Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
11)
Message boards :
Number crunching :
Solar Eclipse Challenge
(Message 84001)
Posted 3004 days ago by  HighTech67
I paused them
This is your problem.
You can't pause tasks for the project you are trying to download tasks for.
Resume those tasks and set you work buffer higher. You will then download new tasks. After the new tasks download, you can then pause the SoB tasks and the new tasks will run. You will have to do this each time you need to download new tasks.
|
12)
Message boards :
News :
Exact BOINC task timing
(Message 82212)
Posted 3071 days ago by HighTech67
When the challenge is over, I will set all computers for "No New Tasks", complete what tasks I have, then reset the project and download a couple of tasks to see if it changes anything before I add the tag in app_config.
There should be no need to put this tag in app_config.
I understand this and should have been more clear with my statement.
What I meant was I would reset the project under the conditions of my above post and try a couple of tasks to see that changed anything. If nothing changed after the reset, then I would add the tag, read the app_config.xml and see if that changed anything that the project reset did not.
I apologize for the confusion.
|
13)
Message boards :
News :
Exact BOINC task timing
(Message 82137)
Posted 3073 days ago by HighTech67
The functionality provided by that app_config flag is being sent out by the server itself, forcing all clients to use that behavior.
However, this will not work if:
A) You have an older BOINC client (prior to 7.3.18, I believe).
or
B) Your computer is already attached to PrimeGrid (which is very likely.)
To get around B), you need to either reset the PrimeGrid project, or detach and attach to PrimeGrid again. Doing either of those steps will delete any in-progress tasks, so only do that after setting "No New Tasks" and letting every task complete.
This will not help with the estimates of tasks that haven't started yet, but once they start running, the estimates will quickly become very accurate.
I haven't tried using the app_config version, but I suspect that this would work (this is what the server is sending):
<fraction_done_exact>1</fraction_done_exact>
Since app_config is used to provide information about individual apps, yes, you would need to do it for every app. Or just reset the PrimeGrid project (once all the tasks are done) and it should happen automatically.
Regarding A), I try to do my best to keep my clients up do date. I updated to the most recent version, 7.4.36, just before the challenge started and before that had been running 7.4.27. Prior to that it was 7.2.42 and before that 7.2.39. I keep the earlier versions just in case the newest one has some problem. Also, I am a hoarder. I guess I either deleted a few versions or missed the updates. Is there some way to be notified when new versions of BOINC are released? I have not been able to find anything like this.
About B), when I updated BOINC to the most recent version, I had only WUProp@Home tasks running. I do not know if it is required or not, but that is the way I always upgrade BOINC. I set all projects to "No New Tasks" and let everything complete. Except for WUProp@Home that is.
So when I downloaded my tasks for the Year of the Sheep Challenge, they were the first tasks to run on the new version of the client. And even after running for quite a while are very far off in their estimate for time left, especially on the one computer. All PSP tasks are off more than my ESP tasks also but that is probably be due to the length required to complete the tasks.
I did not however reset any project when I upgraded. That may not be a bad idea to "activate" any new features in the client software when installing a new version. It also might not be a bad idea for us to reset the clients whenever new BOINC server software is installed. I'm not saying for sure that it will do anything, but as long as there are no tasks downloaded for the project, it doesn't hurt does it?
When the challenge is over, I will set all computers for "No New Tasks", complete what tasks I have, then reset the project and download a couple of tasks to see if it changes anything before I add the tag in app_config.
I read and understand your first post in this thread, but new tasks that have not started will never start showing a more accurate estimate? Do you know if this is in the development stages or anything?
Again, thanx for the help.
|
14)
Message boards :
News :
Exact BOINC task timing
(Message 82135)
Posted 3073 days ago by HighTech67
It seems as if enabling this feature may require that you either reset or detach/attach the PrimeGrid project from your BOINC manager. Of course, never do that while you have any PrimeGrid tasks on your computer or they will be lost!
I have a couple of questions about this.
First let me state that I am by no means an expert nor even what I would call an intermediate user of BOINC, I enjoy furthering the science by donating idle CPU time. I know what configuration works for me so far and often get confused when trying too change things to much at a time so I ask a bunch of questions that are obvious to other people.
But, I do want BOINC to be more accurate in giving me time estimates, much more accurate in some cases. I have one computer, 421177, that even after completing a few PSP tasks think before they start they are still going to take over 9 days to complete and almost 5 days to finish when it took them 2 days and 7 hours to get to a little over 50%. My other computers are also off but not by that much.
Now to the questions.
1) When you say feature, are you talking about the tag in app_config.xml or is there some option you can check in the BOINC Manager to enable this also?
2) I have looked at http://boinc.berkeley.edu/wiki/Client_configuration and from what I can understand from there, the actual tag is <fraction_done_exact/>. Is this correct? Basically, does it need the forward slash at the end is what I am asking in case it didn't show up in the post?
3) It is also my understanding from the above site that this tag has to be entered for each <app> section it the app_config.xml file. Is this also true?
Thank you for the help.
|
15)
Message boards :
Number crunching :
Shakespeare's Birthday Challenge
(Message 75698)
Posted 3333 days ago by HighTech67
So i've had my 2 pc's only doing PPS tasks for the last few days will they automatically be part of the challenge or do i have to do something special to be counted
They will automatically be in the challenge.
But, if you have a large amount of workunits already downloaded before the start of the challenge, they will NOT count. Only workunits downloaded and started after the official start of the challenge are counted.
|
16)
Message boards :
Number crunching :
Haswell AVX2 version of LLR available
(Message 75644)
Posted 3334 days ago by HighTech67
It only affects tasks sent out after the change is made on the server. If the task has already been sent to your computer, even if it's not started running yet, you'll run the old app.
Michael,
Thank you for your rapid reply.
One other question, will the old apps somehow self-delete? Or will they stay on everyone's computers taking up space for no good reason after all tasks sent before the release of the new version is complete on said computer until everyone does a project reset on each of their computers?
|
17)
Message boards :
Number crunching :
Haswell AVX2 version of LLR available
(Message 75639)
Posted 3335 days ago by HighTech67
The rollout of the new LLR on PPS went very smoothly, so we're going to install the FMA3 LLR on the remaining LLR projects tomorrow rather than waiting until after the challenge.
I expect to install the apps sometime early on Sunday April 20th (UTC), which means Saturday evening or afternoon in the western hemisphere.
Will this effect any currently running tasks?
I have two computers running three Genefer CPU tasks each and I know that at least four of the six will not by done by then, maybe more. I would hate to lose all the time I have spent on these tasks.
|
18)
Message boards :
Generalized Fermat Prime Search :
GFN (Short) GPU vs CPU use
(Message 71634)
Posted 3447 days ago by HighTech67
Short answer: Don't worry about the CPU usage of the OpenCL app.
The CUDA driver doesn't do this by default, but I think it has an option to work in the same way. I don't like this mode because I need to repeatedly answer this question. :)
Thank you for explaining this. I will no longer worry about this then and will run the OpenCL app on both my 660 Ti and 650 Ti cards. I will also run the CUDA app on my GTX 460 card to see if it can complete a short task.
|
19)
Message boards :
Generalized Fermat Prime Search :
GFN (Short) GPU vs CPU use
(Message 71624)
Posted 3447 days ago by HighTech67
I ran my first four GFN GPU tasks in prep for the upcoming challenges last week and noticed something I thought was odd.
I have two computers I think will be able to run GPU tasks for these challenges. One (JTMD4) has an NVIDIA GTX 660 Ti and the other (JTMD3) has an NVIDIA GTX 650 Ti.
On both of them, when running the OpenCL app, CPU use was over 95% for the entire time. For the 660 Ti, task length was 12:12:13 with a CPU time of 11:40:42. For the 650 Ti, task length was 21:29:10 with a CPU time of 20:38:57.
When running the CUDA app, CPU use was under 3.5% for the entire time. For the 660 Ti, task length was 15:57:37 with a CPU time of 00:32:02. For the 650 Ti, task length was 01 d,01:42:35 with a CPU time of 00:29:34.
This means I definitely have shorter run times with the OpenCL app. I am just curious about the CPU use. I let the project dictate everything. I made no changes to block size nor in my app_config.xml file.
Is this normal behavior?
|
20)
Message boards :
Number crunching :
Winter Solstice Challenge (2013)
(Message 71478)
Posted 3456 days ago by HighTech67
Hello,
I have a problem getting GPU work for my AMD HD7970....
I re-checked several times configuration on PG and I have set even new profile to try out with and without CPU, with and without Nvidia, eventhough I have never used Nvidia card on PG....
I tried other projects, SETI and Milkyway and they are goiving me work immediatelly without problems, all other projects are stopped....
I tried to play with cc_config and client_state and no success it downloaded CPU tasks immediately, but no way to force it to download GPU
It requests them and message follows:
Requesting new tasks for ATI GPU
Got 0 new tasks.
With enabled work_fetch_debug not any error or problem visible....
Do you have any ideas guys please ?
EDIT:
Sorry I forgot to write....Core 6.12.34+Win 7 x64
You may want to try my solution in Message 71443 earlier in this thread as I was having a similar problem.
|
Next 10 posts
|
