![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Gilthanis
[H]ard|DCer of the Year - 2014
- Joined
- Jan 29, 2006
- Messages
- 8,729
I have been reading on some of our users posts here and other places recently and there seems to be a lot of misinformation, confusion, ignorance (not meant to be insulting but by definition), or mis-labeling going on.
First, lets get something clear. BOINC is not a project. WCG is a project. WCG does not mean or equal BOINC when tossing references around. Let me explain.... BOINC or rather Berkeley's Open Infrastructure for Network Computing (notice no mention of WCG) is merely the client and manager used for running various science apps. When you say BOINC in a very general manner, you are actually referring to over 80 DC projects. For those that have never stepped outside of FAH, that is like having 80+ different FAH's with very different approaches all deciding to use the same software that is freely available to them. Granted not all of the projects are like FAH. That is just to give you an idea how big DC'ing is outside of that very small community. Why do I say FAH is small? Because compared to SETI (the first BOINC project and for which BOINC was originally created) it is nothing but a dwarf in size. Now, I'm not saying the importance of the science at FAH is anything to snuff at, but rather there are quite a few who think it is the standard or the largest or some other hair brained conception. We can go into more details if people really need them, but I would rather go into addressing point structures in this post so that others can get back to arguing with PG or really FAH in general over the latest BA adjustments.
Each BOINC project handles points in their own way. Please don't for a second think you can really compare two BOINC projects point systems let alone compare them with FAH. FAH seems to hand out a great deal of points if you were to really compare MOST BOINC projects to them. Now, there are always exceptions. I will give a few examples of how vastly different the BOINC projects can be.
WCG and SETI are two of the largest BOINC projects out there. Their science is vastly different. They also have some of the lowest point rewarding systems in the BOINC community. SETI pretty much decides what gets put into the client and server software because BOINC is David Anderson's baby. SETI now uses the new point/credit system called Credit New which was meant to be used to level the playing field between all BOINC projects that used it and thus was a self adjusting credit system that levels itself out as time goes by. Most projects that tried it has abandoned it due to how terrible it is. WCG (last I heard) was still using a modified version of it. WCG also still has some old users with the UDP credits, which is why at WCG, you have 7 times the credit that is actually reported by BOINC. WCG has claimed that they plan on fixing this at some point, but like all of its promises, nothing moves very fast. I would point out it took 5 years to release their first GPU app, but as they have stated, that is up to the scientists not WCG. But, back on topic of points, they tend to give credits based on the performance of your device. Now, WCG does it a little different than SETI because you kind of get an average of what you claimed and what your wingman claims. So, someone hits the lottery while someone else loses a little. Since it is all within the same project, it pretty well evens out in the end.
Then you go over to Einstein (another large project that depending on which stat you are using may also dwarf FAH). At Einstein, they have both CPU and GPU work. But, depending on which app you are running depends on the scoring mechanism. One app has work units pretty well the same size so it is fixed credit. Other work isn't, so it is a weight and measure system. You will also find that depending on the science, your hardware, and other factors that certain apps will yield better point returns.
That brings us to another example of how a project can be drastically different. Prime Grid has both CPU and GPU work, but they also give bonuses based on a few factors. Obviously the GPU work yields the most points since it completes work units leaps and bounds faster than CPU work, but there are also other incentives. This past year, it was announced that they would start encouraging people to run the longer running work units by adding an ever evolving non-retroactive bonus for completing work on them. Why did this occur? Well simply put, if your science pays out roughly the same points/hour on the same hardware among all of your subprojects, people will run the short work units to minimize the chance of errors, failures, or missing deadlines. They also get a quicker return on their points and their RAC's are more consistent. So, how do you get people to run the more complex work? You give them bonus points for the extra sacrifice. It doesn't mean you are doing more science or better science. It means you are taking more risk and/or sacrificing. However you want to look at it. This also applies when the project is more focused on a specific work unit. They may give bonuses for each of those work units returned within the given time frame.
Then you have projects like DistRTgen that is always thrown out there like it is the black eye of the BOINC world. Its sole purpose is to crack security. Simply put, this makes a lot of people uneasy to support because the Rainbow Tables are free to anyone and therefore could be very easily used for illegal activity. Now, what does this have to do with points you may ask. Simple, to attract people to run their work, they had to reward the small number (compared to the rest of the DC world) of people that were willing to bring high end cards to the project. They "over reward" their contributors and so anyone who is trying to be in the top spot for points over all BOINC projects will run these work units to achieve that goal. Is this ethical? I leave you to decide.
POEM, this is another well known protein bio/science project that runs both CPU and GPU apps. They can't support the demand for GPU apps and thus run out quickly. However, people not only go there for the science, but they are very generous with their points for the GPU apps. Once you have run their GPU apps, you will find yourself not wanting to run CPU work units there which seem to be in abundance. Sounds like the SMP issue going on with FAH these days.
Another good example is GPUGrid.net. They award points per work unit, but also give you a bonus depending on whether to return it within 24 hours. This bonus can be up to 50% more. So, obviously a high end card will be leaps and bounds more beneficial for the sake of points and will create an even larger margin between the cards that can and the cards that almost can...
Now, I'm not going to run through every BOINC project and give all the ins and outs of their credit systems, but wanted to clear the air a little bit. Every project has growing pains and deals with people being unhappy with the credits in some way at some point. You even get irritated with how management runs the project. You just have to decide your limits. I can assure you that even WCG falls short in many areas and could have a lot of improvements made if management would simply act rather than stagnate.
First, lets get something clear. BOINC is not a project. WCG is a project. WCG does not mean or equal BOINC when tossing references around. Let me explain.... BOINC or rather Berkeley's Open Infrastructure for Network Computing (notice no mention of WCG) is merely the client and manager used for running various science apps. When you say BOINC in a very general manner, you are actually referring to over 80 DC projects. For those that have never stepped outside of FAH, that is like having 80+ different FAH's with very different approaches all deciding to use the same software that is freely available to them. Granted not all of the projects are like FAH. That is just to give you an idea how big DC'ing is outside of that very small community. Why do I say FAH is small? Because compared to SETI (the first BOINC project and for which BOINC was originally created) it is nothing but a dwarf in size. Now, I'm not saying the importance of the science at FAH is anything to snuff at, but rather there are quite a few who think it is the standard or the largest or some other hair brained conception. We can go into more details if people really need them, but I would rather go into addressing point structures in this post so that others can get back to arguing with PG or really FAH in general over the latest BA adjustments.
Each BOINC project handles points in their own way. Please don't for a second think you can really compare two BOINC projects point systems let alone compare them with FAH. FAH seems to hand out a great deal of points if you were to really compare MOST BOINC projects to them. Now, there are always exceptions. I will give a few examples of how vastly different the BOINC projects can be.
WCG and SETI are two of the largest BOINC projects out there. Their science is vastly different. They also have some of the lowest point rewarding systems in the BOINC community. SETI pretty much decides what gets put into the client and server software because BOINC is David Anderson's baby. SETI now uses the new point/credit system called Credit New which was meant to be used to level the playing field between all BOINC projects that used it and thus was a self adjusting credit system that levels itself out as time goes by. Most projects that tried it has abandoned it due to how terrible it is. WCG (last I heard) was still using a modified version of it. WCG also still has some old users with the UDP credits, which is why at WCG, you have 7 times the credit that is actually reported by BOINC. WCG has claimed that they plan on fixing this at some point, but like all of its promises, nothing moves very fast. I would point out it took 5 years to release their first GPU app, but as they have stated, that is up to the scientists not WCG. But, back on topic of points, they tend to give credits based on the performance of your device. Now, WCG does it a little different than SETI because you kind of get an average of what you claimed and what your wingman claims. So, someone hits the lottery while someone else loses a little. Since it is all within the same project, it pretty well evens out in the end.
Then you go over to Einstein (another large project that depending on which stat you are using may also dwarf FAH). At Einstein, they have both CPU and GPU work. But, depending on which app you are running depends on the scoring mechanism. One app has work units pretty well the same size so it is fixed credit. Other work isn't, so it is a weight and measure system. You will also find that depending on the science, your hardware, and other factors that certain apps will yield better point returns.
That brings us to another example of how a project can be drastically different. Prime Grid has both CPU and GPU work, but they also give bonuses based on a few factors. Obviously the GPU work yields the most points since it completes work units leaps and bounds faster than CPU work, but there are also other incentives. This past year, it was announced that they would start encouraging people to run the longer running work units by adding an ever evolving non-retroactive bonus for completing work on them. Why did this occur? Well simply put, if your science pays out roughly the same points/hour on the same hardware among all of your subprojects, people will run the short work units to minimize the chance of errors, failures, or missing deadlines. They also get a quicker return on their points and their RAC's are more consistent. So, how do you get people to run the more complex work? You give them bonus points for the extra sacrifice. It doesn't mean you are doing more science or better science. It means you are taking more risk and/or sacrificing. However you want to look at it. This also applies when the project is more focused on a specific work unit. They may give bonuses for each of those work units returned within the given time frame.
Then you have projects like DistRTgen that is always thrown out there like it is the black eye of the BOINC world. Its sole purpose is to crack security. Simply put, this makes a lot of people uneasy to support because the Rainbow Tables are free to anyone and therefore could be very easily used for illegal activity. Now, what does this have to do with points you may ask. Simple, to attract people to run their work, they had to reward the small number (compared to the rest of the DC world) of people that were willing to bring high end cards to the project. They "over reward" their contributors and so anyone who is trying to be in the top spot for points over all BOINC projects will run these work units to achieve that goal. Is this ethical? I leave you to decide.
POEM, this is another well known protein bio/science project that runs both CPU and GPU apps. They can't support the demand for GPU apps and thus run out quickly. However, people not only go there for the science, but they are very generous with their points for the GPU apps. Once you have run their GPU apps, you will find yourself not wanting to run CPU work units there which seem to be in abundance. Sounds like the SMP issue going on with FAH these days.
Another good example is GPUGrid.net. They award points per work unit, but also give you a bonus depending on whether to return it within 24 hours. This bonus can be up to 50% more. So, obviously a high end card will be leaps and bounds more beneficial for the sake of points and will create an even larger margin between the cards that can and the cards that almost can...
Now, I'm not going to run through every BOINC project and give all the ins and outs of their credit systems, but wanted to clear the air a little bit. Every project has growing pains and deals with people being unhappy with the credits in some way at some point. You even get irritated with how management runs the project. You just have to decide your limits. I can assure you that even WCG falls short in many areas and could have a lot of improvements made if management would simply act rather than stagnate.
Last edited: