OSDev.org

I've decided to develop my OS in x86-64. Do you think that is a correct decision? I've read that Intel has tested smartphones and tablets with x86 architecture. If it continues like this, ARM architecture will die and Intel will be expanded in all the market? Or will be backwards (ARM will win the computers, laptops,... market)?

The only constant is change.

Whichever architecture you choose, it will eventually lose market share and die. Maybe sooner, maybe later, but it will. The only way around this is to make your operating system flexible enough that it can easily be ported to other architectures. 99% of an operating system is trivial to port. If you make the right design decisions the other 1% can be moderately easy. If you make the wrong design decisions the other 1% can be near to impossible.

x86 has no real chance against ARM in embedded markets. Intel has real trouble getting the power consumption down to ARM levels, this despite Intel is far ahead in process technology. Intel for now only sell chips and not IPs where ARM sell IPs only. As I see it, if you want to be a contender in mobile SoCs, IPs is a requirement. Intel might start sell IPs but then ARM has already years of experience and a large product portfolio. Right now I don't see Intel competing with SoC makers like Texas Instruments, they are simply too slow developing new products.

ARM on the other hand will start to penetrate the server markets because they want reduce power consumption. Google for example consume electricity like a quarter of a nuclear power plant, so there is money to be made if you can reduce it.

On the other hand, ARM will not win the desktop or the laptop as there are so many programs out there compiled for x86.

BTW. ARM after going out of order execution, we have seen that ARM is starting to consume electricity too. I actually believe that having a WLIV ISA where the compiler determines which instructions can be executed in parallel is better suited for this when power consumption is important. In a few years ARM will be seen as the power hog, where processors that use WLIV will start to displace ARM in its traditional markets.

If you are going to target your OS for mobile phones and such, then I say develop for ARMs oterwise x86 will do fine. In general it shouldn't really matter but you easily can cave in using one architecture only.

WLIV = Wery Long Instruction Vord? You spelled it wrong twice.

Anyway, wanted to note that if you take all of Google's data centers and put them together, they average 2.2 megawatts and they don't even use dynamic voltage-frequency scaling because the savings don't justify the costs of implementing it.

As far as the x86/ARM thing goes, I think things will stay the same in the near future but JackScott is certainly right in the long run. It may seem weird to those are not very familiar with the history of computing but it was just as weird to others in the past.

Before the iPad and Android tablets, a 'tablet computer' referred to Tablet PCs which were x86 computers running full copies of Linux and Windows (there was even a Windows XP Tablet PC Edition). Now they are known as 'Microsoft Tablet PCs' to distinguish them from the new breed of Android/iOS tablets.

Back in the day there were two kinds of Tablet PCs: Slate tablets, which were basically just a screen and you had to use a stylus to operate them, and Convertible tablets (which I own one of): they basically looked like a laptop but you could close the screen facing up so you could hold it in one arm and use it in 'slate mode'.

I used my convertible tablet all throughout my university studies with Microsoft OneNote (the best note taking tool ever IMO, you could mix images, sound clips, typed text, and handwritten notes on an endless page and they were all searchable).