G
Guest
Guest
Hi all,
Sorry if this is a dumb question, I am merely a peasant with meager education
I play FSX which as most know is best played on a powerfull single core CPU rather than multicore due to bad mutithreading technique in the code.
Well why do multicore CPU's need to run with multi threading software for maximum efficiency because surely it is more practical to set the architecture so that multi cores are seen and treated as a single core by the application. The multi core CPU should not rely on the software to create the thread, it should itself be responsible for passing the code to the required core as and when required so acting like a single core.
From what I know of my AMD CPU, everything travels through the same L2 cache anyway.
Imagine a single lane highway that has 4 toll booths to manage the traffic so there is not a hold up. The traffic goes to each toll booth depending on how long the queue is at each booth (toll booth = core in this case). That is a perfect system but todays multi core CPU's do not work like that as far as I know but may be wrong.
Today's CPU's are like the traffic system above but the cars do not go to the best toll booth with shortest queue. They go to an assigned booth depending on pre given instructions that are not relevant to the core load (queue). Am I correct?
So why not just have a 4 way switch in the CPU that directs the load to a particular core depending on load rather than waiting on thread instructions? Simples!
Chris
Sorry if this is a dumb question, I am merely a peasant with meager education
I play FSX which as most know is best played on a powerfull single core CPU rather than multicore due to bad mutithreading technique in the code.
Well why do multicore CPU's need to run with multi threading software for maximum efficiency because surely it is more practical to set the architecture so that multi cores are seen and treated as a single core by the application. The multi core CPU should not rely on the software to create the thread, it should itself be responsible for passing the code to the required core as and when required so acting like a single core.
From what I know of my AMD CPU, everything travels through the same L2 cache anyway.
Imagine a single lane highway that has 4 toll booths to manage the traffic so there is not a hold up. The traffic goes to each toll booth depending on how long the queue is at each booth (toll booth = core in this case). That is a perfect system but todays multi core CPU's do not work like that as far as I know but may be wrong.
Today's CPU's are like the traffic system above but the cars do not go to the best toll booth with shortest queue. They go to an assigned booth depending on pre given instructions that are not relevant to the core load (queue). Am I correct?
So why not just have a 4 way switch in the CPU that directs the load to a particular core depending on load rather than waiting on thread instructions? Simples!
Chris