Page 1 of 1
What good are points?
Posted: Thu Mar 19, 2020 5:04 pm
by jeffmoor62
Hi All,
Apologies if this has been discussed, but I didn't see it. I've seen posts about how points are calculated and some about them not being updated. Why do I want them in the first place? Is it essentially just bragging rights?
Re: What good are points?
Posted: Thu Mar 19, 2020 5:16 pm
by Joe_H
As far as the F@h project is concerned, yes, just bragging rights.
Some teams have rewards, virtual and actual, based on points earned. Others run contests of various sorts with some kind of reward, honor, etc.
Re: What good are points?
Posted: Thu Mar 19, 2020 5:20 pm
by Jesse_V
They have no direct implications to the underlying research, they are mostly a reward for completing the work. Each project is benchmarked and points are awarded based on how quickly and how reliably you are able to return the results. For example, if I have a machine that generates 100k points/hour and you have one that generates 200k/hour, then I know that your hardware is about twice as fast for Folding@home. This gives people a pretty good way of discussing and bragging about their machines and a great way to set up competitions to the most productive members of a team.
You get the most points with a passkey. If you get 10 workunits done with it with 80% before the deadlines, then you'll get bonus points, which are substantial extra number of points for your efforts and a clear way of increasing your bragging rights. The more points per day, the better.
Re: What good are points?
Posted: Thu Mar 19, 2020 6:03 pm
by JimboPalmer
When I try out different configurations of my hardware, point can tell me if this configuration does more science (does more points) that some other configuration.
Once I am done with that, just bragging rights.
Some measure of how rapid the growth has been in the last week, I was in second place on my team in Points This Week last Thursday. I am in 8th in PTW now. without points how would you see the growth?
Re: What good are points?
Posted: Thu Mar 19, 2020 6:11 pm
by kiore
I agree with Jesse and Jimbo, points useful as feedback to help understand which configurations do more work and act as a marker of value. Of themselves they have no value of course I cannot buy anything with them or trade them, but I do value my own as a representation of work I have done. Bragging rights, yes I suppose but I tend to view them as feedback from the project.
Re: What good are points?
Posted: Thu Mar 19, 2020 6:33 pm
by Jesse_V
kiore wrote:I agree with Jesse and Jimbo, points useful as feedback to help understand which configurations do more work and act as a marker of value. Of themselves they have no value of course I cannot buy anything with them or trade them, but I do value my own as a representation of work I have done. Bragging rights, yes I suppose but I tend to view them as feedback from the project.
Well in some teams there are actual rewards or cryptocurrencies attached to the points. It depends on the team, if you're part of some dedicated group.
Re: What good are points?
Posted: Fri Mar 20, 2020 3:58 am
by v00d00
Points are easy to come by nowadays. Back in years gone by you actually had to work hard for them. Ask the older folders about certain workunits from the 6xx series that circulated back around 2004-2005 and took one whole day to complete and were worth not a lot of points, back when processors were a single core (and on older hardware could take considerably longer).
Workunits. That is a good measurement of how much you contributed to science. Some people might have millions of points, but not many workunits. Do 10,000 workunits and i'll be impressed (20,000 and you are one officially a pro folder).
PS. The new certificates aren't a patch on the old ones.
Re: What good are points?
Posted: Fri Mar 20, 2020 4:04 am
by Jesse_V
That's because the hardware has gotten SIGNIFICANTLY faster and people really don't like if a particular project is under-benchmarked on the points. I don't think workunits are a fair measure either because they can divide up the simulations in any way they like, and modern systems have the RAM and CPU cache to handle much larger molecular units now than before. So there's inflation on the points, but at least the projects are benchmarked so they are all comparable at each point in time.
Re: What good are points?
Posted: Fri Mar 20, 2020 8:13 am
by JimboPalmer
522,992,461 points, but it took me 159,631 Work Units.
An semi-infinite number of slow CPUs will do that over 11 years. (see, Bragging rights!)
But they really are useful to decide is dual channel RAM is faster than single channel RAM. (it is) Is avx-256 faster than sse2? (it is) is Core_22 faster than Core_21? (it is)
But mostly, bragging rights.
Re: What good are points?
Posted: Fri Mar 20, 2020 4:47 pm
by VirusCorona
Hello,
After a few hours of folding (yesterday afternoon, Polish time) I got 370 points. After I came back after about 8,5 hours, there was an error on my laptop (it often happens if it's running for some time without my interaction, like it was missing me) - screen got black and it didn't respond, so I could only restart it. After restart it was folding the whole night and day (this error didn't happen again), but I still have 370 points. Does it mean that it's work is useless and I have to fix something? Or just wait for points counter update? After the restart I'm sometimes getting notifications that entry point for some procedure wasn't found.
Re: What good are points?
Posted: Fri Mar 20, 2020 6:31 pm
by Joe_H
Points are entered into the stats after a WU is returned, and currently there is a backlog waiting to be entered into the database. In addition the stats lookup is now set to access a copy of the database to lessen the load on the server for the database, I do not happen to know how often that copy is updated.
As for the other issues, if you could post your log and describe the problem in a new topic, it can be looked at and possible fixes suggested if needed.
Re: What good are points?
Posted: Fri Mar 20, 2020 9:07 pm
by v00d00
Jesse_V wrote:That's because the hardware has gotten SIGNIFICANTLY faster and people really don't like if a particular project is under-benchmarked on the points. I don't think workunits are a fair measure either because they can divide up the simulations in any way they like, and modern systems have the RAM and CPU cache to handle much larger molecular units now than before. So there's inflation on the points, but at least the projects are benchmarked so they are all comparable at each point in time.
I understand entirely why they use them. if you want some people to continue to fold, you have to stroke egos, im not against doing that either as its a case of whatever is required to complete the goal, which in this case is a noble one. So if giving people points means getting more science done quickly, then I see no moral or ethical problem with it.
But ultimately they arent worth anything. Although if you can find a team that will give you virtual currency for folding that would be sweet.
Re: What good are points?
Posted: Fri Mar 20, 2020 9:09 pm
by Jesse_V
v00d00 wrote:Jesse_V wrote:Although if you can find a team that will give you virtual currency for folding that would be sweet.
There are a couple of teams that do that, yes. And those teams are still one the most popular and powerful teams in all of Folding@home.
