Statistics Screen Shows No Progress Since February 11
1 Mar 2008 17:38:17 UTC
Topic 193544
(moderation:
Despite the fact that the message screen shows that tasks are being completed and reported. I upgraded from BOINC client version 5.4.11 to 5.10.30 on February 24 because I was getting messages about no response from the screen saver and the task exiting with status zero but with no result file. After the upgrade, the benchmark results reported 685 FP MIPS and 1283 Integer MIPS. On February 29, the benchmark results reported 425 and 740. It seems unlikely that my computer slowed down this much.
I just did an Update and my statistics screen jumped. The slope for the last day is about what it was before this process started, but the intervening 18 days have a horizontal slope.
Checking my tasks, I just saw two tasks reporting client errors on February 17 and 20. Perhaps the upgrade resolved the original problem. I'm still curious why the statistics screen required a manual Update to show the latest progress and what happened to the benchmarks.
Statistics Screen Shows No Progress Since February 11
)
Are you still running the screen saver? If so, you might want to turn it off. E@H was freezing on my Windows NT box and giving compute errors until I turned the screen saver off. Once I did that, E@H had no problems. In addition, the screen saver was slowing the computation rate, so once I turned it off, there was a speed increase. Granted, that was back in 2006, but I still leave the screen saver off, even with the most recent versions of BOINC.
Drop in benchmarks can signal
)
Drop in benchmarks can signal the processor slowing down due to heat. Do you need to clean out the dust bunnies?
Kathryn :o)
Einstein@Home Moderator
RE: Despite the fact that
)
The benchmarks ran again on March 5. The numbers were 677 and 1193. I have not done anything (such as cleaning out my computer). The room temperature does not vary widely.
After the benchmarks ran, my messages screen shows the following messages:
3/5/2008 7:15:53 PM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, requesting exit
3/5/2008 7:15:54 PM|Einstein@Home|[error] einstein_S5R3 not responding to screensaver, killing it
3/5/2008 7:15:57 PM|Einstein@Home|Restarting task h1_0909.40_S5R3__364_S5R3b_0 using einstein_S5R3 version 426
3/6/2008 2:16:39 AM|Einstein@Home|Sending scheduler request: To fetch work. Requesting 172 seconds of work, reporting 0 completed tasks
3/6/2008 2:16:44 AM|Einstein@Home|Scheduler request succeeded: got 1 new tasks
3/6/2008 5:40:10 AM|Einstein@Home|Computation for task h1_0909.40_S5R3__364_S5R3b_0 finished
3/6/2008 5:40:10 AM|Einstein@Home|Starting h1_0909.40_S5R3__332_S5R3b_1
3/6/2008 5:40:11 AM|Einstein@Home|Starting task h1_0909.40_S5R3__332_S5R3b_1 using einstein_S5R3 version 426
3/6/2008 5:40:12 AM|Einstein@Home|Started upload of h1_0909.40_S5R3__364_S5R3b_0_0
3/6/2008 5:40:17 AM|Einstein@Home|Finished upload of h1_0909.40_S5R3__364_S5R3b_0_0
Despite the fact that my result was uploaded, and my on-line task list shows that I received credit, my local statistics screen has flat-lined again from March 1 to March 6, even after an update.
I just did a second update. Now my graph shows a slope between March 1 and March 7 of approximately half of what it has been in the past. Perhaps the low benchmark results above affected the amount of work done as calculated from the processor time.
Another question: Why was the amount of work requested only 172 seconds? It should probably have said 173 hours based on the task statistics of the work unit that was actually assigned.
RE: The benchmarks ran
)
Benchmark numbers are not very reliable and can easily be affected by other tasks that are running when BOINC decides to recheck them (every 5 days). For projects like E@H, benchmarks are not really important as credit is assigned server-side and not by the old benchmarks and CPU time method. If you see suspect benchmarks you can always force a manual recheck through BOINC Manager.
Benchmarks play no part in the actual time taken to crunch a task. They are only used as part of the process in making an initial estimate of crunch time for cache control purposes only. The actual CPU time will always be recorded and this will be used to refine future estimates through a duration correction factor (DCF).
I never run the screensaver so I have no knowledge of any associated messages. My philosophy is that the screensaver is just a potential CPU cycle stealer so I always set the screensaver to be a blank screen. For your machine, a P4 1300, disabling the screensaver would probably give you a significant improvement in production rate.
These are totally normal messages and not indicative of any problem. As soon as BOINC notices that your work on hand is even just 1 second less than what your preference states, a request for new work will be issued. In the above example it was almost 3 minutes before BOINC noticed. The scheduler, on receipt of a request for work will honour that request by sending a full new task. It cannot send anything less. Of course, having received a long running task, your local BOINC client will not be making any further requests for quite a while.
Statistics are NOT realtime. It may take anything up to a day for a statistics update to occur. You need to allow for this in your expectations.
By the time of your second manual update, a stats update had obviously occurred. The slope of the line will be entirely dependent on the validated credit your machine has produced in that interval. You may have produced lots of results but if they hadn't been validated before the stats update you wouldn't see these credits reflected in the stats. Benchmark values have absolutely no influence on your machine's ability to produce results and consequently no influence whatsoever on your stats.
As previously explained, the BOINC client will only request whatever is the discrepancy between your cache setting and your actual estimated work on hand. The server, however, will always send you a full task.
You seem to be concerned about how slow your machine is. If you want to be more productive there are three things you could consider:-
* Check out the latest power user optimised app - version 4.36,
* Replace your P4 with a Tualatin P3 :-).
Well, at least any two of the above.
Here is the results list of one of my Celeron 1300s chosen at random. Compare your results list with what my machine is producing and you might understand how bad the low end P4s really are.
Cheers,
Gary.