| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: Credits, Points, CPUs, GPUs, Badges ... |
| No member browsing this thread |
|
Thread Status: Locked Total posts in this thread: 21
|
|
| Author |
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
If anyone here is competent to assess the g-pops in ratio to f-pops or i-pops, s/he needs to speak up. Otherwise, grant same points per GPU hour as the pool of CPU's they're attached too (If that is actually measurable). This way a GTX260 result coming off a P4 host provides equal credit to sitting in an I7. Then take that credit mix of CPU's and apply that as were the card a core. Good point. My question would be does a person who runs their P4 for an hour make less of a personal commitment than the person who runs their i7 for an hour? And they less than the person who runs their GPU for an hour?  |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges:
|
Since all the internal World Community Grid projects are centrally controlled, allocating equal credit is easy. This is not the case with other, autonomous BOINC projects. Each project at World Community Grid earns (on average) exactly the same credit/CPU second. But did you not say each science should choose? Are you in favor of a central authority over all? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Dataman, our projects here are not autonomous BOINC projects. The scientists are independent, but the server management and project hosting is done entirely by World Community Grid. The scientists have no say in how much credit is awarded. Additionally, the WCG projects are not competing against each other. They are all guaranteed enough computing power to get their task done, as agreed in the planning stages.
This is different to other, non-WCG BOINC projects. |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges:
|
Dataman, our projects here are not autonomous BOINC projects. The scientists are independent, but the server management and project hosting is done entirely by World Community Grid. The scientists have no say in how much credit is awarded. Additionally, the WCG projects are not competing against each other. They are all guaranteed enough computing power to get their task done, as agreed in the planning stages. This is different to other, non-WCG BOINC projects. Yes, different than some but not all. There are others that host independent science projects. YoYo does it remarkably like WCG and their projects each have their own credit structures. Strangely no one there seems to care they are different ... but then my German language skills are poor. We all know how WCG works and it works well. I am more interested in the DC world at large. I would be interested in what WCG admin thinks too as they will have to face this when they implement a GPU project. Honestly, I am not trying to start a "Greenies Thread". I am perplexeed by these issues. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
yoyo@home hosts non-BOINC projects, so really there is no basis for comparison at all.
|
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
When someone living on a part-time minimum wage job spends twenty dollars to buy a good hot meal and a good jacket at the thrift store for a homeless friend and when someone with a six figure income with a home in the tony part of town gives seventy five thousand dollars to a local commmunity theater group that means they can survive for the season, who contributed more? How do you decide that objectively? To objectively compare DC projects, you need an objective measurement. Objectively measuring crunching will have to look at the cruncher's machine. By not looking at the project though, you do not account for the complexity or unique issues behind the science of that particular project. But can you do that and stay objective? It is in the best interests of the DC world to try to keep the credit issue reasonable and fair across projects. No, there's not a nice clear solution but the projects need to work together as best they can on as good a solution as they can find. GPU crunching is and will be intense fuel to this fire. I don't know if BOINC itself should provide some standardized benchmark for cross project comparison that projects can adopt or reject (which would make it pretty evident who did and didn't adopt it) or simply the various prohects reaching some sort of common understanding about setting their project's credits multiplier. However it is or is not addressed, the issue should be number one on the list at the next BOINC conference.
---------------------------------------- |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7854 Status: Offline Project Badges:
|
What I was interested in discussing is how the disparity could be changed ... if at all. Sorry Dataman, misread your intent.  Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Makke Beer
Cruncher Joined: Apr 29, 2009 Post Count: 11 Status: Offline |
I would vote for cross-Boinc-project-credit-comparison based on CPU/GPU-time. It's a simple method that has it's drawbacks, but frankly it's the only managable solution I can come up with. In fact, the WCG-badges are based on that, and I don't remember reading any complaints about is (could be wrong, I don't read everything).
----------------------------------------So for example: 24 hours of CPU-time equals 100 points 24 hours on a quad-core equals 400 points 24 hours on a quad-core + GPU equals 500 points I agree with the topic starter that it's getting out of hand lately with the current credit system, with projects rewarding any amount of credit to lure crunchers. Note: again this post is about comparing credit of Boinc projects, not about comparing credit of WCG projects, which works fine. |
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Here is my two cents about credit:
1) Credit is intended to provide the same 'reward' to volunteers based upon performing the same amount science computations 2) It is very hard to determine how much work has really been done since our projects have workunits that vary in difficulty. 3) Benchmarks are tricky (notice the debates and issues surrounding all benchmarks). They need to actually measure the same thing as the work that is being done. For example, our projects vary in resource use. HCMD2 uses very little memory compared to The Clean Energy Project. Thus if you have a quad core chip with poor memory bandwidth, you are going to run CEP1 slower then you will HCMD2 because it will spend more time waiting for data from main memory. A benchmark designed to capture the performance characteristics of HCMD2 will not be appropriate for CEP1. 4) A lot of time has been spent on trying to be as a fair as possible. An infinite amount of more time could be spent. As far as cross-BOINC project parity of points. I do not believe it is possible anymore. It would require a high degree of central control from BOINC in order to enforce any rule that could be devised. Any rule that could be devised would almost certainly have issues since the resource use varies so dramatically between BOINC projects. The discord that this would create would be very destructive to BOINC community. I think that the sites that do cross-project aggregation of BOINC stats should look at alternate metrics (such as rank on a given project) rather then strictly credit in order to provide global rankings. As I said, just my 2 cents. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
2 cents made of solid gold.
Credit disputes have already caused all sorts of silly feuds in cross-project BOINC circles. And at least one enterprising stats site already uses ranking as the basis for cross-project comparison. It works very well. |
||
|
|
|