75Views2Replies

Author Options:

CPU architecture idea/concept? Answered

I don't know much about CPUs or computers for that matter, but why hasn't anyone got the idea to take a multicore CPU architecture, and stagger the clock frequency of each core and have one shared cache memory between them all in the center so that the 4 cores can run one thread of machine code super fast? I'd imagine the reason we don't see clock frequencies greater than 4GHz is a problem with getting silicon to switch at those frequencies (maybe internal parasitic inductive and capacitive effects at the silicon level?) hence why multicore processors have become a thing? The problem I see with multicore CPUs is that programs need to be written in such a way so they can take advantage of parallel processing threads, and some tasks simply do not lend themselves well to parallel processing. Also, what is it about a CPU that determines the performance? It can't be just clock frequency(s) and the number of CPU cores, surely it has to do with some architecture and instruction-set stuff, but what exactly?

Comments

The forums are retiring in 2021 and are now closed for new topics and comments.
0
iceng
iceng

Best Answer 5 years ago

A problem arises with, which CPU (central_processing_unit) gets to access the same memory location. Something would have to arbitrate which CPU gets control.

Memory works by address lines that must be connected to only one CPU core at a time or address lines will fight each other when one core wants A9 high and another wants A9 low there will be damage..

The CPU only uses minuscule electric signals to generate addresses but needs power drivers to overcome inductive reactance of traces running to memory. The faster you clock the CPU the more power you need to overcome trace inductance and the hotter the computer gets...

The only way to win is to shorten the trace wiring between the CPU and periferals, then frequency can increase without over heating or, maybe place some of the memory next to the CPU inside the IC to work high speed while a lower clock fq goes out to the memory sticks to pre-load the CPU IC fast memory. This in fact is why some processors have multiple clocks.

0
-max-
-max-

Answer 5 years ago

Oh, because all cores are wired to the same memory and might request different things at the same time, didn't think about that. The RAM might need to be changed so 4 or however many cores there, addresses of memory can be accessed at an instant. That might be cost prohibitive. (and make for some realllllllly long DIMMs lol)