- Aug 25, 2001
- 56,587
- 10,227
- 126
I was just reading about the BOINC points awarded in the reisel sieve thread being under discussion, and I started to think about this issue.
Should there be a standardized unit of "work computed", that should be used by ALL distributed-computing projects?
I was thinking of something like the Mhz-hour, which is the Mhz multiplied by the CPU time elapsed for the process in question.
But the problem there is scaling the points according to the efficiency of the architecture.
Then again, perhaps that metric would actually be more fair, as people with slower, older architecture PCs, such a P4s and Athlon XPs, would be more likely to contribute to DC, if they could be awarded the same sort of points as faster systems. Only less actual work would be completed.
IOW, should a P4 Mhz-hour be equivalent in terms of scoring, to a C2D Mhz-hour?
In terms of fairness as to what computing resources a project consumes, I think that they should in fact be equal, as the user's computer is being utilized for the project for the same amount of time. In a way, this awards points based on participation, not on output.
Then again, there could be a "scaling factor", that would be computed according to a benchark of the particular project, that would attempt to normalize according to a CPU's archtectural efficiency.
These units would be similar in terms of usage to a watt-hour, and power efficiency.
The scaling factor would be a per-project thing, whereas the overall Mhz-hour scoring would be considered universal, even comparible between projects.
Edit: Call the scaling factor, "performance factor". If a P4 had a PF of 1.0, then a C2D would have a PF of roughly 2.0. A C2D would also have twice the Mhz-hour rating, because it has two cores, assuming the project was multi-threaded.
Should there be a standardized unit of "work computed", that should be used by ALL distributed-computing projects?
I was thinking of something like the Mhz-hour, which is the Mhz multiplied by the CPU time elapsed for the process in question.
But the problem there is scaling the points according to the efficiency of the architecture.
Then again, perhaps that metric would actually be more fair, as people with slower, older architecture PCs, such a P4s and Athlon XPs, would be more likely to contribute to DC, if they could be awarded the same sort of points as faster systems. Only less actual work would be completed.
IOW, should a P4 Mhz-hour be equivalent in terms of scoring, to a C2D Mhz-hour?
In terms of fairness as to what computing resources a project consumes, I think that they should in fact be equal, as the user's computer is being utilized for the project for the same amount of time. In a way, this awards points based on participation, not on output.
Then again, there could be a "scaling factor", that would be computed according to a benchark of the particular project, that would attempt to normalize according to a CPU's archtectural efficiency.
These units would be similar in terms of usage to a watt-hour, and power efficiency.
The scaling factor would be a per-project thing, whereas the overall Mhz-hour scoring would be considered universal, even comparible between projects.
Edit: Call the scaling factor, "performance factor". If a P4 had a PF of 1.0, then a C2D would have a PF of roughly 2.0. A C2D would also have twice the Mhz-hour rating, because it has two cores, assuming the project was multi-threaded.
Facebook
Twitter