Q. So you are working on it?
A. Not really. Too many outstanding compatibility problems still. It's low priority.
Q. And will using 3 or more cores be beneficial to pcsx2?
A. Eventually. But by how much it will be beneficial is hard to determine. It certainly won't be by a factor of 2 or anything like that. Maybe a 20-25% speedup in the end, if we're lucky.
Jake Stine (Air) - Programmer - PCSX2 Dev Team
Thanks, makes sense you're working on compatibility at the moment. But a 20-25% increase in speed is pretty good though, but then again it'll be a lot of work.
09-21-2009, 10:53 PM
(This post was last modified: 09-21-2009, 10:54 PM by ChronicNL.)
well 20-25% FPS more would be great tho and could make the difference to obtain full speed however its far easier to just overclock your CPU
and there is affinity/priority change in taskmanager wich works and gives for example 50fps where you would get 42/43 without.
it varies on the software you do it with tho
the logic behind this is if pcsx2 uses core 1 and 2, core1 will have a lot of things to do like run windows and such
core3 and 4 have far less to do so more for pcsx2
...the cores are like core0-3 & yes, first core (core0) runs system, so it's so to say always in use, last core (core3) usually is used by programs for capturing video etc. (like Fraps), so second & third are optimal for PcSx2 (core1 & core2)
4x 4GB DDR3 2400MHz
AMD Radeon R9 Fury X
Hey, programs does not use CORES, they don't know about this! Programs use threads, and it's OS job to handle threads in most efficient way. ANd OS newer do such a stupid thing, as using core-0 more ofthenly than all others. Even more, in every unbalanced two-threads load there is no need to use 3-rd core for low-level tasks: until one thread use 100% of his core, and other stay lower than 80%, 3rd core is no need. All other tasks could be made at this 20% free time. And it's just no difference between 2 and 4-cores CPU at such case, except 4-cores are slower per same clock.
To obtain benefit from 3rd core program should divide the most heaviest thread into the two or more, and divide efficiently. And recompiller logic are pretty linear, it's turing machine.