Finally someone that has understood something. Performance is not mesasured in MHz or any number other than the execution time of the job you want to run.
I would like to add though: You do not need the last part of your equation. Execution time for a certain function = Instructions/Function * average Cycles/Instruction * Time/clock cycle + hardware or IO delays The delays you are talking about means that the processor must stall for a number of cycles, thus lowering your CPI. Ofcourse you could write it like you have if you separate these stalls from other stalls and don't include them in your CPI.
The common way of writing "the iron law of performance" is: T=I*CPI*Tc (Time = Instructions * Clocks per Instruction * Clock cycle time)
Actually the GeForce has a lower clockspeed than the TNT2 Ultra. It is not even faster in some situations. But, and it is a big butt, the GeForce has an onchip transform engine which offloads your CPU in high polygon count scenes. That is thhe big deal with the GeForce.
Finally someone that has understood something.
/peter
Performance is not mesasured in MHz or any number other than the execution time of the job you want to run.
I would like to add though:
You do not need the last part of your equation.
Execution time for a certain function = Instructions/Function * average Cycles/Instruction * Time/clock cycle + hardware or IO delays
The delays you are talking about means that the processor must stall for a number of cycles, thus lowering your CPI. Ofcourse you could write it like you have if you separate these stalls from other stalls and don't include them in your CPI.
The common way of writing "the iron law of performance" is:
T=I*CPI*Tc (Time = Instructions * Clocks per Instruction * Clock cycle time)
Actually the GeForce has a lower clockspeed than the TNT2 Ultra. It is not even faster in some situations. But, and it is a big butt, the GeForce has an onchip transform engine which offloads your CPU in high polygon count scenes.
That is thhe big deal with the GeForce.