Quote from: Oshyan on December 08, 2007, 02:07:18 AM
The final release will support multi-core. This would obviously remove the need to split up resources between instances of TG2.
I've got a query on this...
When an application is multi-threaded, I've read that performance never ramps linearly - so a quad core wont be 4x faster (how much slower I guess(??) would depend on the tasks' suitability to being threaded, and programming skill - but potentially things could be significantly slower). If this
is the case it would surely be more efficient to render with 4 separate instances split-up intelligently, rather than 4 cores munching away on 1 complete instance? Or is there something I'm not considering?
Don't really know much about it other than random bits that I've read, but am curious as I'm about to embark on the multicore adventure myself.
Oh, and I was wondering - as Matt currently has his nose deep in the innards of the renderer, do you know if he's going to be implementing any of the new SSE4 instruction set which Intel's new chips are now sporting... assuming any of them are of use to the rendering process, and that you are allowed to say, of course.