Thanks for that concise answer. I really appreciate it.
Do you know why people were saying on MacRumors and 9to5Mac that Apple was planning an Intel processor with an arm coprocessor handling low power then? This is not me being a jerk I’m just really curious why that’s possible but my theoretical 3 processor idea isn’t. I’m really curious because to someone not well learned on the subject it seems all too similar because one of the processors has to know whether to send the program to the other processor.
Honestly those rumors might have been stupid becuase it isn’t actually possible but I feel I must be missing something.
Well, think of something like the touch bar and the T1 chip, which is a SoC not entirely different from what's in the Apple Watch. It's not too far from being its own independent subsystem. It's not that far fetched to imagine something similar with its own very small OS that can handle very simple tasks like waking up the main computer if it sees certain network packets. And I'm sure they'd be doing more than that if the rumours were indeed true.
Or if you wanted to imagine a SoC that is effectively an iPad on a chip, that borrows display, speakers etc from the MBP, then that would probably work fairly well too. That SoC would run iOS, the Intel chip would run macOS, and they wouldn't be scheduling apps between them.
Both of these examples are pretty similar to how a dGPU works. When you want to run a compute job on the GPU, you have to send it the program (called a kernel, just to keep things as confusing as possible). Then you send it the data (copy from main RAM to GPU RAM). Then you tell the GPU to run the program on that data, and possibly write output data to somewhere else. Then the output data has to be copied back to the main RAM. The GPU is effectively a coprocessor, but it's not SMP as when you have a dual socket workstation. They don't (typically) share RAM with the main CPU.
The key here is data sharing. If you want two processors to have shared access to the same data, then they have to communicate on how they access that data, and that makes it more complex. Probably also a bit slower. At the same time it's pretty convenient to share data and not have to spend time copying it from one chip to another. Engineering tradeoffs...
Now engineers sometimes come up with ingenious ways of creating magic. If you remember back in the 80's, the Amiga had one main CPU and several coprocessors that could indeed work independently with main RAM and other resources. This was, if I remember correctly, achieved by clocking the coprocessors at a fixed multiple of the main CPU, and then phase shifting the clock so they never actually conflicted with each other. Which was pretty neat for the original box, but fairly inconvenient later when they wanted to bump the clock speed of the CPU independently. And this was the early 80s mind you. What's possible or not possible is constantly in motion.
Anyway, while many weird things get written by people who don't know tech, the low power chip idea was probably fairly feasible.
[doublepost=1527521949][/doublepost]And just to keep this at least remotely on topic -- I do think it would be pretty cool if Apple would drop in a cheap ARM coprocessor in the MBP. Not just for sending mail in its sleep, but also for iOS app development, or just being able to run iOS apps on the MBP at full speed. I quite doubt we're going to see that this year though.
