Message boards :
Number crunching :
Completed, marked as invalid
Message board moderation
Author | Message |
---|---|
Send message Joined: 24 Apr 10 Posts: 4 Credit: 77,216 RAC: 0 |
Most of my WU's are marked "Completed, marked as invalid" But also marked that "granted credit = 25" Is this a validator problem or is it really invalid http://wuprop.boinc-af.org/workunit.php?wuid=88275 I have 3 out of 40 plus completed Thank, Steve |
Send message Joined: 4 May 10 Posts: 8 Credit: 238,189 RAC: 0 |
Just curious, what other projects were you running? |
Send message Joined: 24 Apr 10 Posts: 4 Credit: 77,216 RAC: 0 |
Nothing on CPU Boinc, But running PRPNET on all cores from the Project Staging Area on PrimeGrid. http://vcn94.homelinux.org/PRPNet/user_search?uid=Steve_Martin@SETI.USA I also ran Primegrid's Earth Day Computation. End of April. Also run GPU WU's from Gpugrid, Milkyway, and Collatz Conjecture. |
Send message Joined: 4 May 10 Posts: 8 Credit: 238,189 RAC: 0 |
Nothing on CPU Boinc I suspect that's the reason, I don't think this project looks at GPU work units yet. |
Send message Joined: 17 Apr 10 Posts: 8 Credit: 94,138 RAC: 0 |
Nothing on CPU Boinc It must, I only run Collatz, MW, GPU Grid and DNETC on my GPUs (5 systems, including OS-X), and it has recorded data on these projects ... of course, it does not distinguish these tasks as being run on the GPU, it records them, apparently against the CPU for purposes of recording. So, there will be data skew in the sense that some CPU tasks for Collatz (for example) on OS-X will be recorded as running in `58 minutes and yet the same class task on the same class CPU will also be recorded as taking several hours ... depending on if the participant is running them on the CPU side or on the GPU (as I am)... Likewise, when I was still running MW on my CPUs they were taking I think up to 4 hours to run on an i7 920 ... yet, same i7 I have is going to record them at 2 minutes on one system and 5 minutes on another, same class CPU, yet the data will be spread from minutes to hours all because the tasks can be run on the CPU, or in my case on HD5870, HD4870, GTX295, or GTX280 ... so, in that sense the data is there, but, the data is not properly collated and can give misleading information ... On "pure" CPU only projects that are running on a class of CPU the data can be trusted with the usual caveats ... Then again, over time, they may figure out how to capture enough data to start to segregate the data by GPU as well ... of course there the issue will be that there will still be a spread because of the number of people that have factory OC cards and those that push those limits even higher ... |
Send message Joined: 24 Apr 10 Posts: 4 Credit: 77,216 RAC: 0 |
So, I assume to get any credits on this project, I need to run all the CPU cores at 100% on Boinc CPU Projects. Am I right in what I am reading? Steve |
Send message Joined: 28 Mar 10 Posts: 2871 Credit: 538,601 RAC: 132 |
I looked results you had returned, there was only data from this project. So your results was invalidated. Project collects only data from CPU WU. |
Send message Joined: 24 Apr 10 Posts: 4 Credit: 77,216 RAC: 0 |
I looked results you had returned, there was only data from this project. So your results was invalidated. See if I get this right. If I set a quad core to run 25% of the processors, (1 core) run 1 wu from wuprop, 1 wu from a gpu only project 25 wu's from freehal, 1 wu's from a cpu only project and non boinc projects on the remaining 3 cores. Would this work? Steve |
Send message Joined: 2 Apr 10 Posts: 8 Credit: 108,152 RAC: 0 |
As I see it, yes. All info of FreeHAL, GPU projects, and the single CPU project will be taken into the database, and with the cpu-project it will be OK for the validator. For overheating purposes I only use 3 of my 4 cores (I really HATE these CPU-fans provided by AMD) and for me everything is OK, so it does not count if you use 1 4 or 16 or zillion CPU cores as long as you use at least ONE for a CPU project. |
©2024 Sébastien