south side sammy wrote:please don't speak about as if I'm not here. I'm sitting in the same room as you.
I certainly hope you aren't.
south side sammy wrote: I only stated what I wanted 15 times. I'm sure memory/memory bandwidth is part of it but most certainly NOT all of it. GPU-z won't really give me what I want. A good look at what the architecture of the card puts forth inre to physx... and just how much of a "physx" card would be needed to provide that.
I don't want to know anything about ancient technology ( gtx260, etc. ) I'm talking about now. I can't believe nothing has advanced ( in the past 4-5 years ) to get to the bottom of this. There has to be some kind of calculation to measure this.
Couple of things here.
1. No one mentioned memory. It has very, very, very little to do with memory. We've already mentioned shader clock and gpu/cuda cores. Those are the two factors.
2. Ancient technology? well, here's the thing.... physx is easy to do. Remember, going to back to 8xxx hardware, physx could be run. Sure, it'll be a hit on performance. But, if you simply use another card, like a GT 450, that hit disappears. Or you could add a 9800, or whatever else you want.
There isn't a scaling in physx. You won't have a better experience with a better card. The fact that it only takes 10% off of the performance should tell you something. As a rule of thumb, you'd only need 10% of what your current card is running. So, look at core count and shader clock of your current card, multiply by .1, and see what card fits the bill. But really, just getting an old card (will be power hungry) or a new lower end card (will NOT be power hungry) are the only two options that will result in good results. Anything more and you are wasting money.