Are "steps" always equal? [No]

[BrownBullhead]

Are "steps" always equal?

Let me first describe my environment: Windows Vista 32-bit Home Premium, 2 GB RAM, Intel Quad Core Q6600 CPU @ 3.0 GHz.

I ask because several days ago, when I was running 5.91 console, my client downloaded a WU of 2,500,000 steps (2 million+) which completed each 1% (25,000 steps) per 14 to 15 minutes. The completed WU took just over 24 hours. ***This was on Project 4012***

Now, running 5.92 console, my client downloaded a WU of only 250,000 steps (1/10 the above example) which is only completing each 1% (2,500 steps) per every 18 minutes ***This was on Project 2665***. 1/10 the steps in this 5.92 WU are taking 25% longer time when compared to 5.91?

Or, are steps "different" per WU? I guess what I am asking is: Is a step always a step?

[DanEnsign]

Different work units correspond to different model systems. The main determination of the speed of one step in a simulation will be the number of atoms represented in the model. 2665 is probably one of Peter's humongous projects.

I'm not sure how you would get a 4012, since that's a PS3 project, but perhaps it's mislabeled (the best way to tell the actual identity is to look where it says Project: 3051 (Run 6, Clone 92, Gen 0) or what have you). In any case it's probably a CPU version of a PS3 project (for testing) which are all extremely small number of atoms and so the steps will take a lot less time.

In case you're interested, one "step" corresponds to solving Newton's laws for all the atoms in the system on a discrete time step. You can see why this would take longer for more atoms.

Dan

[BrownBullhead]

Please accept my apologies; I do run on a PS3 as well and I quoted the wrong project from my personal stats when writing that post.

I read some other threads and I learned that 2665 is not a typical WU. Thank you for your feedback, although I must say I am not a very scientific guy; just want my otherwise idle equipment to work toward the "greater good".