I’ve been trying to understand how the BOINC credit system works but it seems somewhat random to me.

Why isn’t it based on a straightforward model where points are awarded proportionally to the actual computing time, adjusted for the number of cores and clock speed of the CPU or the model of the GPU used?

Wouldn’t this make the system more fair and transparent? I’d love to hear your thoughts on why the current method was chosen.

For example the credits I've in Asteroids@Home are nothing compared to the ones in Amicable Numbers. I understand there's a logic in the current system but is it the best one we can have?

In reply to Luca's message of 2 Mar 2025:

IWhy isn’t it based on a straightforward model where points are awarded proportionally to the actual computing time, adjusted for the number of cores and clock speed of the CPU or the model of the GPU used?

Because the time it takes to process a task, is, quite simply, meaningless.
What matters is the work actually done- the number of computations actually performed in order to produce a Valid result. How long it takes to do that is a matter of efficiency; but what matters is the amount of actual work done, not how long it takes to do it.
That was why the Cobblestone was created- a 1 GigaFLOP machine, running full time, produces 200 units of credit (200 Cobblestones) in 1 day.
If a system is more powerful than 1 GFLOP, then it will get more credit per day. If an application is more efficient, and that 1 GFLOP machine can actually process 1 GFLOP of work per day (eg using MMX, SSSE, AVX, AVX512 etc) then it will get more Credit the exact same machine would get with a less efficient application. A system more powerful than 1 GFLOP with a more efficient application will get even more Credit each day.

For example the credits I've in Asteroids@Home are nothing compared to the ones in Amicable Numbers. I understand there's a logic in the current system but is it the best one we can have?

While Credit New is an abomination, the fact is that each project is free to do whatever they want when it comes to issuing Credit for work done, and many choose not to use Credit New. If every project was to implement Credit New as it was intended, then credit between projects would be comparable. They don't, so it isn't.

---

Grant
Darwin NT.

While Credit New is an abomination, the fact is that each project is free to do whatever they want when it comes to issuing Credit for work done, and many choose not to use Credit New. If every project was to implement Credit New as it was intended, then credit between projects would be comparable. They don't, so it isn't.

I only use credit as a way of alerting myself to problems. In the past, CPDN has been fairly vanilla in how it allocates credit but with some of their newer task types using 8GB of RAM/tasks a decision was made to give these tasks slightly more credit. (I am not sure how much.) I think this is justified as except on really high end boxes this is likely to restrict how many tasks from other sources can run at the same time.

I know there are projects which give a much greater return than others for the same time on a core but really don't have a clue about which is which.