[QUOTE="Riaz85"]
[QUOTE="artiedeadat40"]
Those effects could easily be done on the cpu if properly coded for x86.
artiedeadat40
ok that is not true, a CPU can do some really convinceing physics dont get me wrong, but they dont compare to the physics that GPU's can reproduce. its not the quality of physics that is the issue, its the quantity. A quad-core CPU dedicated to calculating physics for objects would most definetly fail to handle over 200 objects, PhysX was close to manufacturing PPU's that could calculate physics for 50,000 objects, the architecture for their PPU's are very simular to the GPU's nVidia is releasing, they do large number crunching calculations perfect for physics. GPU's were already well suited for physics calculations long before their was a need. CPU's need to stay flexible and versatile, why would you want them getting bogged down running your physics anyway? my point is if you do the research its clear, GPU's are far superior at physics then CPU's.
I have yet to see any proof to these claims and don't give me the Physix on Mirrors Edgesince those effects are not coded to run on x86.Cryengine 2 can pull off some amazing physics with only 2 cores/threads.
when half-life 2 came out they were limited to calculating 10 physics object simultaneously on a P4 EE, the dev said if the CPU was dedicated to physics it could calculate for 46 objects. its the same x86 framework they use today, and im being genrous, the core for a P4EE was more powerful then a core in a quad-core. this is old news and i wish i could cite the source but it was released forever ago. and if they could why wouldn't they code those physics for x86, think about it. thats proof right there, in Mirrors Edge they accomplish with GPU physics what they couldnt for CFPU physics. and its not a theory the GPU's are better at number crunching then CPU's it is a fact.
Log in to comment