Why did my screen saver freeze?

I'm running BOINC 5.4.9 on Windows XP Professional, SP2. I recently told BOINC not to accept new work (because I'm going on vacation and will shut the computer down; I don't want to leave Rosetta waiting for results from me). This morning I came back to my computer and found that the screen saver had frozen - I couldn't clear it. I had to cold boot the system, not a good solution. When everything came back up, I found that BOINC had finished its last WU and was waiting to report. Is this something I can avoid?? I really don't want to have to power down to get past the screen saver.

---

--hedera

Never be afraid to try something new. Remember that amateurs built the ark. Professionals built the Titanic.

Why did my screen saver freeze?

(Hedera)
Hedera Welcom to Rosetta. I'm not sure on that one. There have been some problems with the screen saver lately. Did the work unit complete successfully or did it report errors?

You might bring it up in the thread for issues with the Rosetta version you were running (probably 5.25)

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

How common is credit cheating in BOINC? Should I calibrate my computer by running a CPU benchmark, and if so, where do I enter the calibration results?

How common is credit cheating in BOINC? Should I calibrate my computer by running a CPU benchmark, and if so, where do I enter the calibration results?

Credit cheating isn't as common as some think. Boinc uses it's own benchmark and your benchmarks are:

Measured floating point speed 632.18 million ops/sec
Measured integer speed 1989.36 million ops/sec

compare this to others using powerbooks/powermacs



You'll see your benchmarks are pretty much on target. You should be claiming roughly 5.4615417 per hour of run time.

tony

Thanks!

I have another problem - I have not been able to update my team information. I want to change our type from Unknown to Junior College, but when I click update nothing changes. Do I have to have more than one member of the team in order to be able to change the info?

Any suggestions appreciated.

Charlie

---

Thanks!

I have another problem - I have not been able to update my team information. I want to change our type from Unknown to Junior College, but when I click update nothing changes. Do I have to have more than one member of the team in order to be able to change the info?

Any suggestions appreciated.

Charlie

Already fixed. LOL Some team data changes can take up to a day for you to see them (such as new members active count and such). Full data table sweeps are "expensive" when it comes to DB access so they write them to XML and set the scan frequency to higher periods to free up their resources. Team data is one of them.

tony

How common is credit cheating in BOINC? Should I calibrate my computer by running a CPU benchmark, and if so, where do I enter the calibration results?

Careful little grasshopper. ( Man I do miss the Kung Fu Series with David Caradine)

Why did my screen saver freeze?

(Hedera)
Hedera Welcom to Rosetta. I'm not sure on that one. There have been some problems with the screen saver lately. Did the work unit complete successfully or did it report errors?

You might bring it up in the thread for issues with the Rosetta version you were running (probably 5.25)

Screen saver issues are less common now. But the running of the Screen server can and does reduces the number of cycles your computer crunches the project ( Drat that is the over clocker speaking) and yes some incompatibilities with windows can cause the screen server to freeze. It used to be that way some months ago.

I know it is not as pretty as with the screen saver but, it is safer to run it without.

How common is credit cheating in BOINC? Should I calibrate my computer by running a CPU benchmark, and if so, where do I enter the calibration results?

Careful little grasshopper. ( Man I do miss the Kung Fu Series with David Caradine)

The BOINC Manager automatically does a benchmark when it's first started, and again an a preset interval - something like 7 or 10 days. That information is uploaded to projects automatically during the routine communications involving Work Units.

---

Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)

How common is credit cheating in BOINC? Should I calibrate my computer by running a CPU benchmark, and if so, where do I enter the calibration results?

Careful little grasshopper. ( Man I do miss the Kung Fu Series with David Caradine)

The BOINC Manager automatically does a benchmark when it's first started, and again an a preset interval - something like 7 or 10 days. That information is uploaded to projects automatically during the routine communications involving Work Units.

Benchmarks happen when upgrading/downgrading boinc version, then every 5 days after that.

tony

Newbie here....

Am curious about how new work is assigned.
I see in my messages tab that it appears the client is requesting work, but what the server parcels out is very different from the request.

Do the work assignements consider the hardware of the client in terms of processor speed, number of CPUs, cache size and memory size? I'm not concerned about credits as I have a very simple machine, but it would seem that including these factors would help improve overall project throughput.

Thanks

Al

Newbie here....

Am curious about how new work is assigned.
I see in my messages tab that it appears the client is requesting work, but what the server parcels out is very different from the request.

Do the work assignements consider the hardware of the client in terms of processor speed, number of CPUs, cache size and memory size? I'm not concerned about credits as I have a very simple machine, but it would seem that including these factors would help improve overall project throughput.

Thanks

Al

Hi Al, Welcome to Rosetta.

There are two major parts to running any project under Boinc. One is Boinc and the other is the Project(in this case Rosetta). Boinc consists of three parts, the manager (Boincmgr.exe, the interface you see and work with), the Daemon (Boinc.exe, does all the work in the background), and the Screensaver (Boinc.scr, it handles graphics/screensaver). Boinc provides ALL maintenance functions for the projects. It decides which to run, when to run it, what to download, how much to download, how much to cache, etc. Rosetta provides the Application file, which does the actual crunching of numbers. To put it in perspective when a project runs under boinc, Boinc will generally consume less than 3% of the processor (even less on modern fast puters <1%) to perform it's function and the rest of your processor will be used by the application (rosetta).

I've found the Boinc Wiki to be an excellent source of information. If you look at how the Work Scheduler works it will answer many of your questions. For boinc to determine the correct amount of work, without allowing a user to miss a deadline it uses the RDCF (Duration Correction Factor) when requesting work. It will take a dozen to a couple dozen returned wus for the DCF to be corrected to the actual completion time of a WU.

OK, It's early and I'm starting to lose track, have I actually answered the question?

what the server parcels out is very different from the request.

I take it you received significantly more work then you requested? If you would give us an idea what you observed that would be helpful. It's very likely that the estimated runtimes of your tasks hasn't settled in to reflect your preference for the runtime preference.

It looks like you are crunching with the default 3 hour runtime, and have been for some time, so perhaps that's not what you meant. Currently you have two tasks on deck. That's about 6 hours of work. Is that not what you expected?

...and yes, the BOINC environment takes those types of factors in to account. Perhaps you've got one where the first model is taking a long time to crunch. This would mean your progress % is still very low, even after an hour of crunching. This is described here in the first two Q&As, not to worry.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Thanks folks,
My question(3) was kind of vague and your answers cleared up most of them. I do need to learn to pay attention to Wiki more, it's come a long way. The only dangling questions are that my client sends a request like:

9/20/2006 11:12:20 AM|rosetta@home|Requesting 584 seconds of new work, and reporting 1 completed tasks
9/20/2006 11:12:25 AM|rosetta@home|Scheduler request succeeded
9/20/2006 11:12:27 AM|rosetta@home|Started download of file 1ogw***.gz

and the work task 1ogw***.gz generates a task
9/20/2006 12:57:00 PM|rosetta@home|Starting task 1ogw__LARS_ABRELAX_ALL_NATIVE_BARCODE__1241_2769_0 using rosetta version 525

which shows as a 3:30:00 size task.

The other question is that when this task starts running the "To Completion" column runs backwards (i.e increasing) for a while, then drops a fair amount (20min or so) then runs backwards a bit for a while, then drops a fair amount again.... It seems that the progress column jumps about 8% whenever this happens.

None of this is significant, and not something I really need to do, it's just a matter of curiosity.
I'm an "old school" computer geek (started on the PDP8), and I'm enjoying the horsepower my new $400 desktop has. It's quite amazing, and bodes well for distributed processing. There's a lot of CPU horsepower just sitting there. I can get work done with the rosetta application set to high priority and I barely notice the loss of response. Very impressive.

Thanks again,

The time increasing/decreasing is an effect of the percentage complete not being very accurate. So when it shows that it's done, say 1.054%, it's actually at, say, 5% complete of the actual work - but it's hard to say if it's 1% or 5% until it's done the model complete, at which point we can "jump" to 5%. Then it stalls again, only updating from 5.005% to 5.034% and such for several minutes (or hours). It's just impossible to say how much of the overall work is done during each step, because it depends on "what you find" during the step.

Just live with it... ;-)

--
Mats

---

Just live with it... ;-)

--
Mats

Just curiosity anyway ;)
I think I need to read the Wiki entry a lot more. There is a lot more going on then meets the eye here.

---

Just live with it... ;-)

--
Mats

Just curiosity anyway ;)
I think I need to read the Wiki entry a lot more. There is a lot more going on then meets the eye here.

don't try it in one sitting......Last I heard (6 months ago or so), it was more than 1700 pages. LOL

HP_Owner: Your client told the server, that it needs at least 584 seconds of new work to fill its queue. So the server did send you one WU (and not more). The size of the next WU to be sent is not taken into account when sending one WU. It is only used to compute if additional WUs are needed.

Norbert

---

HP_Owner: Your client told the server, that it needs at least 584 seconds of new work to fill its queue. So the server did send you one WU (and not more). The size of the next WU to be sent is not taken into account when sending one WU. It is only used to compute if additional WUs are needed.

Norbert

Makes sense now. The minimum quanta (by default) is 3 hrs.
Another thing I've noticed, that I don't understand is that with certain work units (the one I'm running right now), it's been running for 25 minutes and shows no sign of progress.
Is there a point at which it make sense to abort, or do I just let it run...
The CPU is fully engaged working on this WU...

---

HP, 25 minutes is not much time to assess if there is any problems. When you crunch a work unit, your science work is measured in completed models crunched. And with the WU runtime preference, the client can basically end a WU at any end of model. So, for example, you have a 3hr time preference. If your first model takes 2 hours to run, your progress % will show 1.xxxx for the whole 2 hours, and the estimated completion time will be increasing.

At the end of that model, it will reestimate completion time and % done, determine that a second model on such a tough protein would take 4 hrs, and it will decide now would be a good time to end this WU to avoid exceeding the time preference. So, your progress jumps from 1.xxx% to 100%.

Now we start on the next protein and it is smaller. We can crunch 6 models an hour on this one. The exact same thing happens, during that first model your progress shows 1.xxx% and your time to completion is increasing... but after 10min, the first model is completed. The numbers are recalculated and you see a more linear, granular progression.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/