Posts by Fardringle

1) Message boards : Number crunching : credit/points to low? (Message 235)
Posted 4 Mar 2023 by Fardringle
Post:
Yes, credits on this project are rather low. Not the lowest rewards for a BOINC project, but pretty close.
2) Message boards : Number crunching : New app HEX (Message 202)
Posted 8 Nov 2022 by Fardringle
Post:
We are testing a new app.
The first few test runs were canceled at some stage.
We found errors therefore we canceled.
The current series is TESTS_B_1M_[X] and according to the latest information there are no bugs found.
However, it is still a test version of the WU series.

This application has Quorum=2.

Additional information about the iThena HEX application will appear later.


It looks like the tasks are all completing successfully for the new app on my Xeon E5-2967 v3 running Linux Mint 20 since the new version was installed on November 5th.

However, they are taking about 50% longer than the previous GRAPH tasks for the same number of credits.

And you really should put the option back to allow people to choose which applications they want to run so that individuals can decide whether or not to run new/test applications.
3) Message boards : Number crunching : Out of Work (Message 134)
Posted 4 Dec 2021 by Fardringle
Post:
Yup. It looks like we are out again.
4) Message boards : Number crunching : XML data has partially disappeared (Message 108)
Posted 13 Nov 2021 by Fardringle
Post:
I was wondering why the stats exports were not being updated.

There should be a news post/announcement letting everyone know that this feature has been enabled, and each person needs to go to their account preferences to allow their stats to be exported.
5) Message boards : Number crunching : Lots of errors on one of two machines (Message 99)
Posted 11 Nov 2021 by Fardringle
Post:
I find I need to stop the Measurements application then run the Computation application otherwise I get immediate Computation errors. When it is running I can un-suspend the Measurements tasks.

I haven't seen any problems running the .Measurements and .Computation projects at the same time on Windows and Linux computers.
6) Message boards : News : iThena GRAPH: windows_intelx86 and windows_x86_64 (Message 98)
Posted 11 Nov 2021 by Fardringle
Post:
Well over 100 tasks completed so far on my Ryzen 9 3900X running Windows 10. Zero failures/errors.
7) Message boards : Number crunching : Somewhat limiting.... (Message 37)
Posted 1 Nov 2021 by Fardringle
Post:
I've noticed problems.
I increased 10x the possible execution time.
I set it to 40 credits per WU at this moment.
There is a new specification for tasks in the new series "TESTS_1A_T..."


Thank you for increasing the time limit and the credits. That will allow more computers to complete their tasks successfully so that you can see what the time limit really needs to be based on actual completion times.
8) Message boards : Number crunching : Somewhat limiting.... (Message 36)
Posted 1 Nov 2021 by Fardringle
Post:
hello
on one of my PCs
Intel(R) Core(TM) i7 CPU 870 @ 2.93GHz [Family 6 Model 30 Stepping 5]
(8 processors)
computer 94
I received 152 WU at the same time, is this normal?
thank you


That is probably because the initial estimated run time for the tasks is only about 15 minutes, even though they actually take several hours to complete.
9) Message boards : Number crunching : Somewhat limiting.... (Message 29)
Posted 1 Nov 2021 by Fardringle
Post:
On a side note, based on the task in the link in the "First Success" post made earlier, I realize that this is still the basic testing stage, but if you want people to continue to participate in the new CPU intensive project, you need to give a LOT more than 8 credits for CPU tasks that take several hours to complete!
10) Message boards : Number crunching : Somewhat limiting.... (Message 27)
Posted 1 Nov 2021 by Fardringle
Post:
Agreed. It seems that these tasks have a very wide range of run times depending on the computer, and your estimate and max run time limit are too low.

Tasks running on my Xeon E5-1650 v4 are estimating run times of around 3.3 - 3.5 hours. And tasks running on my Ryzen 9 3900X are estimating run times of more than 5 hours. Both of these computers are normally a lot faster than the Xeon E3 1220 test machine mentioned in the announcement message, but are not running the .Computational tasks that fast and need to have more time allowed to finish the tasks.

Also, these tasks are actually using more than one CPU thread per task. On the E5-1650 with 6 cores/12 threads I had to reduce the number of running tasks down to 8 before the total CPU usage dropped below 100%. On the 3900X with 12 cores/24 threads, I had to drop it down to 18 running tasks to get the total CPU usage below 100%.

If the tasks were actually just using one CPU thread each, the totals would have dropped below 100% at 11 tasks on the E5-1650 and 23 tasks on the 3900X...

This might explain the higher run times than the original estimate as well. For those of us who usually run a number of tasks equal to the number of CPU threads, the tasks aren't getting the amount of CPU time that they want.




© 2019-2024 iThena. All rights reserved. | Private Policy

Page generated on 5 Mar 2024, 5:25:27 UTC in 0.1966 seconds.