buy more CPU than what they need plus should they change their mind they can just upgrade. In time, all CPU makers will go with the subscription model since it makes sense in terms of manufacturing, customer satisfaction and company profits.

Environmentally, it also makes sense. There won't be a need to dump your current hardware just because you want to upgrade.

Note: I meant pay-as-you-go, not subscription.

[thesuperbigfrog]

>> The subscription model is better since customers don't have to buy more CPU than what they need plus should they change their mind they can just upgrade.

On the contrary, with the subscription model, customers must continue to buy the CPU they need and how much they need to pay for it is subject to Intel's whims.

>> In time, all CPU makers will go with the subscription model since it makes sense in terms of manufacturing, customer satisfaction and company profits.

Then we need open hardware. I will not buy hardware that holds my computing hostage.

[hakfoo]

I find the environmental angle pretty hard to buy.

The ideal customer for this sort of scheme is "I'll buy a 16-core today, and if my business needs grow substantially in two years, I'll unlock it to 24 cores."

This would, under the current scheme, involve opening the case and swapping a CPU. I don't think this happens for any number of reasons:

* Service contracts (and associated obscene costs for manufacturer upgrades) discouraging people from DIY upgrades. Maybe selling it as a service reduces some recalctriance there, but it requires absolute 100% "this will not void your warranty" reassurances.

* Upgrade needs outside the CPU (i. e. new networking or drive interfaces)

The logistics are also sort of ugly. Even if you find a willing customer, the ceiling of the upgrade is presumably the top of the line of that generation. After a few years, this might not be compelling, without some very aggressive pricing. In desktop terms, if you had a Ryzen 1300 today, how much would you be willing to pay to "unlock" it to a 1800X knowing that there are far better 5000 series parts available for cheap?

The idea of "sell everyone the top CPU, and fuse it down in software" also assumes high yields on top-end parts, something Intel has had problems with in recent history. If you're only producing, say, 10% of your parts hitting top specs, will there be enough to go around for both "people who will pay up front for the full chip" and to stock the "upgradeable" pool?