Of course, I have long held the radical notion that we do not do enough checking and that more projects should be doing at least a replication of 3 min quorum of 3 with the inclusion of test tasks that actually certify the machines that we are running on ... to make sure that we are, in fact, getting good results ... and not just results that agree with someone else's results... agreement among wrong answers is worse than simply having wrong answers ... because now we are convinced that the wrong answer is somehow more right than it really is ...

I too have always felt a "test" unit should be periodically sent to each pc to help ensure everything is still working up to snuff. And I have always thought each project should have more that a 1/1 ratio. Now 15/19 or even 10/19 can be a very picky project and probably not for me personally. That would have to be close to finding a cure for something, or something else like that, for me to crunch for them.

---

I must admit, that WCG has some good projects, too. One has initial replication of 1! But yesterday I realized that WCG uses initial replicator of 19 (!) in many
projects.

Pentti Kiesi, you might want to check this thread on the WCG site before you judge so harshly. It explains how for the Human Proteome Folding project, each of the 19 tasks is different.

I must admit, that WCG has some good projects, too. One has initial replication of 1! But yesterday I realized that WCG uses initial replicator of 19 (!) in many
projects.

Pentti Kiesi, you might want to check this thread on the WCG site before you judge so harshly. It explains how for the Human Proteome Folding project, each of the 19 tasks is different.

Thanks, that is very useful ...

Though I was not that troubled by the 19 replication ... it is still a comfort to have a good explanation ...

Now if they would just start to use the GPU for some of the sub-projects ... :)

I really want to use the GPU for something other than GPU Grid ... not that what they are doing might not be valuable ... I just am not into SaH and that is the only other alternative ... besides, SaH has more than enough people ....

---

I must admit, that WCG has some good projects, too. One has initial replication of 1! But yesterday I realized that WCG uses initial replicator of 19 (!) in many
projects.

Pentti Kiesi, you might want to check this thread on the WCG site before you judge so harshly. It explains how for the Human Proteome Folding project, each of the 19 tasks is different.

Thanks, that is very useful ...

Though I was not that troubled by the 19 replication ... it is still a comfort to have a good explanation ...

Now if they would just start to use the GPU for some of the sub-projects ... :)

I really want to use the GPU for something other than GPU Grid ... not that what they are doing might not be valuable ... I just am not into SaH and that is the only other alternative ... besides, SaH has more than enough people ....

Ok. I apologize my words. I was unjustice and didn't understand enough the proteome folding science while making that email. Yesterday evening I read CASP
papers and now I know better.

/Pentti

Ok. I apologize my words. I was unjustice and didn't understand enough the proteome folding science while making that email. Yesterday evening I read CASP
papers and now I know better.

Not a problem from *MY* perspective ... :)

My only disappointment is that there seems to be so many projects doing Folding in one form or another ... kinda leaves those of us that are interested in physics out in the cold a little bit ...

Anyway, we all learned something and that is all to the good ...

---

Hello. Traffic is bidirectional. I am gradually closing down some WCG projects and trying this.

I must admit, that WCG has some good projects, too. One has initial replication of 1! But yesterday I realized that WCG uses initial replicator of 19 (!) in many
projects. It means that computing efficiency is a little bit over 5 %.
Some lost WUs now and then is very little wasting compared to that.

I used this as a positive example to calm down one frustrated cruncher,
who was angry about buggy Clean Energy project. That moment I didn't
think it as a bad thing to have 19 replicas, but my view of point changed,
when I noticed the whole thread had "disappeared".

Now my only criterias are:
1. Benefit to mankind
2. Overall efficiency

So you haven't noticed that some WCG projects are using what appears to be a very high replicator to search more of the available problem space? Those aren't 19 exact replicas; they have 19 different starting points in the rather large problem space.

However, WCG is planning to go inactive for at least 48 hours during the middle of this month for a move to a new location, and probably slow down issuing workunits first, so it's a good idea to be ready to have an alternate source of workunits.

I'm currently trying to use 7 sources of workunits relevant to medical research, and only 3 of them are currently good at supplying workunits. In case you'd like another source of workunits for times when Rosetta@home is having problems, the third one is:

http://boinc.fzk.de/poem/

Don't expect them to present much information on which conditions they're trying to help, though; they currently seem to be more interested in getting the software ready.

Well, I do not know about linux, but I am certainly taking all my windows machines off ... I just killed two more Tasks that were over the clock one at 13 hours and the other at 6 something ...

NOt sure if the project is paying any attention, but they might want to look at the code differences between OS-x which has not had an overrun that I can tell (40 some tasks) and WIndows which can't seem to avoid it ...

Have you considered the possibility that they are not issuing workunits using their newer features to OS-x machines, only workunits that confine themselves to their older, well-tested features?

Windows machines, especially those with large amounts of memory, on the other hand, seem to frequently get workunits using the newer, not-so-well-tested features.

I had to take a two year break from my BOINC addiction and when I came back about a year ago the total number of people doing BOINC had not changed significantly... it still has not ... I don't believe it will until many things change ...

Have you noticed that BOINC is in the process of adding the capability to use GPUs, as well as the CPUs they have used in the past? GPUs are typically capable of running calculations about 10 times as fast. However, their architectures are different enough that it's not just a matter of recompiling the same software to run on a different machine. Programming to run on GPUs is difficult enough that not many of the BOINC projects have tried it yet, especially among those with frequent software updates and/or high memory requirements, such as Rosetta@home. However, it looks likely to eventually allow the whole BOINC world to do about 10 times as much calculation on the same number of machines as at present.