Page 1 of 1
Confused about point allocations and work loads
Posted: Fri Mar 20, 2020 12:02 pm
by Kitutal
Just wondering how this thing works, yesterday I started out with a WU of 2,000,000 steps, was told it was worth something like 25,000 credits and would take eight hours. For some reason those credits were never added to my account, but never mind that for now, I've heard things are a bit slow with the new demand. This morning I get assigned another WU of 2,000,000 steps, but apparently this one will take 10 days to complete and is only worth 2500 credits? My computer is running at the same speed both times, sounding like a jet engine taking off trying to keep up with the work, yet a massive disparity between the amount it seems to need to do and what it is considered worth. There's no way I can keep my computer running for ten days making this much noise, I've tried turning it down to the light power rating and that has no apparent effect. I've tried waiting a while to see if the estimates change, but nothing has happened all morning apart from the first few thousand steps slowly being done.
Re: Confused about point allocations and work loads
Posted: Fri Mar 20, 2020 4:33 pm
by Joe_H
2,000,000 steps on a large protein takes longer than the same number of steps on a smaller one and gets more base points. Without project numbers to compare, hard to say exactly.
The ETA initially calculated when a WU from a project you have never folded on that system will be inaccurate until a percent or two of the WU has been completed. If you have left the WU running long enough the estimate will be closer to the actual time needed. The timeout figure is sized to give people the option to not fold full time, and in practice most WU's can be completed in less than a day.
The stats are getting updated, but there is a backlog of entries to be added. If you provide information on the Project, Run, Clone and Gen numbers for a WU we can check on it, or you can use the same numbers to check the WU Status at apps.foldingathome.org.
Re: Confused about point allocations and work loads
Posted: Fri Mar 20, 2020 4:54 pm
by JimboPalmer
[Understand that i know nothing of your PC and will just make stuff up]
A Work Unit for your graphics card many have 50,000 atoms to track, because your Graphics card may have 640 simple threads it is expected to do them raphidly
A Work Unit for a CPU should be smaller but you CPU may only have 8 somewhat faster threads. Obviously the Graphcs card has a 800 to 1 advantage over the CPU, although the CPU can (slowly) solve problems the Graphics card can't.
Both are used and both help the cause, but they don't act the same.
WUs that mention 21 or 22 work on the graphics, WUs that are a7 are for your CPU.
The individual researcher picks how many steps he/she wants to look at. It depends on her/his goals.
Re: Confused about point allocations and work loads
Posted: Fri Mar 20, 2020 5:20 pm
by bruce
JimboPalmer wrote:[Understand that i know nothing of your PC and will just make stuff up]
The individual researcher picks how many steps he/she wants to look at. It depends on her/his goals.
Each Step represents a microscopic bit of motion for all 50,000 atoms toward some new shape at an end-point defined by that goal. You have been assigned a WorkUnit that asks you to calculate 2,000,000 of those steps, When you finish that assignment, identified by 4 numbers PRCG, (Project, Run, Clone, Generation) you'll return that result and be given another assignment. Assuming, you have completed Gen N, the server will immediately build a new assignment called Gen (N+1) that starts from the end of your calculation and adds another 2,000,000 of those steps to the ones you just finished. That series of Gens is called trajectory P,R,C when it reaches that final goal. Other trajectories for the same protein P are also assigned but with different values of R and C so a number of different end-points are determined.
If, in one of those trajectories, the drug happens to bind to a site on the virus that incapacitates it. we have identified a candidate drug.
Re: Confused about point allocations and work loads
Posted: Fri Mar 20, 2020 8:39 pm
by JimF
bruce wrote:If, in one of those trajectories, the drug happens to bind to a site on the virus that incapacitates it. we have identified a candidate drug.
That takes the process a bit further down the road than I had imagined.
I trust that this can be published in some form rapidly these days.
I would like to think that Dr. Pande (for example) is on the case immediately.