Main Menu
| Home |
| Getting started |
| Benefits to humanity |
| Simulation software |
| Screensaver graphics |
| Current simulations |
| Results gallery |
| Security and safety |
| About us |
| Mailing List |
Participants
| Your account |
| Teams |
| Add-ons |
| Applications |
Community
| Participant profiles |
| Message boards |
| Questions/answers |
Statistics
| Top participants |
| Top computers |
| Top teams |
| Other statistics |
|
Questions and answers Talk live via Skype with a volunteer, in any of several languages. Go to BOINC Online Help. Check the FAQ, select a topic, or do a keyword search to find what you are looking for.
Frequently Asked Questions1: General1.1: What's the credit policy around here? 1.2: I have tons of Pending credit, why is that and when will it be fixed?! 2: Platforms and BOINC Managers2.1: Will you provide optimized AMD clients? How about Linux or Mac? 3: Nanofactory_13.1: The completion time for these WUs ranges from a bit off to way off. What gives? 3.3: When does the NHAH_QSMAKAS app checkpoint? 1: General1.1: What's the credit policy around here? It's described in this Project Characteristics section. 1.2: I have tons of Pending credit, why is that and when will it be fixed?! We understand your unhappiness over some seemingly ever-pending credit. We've done some things to shorten the time that some credit spends in the Pending state (by lowering the maximum number of results sent out per WU, etc). We could increase the number of results initially sent out per WU, which would shorten Pending times, but then there would be too much unnecessary redundancy. It's been a challenge to harmonize the BOINC distributed computing platform with the experiments we're trying to complete, with the BOINC participant expectations, and we've had to ask the participants to share some of the pain. Heck, in the first NHAH project the client app didn't even have a representative progress number - it was just pegged at 50%, mostly because it's near impossible to calculate one. That caused havoc with the BOINC Manager's scheduling mechanisms though, and we (more or less) fixed it for this project. That some credit remains Pending for a long time is something that isn't easily fixed without changing the way that the BOINC server code decides what result to send to who and when. I think that modification is going to have to wait until the next NHAH project. 2: Platforms and BOINC Managers2.1: Will you provide optimized AMD clients? How about Linux or Mac? We plan to support Linux and Mac in addition to Windows. We also plan to do optimized clients. They will come after the ports for the Unix platforms, ie, once we get everyone's platform basically supported, we'll start optimizing them. 2.2: My NHAH WUs don't seem to be processed correctly on my computer - they're all only taking a minute, and I'm hitting the 50 WU limit all the time. By the way, I'm using an alternate BOINC Manager. The NHAH client application doesn't work properly with some alternate BOINC Managers (like Crunch3r's 64 bit build) on some computers. Try the latest official build of the BOINC Manager. 3: Nanofactory_13.1: The completion time for these WUs ranges from a bit off to way off. What gives? The problem is that it is very hard to predict how long any given WU will take. Nanofactory_1 WUs typically take between 2 and 8 hours, but can take as little as 40 minutes and as much as 19 hours. This is because of how the WU is calculated; it is basically searching for a set of parameters that minimize the total energy of a molecule. If its initial parameter guess is close to the minimum, and the WU doesn't have too many atoms, it can just take an hour to hone in the minimum. Less ideal initial conditions can take 8 or more hours to converge. Here's another way to look at it: Imagine a planet. It takes about 4 hours to walk pole to pole if it's smooth. If it has mountains, it may take 24 hours to traverse. You start blindfolded at the north pole and must find a certain pebble on the planet by feeling around with your bare feet. You are told that the pebble lies at the bottom of a gully on the planet. If you're lucky, the pebble will be quite near to you and you'll stumble down the hill to find it in about an hour or so. If you're unlucky, the planet will have many many gullies, and you'll stumble down gully after gully only to find no pebble. Eventually, if you haven't found the pebble in the time allotted, the planet blows up. 3.2: This WU took me over 6 hours and yet another guy it only took about 40 minutes. How does that work? While 99% of all WUs will process with the same calculations, there can be an element of randomness to the searching algorithm that can cause a wide gap in processing times. 3.3: When does the NHAH_QSMAKAS app checkpoint? It checkpoints at each optimization iteration. An optimization iteration is made up of several energy calculation iterations and checkpoints can be up to an hour apart. You can determine when a WU has checkpointed by examining the screensaver graphics. The screenshot at the top of the Screensaver Graphics description page shows an example of when a checkpoint occured. We realize that longer times between checkpoints is not optimal and are working on an app version with a higher frequency of checkpointing. 3.4: My WU crashed after crunching for hours and reported an error! Did I just waste hours of computing? There are two main reasons for a WU stopping and reporting an error. The lesser likely of the two is that there was an error in the client application. The more likely of the two is that the WU didn't converge in some not-nice way, ie, some internal criteria during computation wasn't met, and the WU is basically giving up its search. In both cases however, you will get credit for crunching the WU from one of the scripts roaming the database adjusting credits. 3.5: I have a result that is sitting in pending since forever and a day. It's been sent out to multiple computers and it's just not validating. What's up with that? Due to the nature of the science of the project, some WUs will not converge, or find a minimum energy configuration for the given tooltip deformation. This is no cause for panic or concern. Sometimes a WU will converge on one hardware configuration and not on another. This has to do with differences in how the various CPUs process floating point numbers. That is why we send the WU to several hosts in hopes that some computer will crack it. And it takes time for other hosts to try and converge the WU, potentially fail, then have the WU sent to another host, so your claim ends up pending for some time. Once either quorum is reached for the WU, or it fails to converge a certain number of times, a script processes all the claims for that WU and grants credit for them.
| ||||||||||||||||||||||||||||||||