Deneidez said:
I wonder why you are telling this for us. They sure can do doubles, but its no use to use them because they are much slower than floats. Most of games(or all?) use only 'regular' floats to make 3D worlds. Same goes with physics & other stuff just because its much faster to use floats than doubles and because theres just about nothing you would gain by using doubles. |
I assume you're saying that double precision are slower than single precision, which has no base at all. They provide more exponent width and significant precision than common single flops. Thus it provides numerical stability to the cpu and make it capable of handling more tasks. @fj-warez: the Cell does not suck at double precision. Actually it is on par with the average quad-core CPU, just looking really bad when compared to single-precision performance of the cell itself. And IBM didn't leak anything, they actually announced a Cell variant focused on DP to be used QS22 Blade Servers.







