> > The difference in encode times > >between 4 cores 2.4GHz and 4 cores at 3.2GHz is dramatic--overclocking > is > >still very much alive and very much worthwhile. > > For you, maybe, not for me. I can spend hours editing, and encoding > video... I might not even see an annoying anomaly in the first > viewing....it might be the second or third viewing, months or years > apart. If I over clock I have to wonder.. was this caused by a blip > in the hardware.
Right, because an errant calculation caused might cause the motion estimation calculation for a frame to be off, resulting in...likely nothing discernable. There are so many trillions of calculations involved that the impact of any single calculation being off is likely one garbled macroblock on one frame. It isn't like you're protected from that by not overclocking...all processors and all memory will introduce errors over the long term. Am I running closer to my margins? Yes--but it is worthwhile given the substantial speedup I get. Frankly, I feel that I probably do better validation of the chip when determining if the overclock is stable than the CPU manufacturer does when binning. I'm an atypical overclocker through. I don't push it to the limits and then keep it there. I push it as far as I feel comfortable, validate, then back it off a step or two. > > I have a .38 special. Almost every expert, and authority will claim > that I can shoot the occasional +P round in it and it would be > fine.....but then you got to wonder... is this the day the barrel is > going to blow off in my hand. I don't like the feeling that maybe > this time things won't come out right, when I am spending a lot of > time on my work. What, no car analogy? A little overdramatic, don't you think? > > > > One thing that really isn't considered is the growth of quad core > > > supported apps. My guess is that next year lots of new apps are > going > > > to support Quad core and the years after that even more. That means > > > your Quad core processor is actually going to get faster over time, > > > which will not be the case for dual core. > > > >Sure it will. > > yeah, but the Quad core is going to start out already even, or faster > then the dual core, but mostly unsupported. The Quad core performance > curve, over time is going to be a lot steeper then the dual core. > While there may be a degree of truth to that (you are assuming that the majority of tasks can be scaled to n cores effectively--a likely fallacious argument), that isn't what you said. You stated that a dual core will not "get faster" over time, while a quad-core will. I believe that we will find that those tasks that can have multiple worker threads do so far more efficiently with two threads than n. The number of killer apps for 4+ cores will likely remain small into the foreseeable future. Greg
