9037 - low points even though more atoms and longer deadline

[PFM]

Comparing project 9037 with 8608, points for 9037 are much lower as compared to 8608. Why is that ? Is that normal or does anything need to be changed/fixed ?
9037 - 33k atoms, deadline 3.8/8.2 days, base points 207.
8608 - 10k atoms, deadline 0.43/4.8 days, base points 314.
I dont have filesizes but I dont remember anything out of the ordinary, although not sure if it even matters for the question.
Thanks.

[davidcoton]

Note the longer deadlines on 9037. This increases the effect of QRB, so actual PPD for a QRB-enabled slot will not be in the same proportion. It is always difficult to compare points across projects. Although there is an attempt to set points value with some consistency, this is not possible across all projects and all hardware, particularly with the complications introduced by QRB.

[Joe_H]

Also, if you have run WU's from these two different projects, check the total number of steps involved in the logs. That can vary between projects, some run for 500,000 steps, others 250,000, and yet others use a larger or smaller number of steps than those.

The base points awarded for WU's is determined by the length of time they need to complete on a benchmark machine. QRB does complicate the comparison a bit, but the PPD of different projects on the same equipment should be comparable. PG does make some effort to keep different projects within a range of +/- 10%, but there are out liers.

[PFM]

The base points awarded for WU's is determined by the length of time they need to complete on a benchmark machine.

Thats what I was trying to ask - when comparing these two projects one of them seemed like it would need more 'work' by the machine so I was expecting to see more base points for it.

[Joe_H]

Now that I am home from work, I just looked up a run of each project. Project 9037 runs for a total of 250,000 steps, while Project 8608 takes a total of 2,500,000 steps. That is 10 times as many iterations of the simulation. So, about a third the number of atoms, 10 times the iterations and it end up with base points about 50% higher. The total number of atoms is a starting point of comparison between projects, but is only one factor.

[PFM]

8608 has 2.5 Million steps ? thats interesting. Is steps the same thing as those 1% increment steps in the log file ? If so I dont remember having seen so many steps in the log for any project wu that I folded recently, but I will see if I can find a few days old logs to confirm. Even if it were true it doesnt quite add up. If it were really 2.5M steps then at the rate it was folding it would take years to finish.

[Joe_H]

No, it is just related to those percentages. You will see an informational message in the log for each percent completed similar to this:

Code: Select all

18:23:53:WU01:FS00:0xa4:Completed 25000 out of 2500000 steps  (1%)

How many steps corresponds to each percent will depend on the total number of steps the WU is set to complete. Information passed on through bruce a while back was that each step was 2 femtoseconds as I recall.