Posted 418 days ago by
Ageless
At what point do you cut off the Atom netbook pc's that do have a dual core but only a 1.6ghz one that can take an eternity to crunch a unit here.
Or on a similar note, at what point does the project set the requirement to crunch 24/7/365?
The problem with asking the user-base in the forums what they think of such things is that not everyone who has attached to your project is active in the forums. So while in a thread like this 90% may tell you they're OK with it, the question is if everyone attached to the project and crunching for it is agreeing with that.
There are other options as well, such as the use of the reliable host option for config.xml. Only reliable hosts get the longer running tasks, those with a lesser reliability status will get the shorter work.
A smaller quota may help as well, since every task aborted or run over the deadline will decrease the amount of work any machine may download. A person aborting enough work will eventually have to sit through a task to get the quota back up, or be looking at no work for (at least) a day. Those with big machines with a plethora of CPUs have already said they run the 200-qubits in a lot of time as well, so is there really a reason anyone needs to be able to cache 100 tasks per CPU per day?
BOINC is volunteer computing, giving projects the chance to run their complex computations on the unused CPU cycles of any computer, without needing to pay the people helping out any money. As soon as you set strict requirements on the hardware used and possible the time it should be used, where does that leave the thought behind BOINC? You could just as well go and construct your own supercomputer then and run your project on that only.
It feels discriminatory. Segregation based on computing capabilities. I'm not sure that's a path you want to turn into. And all that based on a few who are essentially cherry-picking their work or whose computers crash and lose the work. For the latter there's an option available in the back-end.
Neo wrote:
(e.g. because the system crashed and somehow BOINC lost the units, which happens)
Turn on the <resend_lost_results/> option in config.xml and people will get that work resent when they next contact the project (assuming their hostID number stays the same).
Posted 420 days ago by
Ageless
You've signed in on BOINC Wide Teams, so any project that imports those teams will already use your email address, making it impossible for you to make the account again with that email address.
So the thing to do is to use the "remember password" option on the login feature, then fill in the email address you used to register your team on BWT and you'll get a link sent in that email address to activate your account and set a password.
In other words, it was you all this time who had this account. :-)
Posted 599 days ago by
Ageless
Does this mean that the reboot had not erased the RAM?
Any reboot erases the RAM as for a millisecond your system is without power. Enough to flush the RAM.
It'll show 60% progress, but I'll tell you that when it restarts it'll do so from zero percent. I've seen the same things on my Windows system, after exiting BOINC and restarting it.
Posted 600 days ago by
Ageless
Help is on its way!
You gonna get us the Little River Band? OK, I'll be there. :-)
Posted 601 days ago by
Ageless
The AQUA application doesn't do checkpointing, which your other project's applications will do. By checkpointing the application writes its state to disk every so many minutes, making it possible to stop the application at any time and continuing with it at a later time, it'll then continue from the last checkpoint.
But since AQUA doesn't do that, BOINC will run it from start to finish without letup. If you feel you need to break in and allow another application to go first, make sure you have your preferences set for leave applications in memory, before suspending the AQUA task, or else you'll need to start all the way from the beginning at the restart of the application.
Of course, all bets are off when you stop BOINC. Then all applications will be unloaded from memory.
Posted 604 days ago by
Ageless
That means that all Windoze users will lose X hours of work.
Only those that have auto-updates on and allow their system to reboot upon receipt and install of those updates. Not all do. I sure don't. So it won't be all Windows users who have to recrunch their work.
Posted 609 days ago by
Ageless
My P4 3.0GHz HT eventually took 23 hours, 38 minutes to complete its 96qubit. You're not hearing a complaint from me, it's still better than working on an Orbit task (which does checkpoint) that resets to zero percent after it had done close to 90% of its work, in only, oh, 800 hours? ;-)
Posted 610 days ago by
Ageless
My AMD 2200XP+ took 48,562 seconds on its 96qubit. So a little under 13.5 hours. My Intel P4 3.0GHz HT seems to be taking 25 hours on its 96qubit, seeing that it was at 21% after 5 hours of running. Not worried about it, it keeps the CPUs warm in this otherwise freezing flat. :-)
Posted 611 days ago by
Ageless
As per this post:
Boinc Admin wrote:
Check pointing is on the high priority list. It will be implemented in the new re-write of the AQUA application.
So it's coming, no need to stress out or be obnoxious over it.
Posted 614 days ago by
Ageless
After running the benchmarks the apps resume normally. No computation is lost.
This depends on BOINC version. Some early 5 versions don't have this option. I was looking for the version in which this changed, but haven't found it yet. Will keep digging.