I was just thinking... with ATi now having those unified shaders, and dynamically allocating them to where they are required...
It's quite similar to what HyperThreading does really... except that it doesn't work at pipeline level, but rather at execution unit level, because the execution units aren't unified.
Now there is dualcore, but it doesn't have the flexible scheduling that HyperThreading offers... which means that no matter how many cores of say 3 GHz you add, a single thread will run no faster than it would on a single 3 GHz core.
So I thought... instead of adding a second complete core... what would happen if you would add its execution units to the first core instead, and expanded the HyperThreading/scheduling? Then the whole system would be more dynamic, and be able to boost single threads aswell, and still allowing multiple threads to run simultaneously, like a normal dualcore already does.
I just wonder... is this something that Intel just didn't think of, because it's unusual... dual CPU is how things were done in the past, and dualcore is just an extension of that...
Of course it could also be that they did think of it, but didn't use the idea because a single core is already designed in a way that there rarely is an 'execution unit' shortage...
That is something I am not too sure of however, since AMD uses 3 ALUs instead of Intels 2 ALUs (although Intels do run certain instructions at double speed).
Anyway, just a thought
It's quite similar to what HyperThreading does really... except that it doesn't work at pipeline level, but rather at execution unit level, because the execution units aren't unified.
Now there is dualcore, but it doesn't have the flexible scheduling that HyperThreading offers... which means that no matter how many cores of say 3 GHz you add, a single thread will run no faster than it would on a single 3 GHz core.
So I thought... instead of adding a second complete core... what would happen if you would add its execution units to the first core instead, and expanded the HyperThreading/scheduling? Then the whole system would be more dynamic, and be able to boost single threads aswell, and still allowing multiple threads to run simultaneously, like a normal dualcore already does.
I just wonder... is this something that Intel just didn't think of, because it's unusual... dual CPU is how things were done in the past, and dualcore is just an extension of that...
Of course it could also be that they did think of it, but didn't use the idea because a single core is already designed in a way that there rarely is an 'execution unit' shortage...
That is something I am not too sure of however, since AMD uses 3 ALUs instead of Intels 2 ALUs (although Intels do run certain instructions at double speed).
Anyway, just a thought