| coolestguyever said: does anyone actually know what the fuck any of that glops shit means or are you all just pretending to be smart? thats an actual question, does anyone know? |
A floating point number is (essentially) a number that is represented in scientific notation "on" a microprocessor ...
A FLOP is a floating point operation which is (essentially) an addition, subtraction, multiplication, division or comparison operation (equal, not equal, less than, greater than, etc.) that is performed by the CPU. (There are dozens of other potential operations on modern CISC processors but that is another story)
When people talk about peak GFLOPs they are (typically) refering to the maximum number of floating point operations that can be performed by a processor which (typically) means that they are performing the simplest operation, using a single precision floating point number, over and over again without ever reading from memory. This is (generally) a bad indication of performance because real world applications tend to use double precision floating point numbers with a variety of different floating point operations (including the more complicated and time consuming operations) and they read from memory quite heavily.
The limitations of peak theoritical and real world FLOPS is one of the reasons you will very rarely see them used as benchmarks is serious discussions about the performance of a system. In fact, the number of FPS that a system could run Quake 3: Arena would still be far more meaningful than its GFLOPs performance even though most modern hardware could run it at far higher than 1080p at a framerate exceeding 500FPS.
