Posts by Greg_BE

1) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91241)
Posted 10 Oct 2019 by Profile Greg_BE
Post:
Thanks guys for the input.
With BOINC tasks I see what is going on now.
So I will just let the 4.07 stuff run its course.
2) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91225)
Posted 6 Oct 2019 by Profile Greg_BE
Post:

Well never thought about memory speed being an issue. Always figured it was the processor.
Memory speed is 1066mhz (but the CPU-Z says something about 2133 and when I look at MSI Command center DRAM it talks about 2133) and 24,526mb capacity. With all the stuff I have running including monitor programs I am using 91-94% of the memory. I don't know how to OC memory, I have always left that to the programs to control automatically when I crank up the CPU speed.


Faster memory would make the tasks go faster, but it wouldn't directly impact the % cpu utilization. The time spent waiting for slow DIMMs counts as cpu time.

"I am using 91-94% of the memory"

That is way too much. If I were you, I would try to cut that % in half, either by adding more RAM or by cutting down on memory-hungry tasks.



Not sure what was hogging the memory so much. I had BSOD for a bad driver and since fixing that I am down to 43%. I think ATLAS is the memory hog and for now Amicable numbers is the high scorer, but that is a GPU project. I'll see what ATLAS is up to later on tonight or tomorrow. I also had a bunch of Firefox tabs open last night. Give you an update later on tonight or tomorrow once the big stuff hits again. When I was writing last update it was the middle of the night here in Europe.

Since Amicable is multithreaded, it will rocket that ram consumption to around 10-13 GB or so. Atlas requires tons of babysitting since it is an absolute memory hog with the tasks requiring Virtual box (I think). My questions on that forum remain unanswered, so I haven't really delved too deeply into it.

...oh. I just read re: GPU. One tip for Amicable. It actually uses a ton of CPU to run the GPU - if you run primarily CPU you will most likely get higher throughput at this point.


Amicable is being nice and using only 6% cpu on 11% GPU. I think FAH is the big CPU hog. 31% Amicable is a memory hog though. nearly a gig. FAH is less memory. Only 126. ATLAS likes 6hrs uninterupted and yeah its a memory hog in its own right. Amicable is running 1:1 on cpu time to elapsed time. I got a ROSIE 4.07 that isn't doing so well. 1 day 3hrs and only 5.5 hrs cpu time. But now that I look at it in BOINC tasks, I see its not due until the 11th. So no wonder it is taking so long. Other stuff is before it in due date time.
3) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91222)
Posted 6 Oct 2019 by Profile Greg_BE
Post:

Well never thought about memory speed being an issue. Always figured it was the processor.
Memory speed is 1066mhz (but the CPU-Z says something about 2133 and when I look at MSI Command center DRAM it talks about 2133) and 24,526mb capacity. With all the stuff I have running including monitor programs I am using 91-94% of the memory. I don't know how to OC memory, I have always left that to the programs to control automatically when I crank up the CPU speed.


Faster memory would make the tasks go faster, but it wouldn't directly impact the % cpu utilization. The time spent waiting for slow DIMMs counts as cpu time.

"I am using 91-94% of the memory"

That is way too much. If I were you, I would try to cut that % in half, either by adding more RAM or by cutting down on memory-hungry tasks.



Not sure what was hogging the memory so much. I had BSOD for a bad driver and since fixing that I am down to 43%. I think ATLAS is the memory hog and for now Amicable numbers is the high scorer, but that is a GPU project. I'll see what ATLAS is up to later on tonight or tomorrow. I also had a bunch of Firefox tabs open last night. Give you an update later on tonight or tomorrow once the big stuff hits again. When I was writing last update it was the middle of the night here in Europe.
4) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91219)
Posted 6 Oct 2019 by Profile Greg_BE
Post:
Hello,
A few things to try, as it were.

Try completely suspending - or at least asking for no new work from other projects apart from Rosetta and see if the CPU issue resolves itself.
Some projects from LHC ask for far more than 1 gb of memory per core and are also multithreaded. Add Virtual box to that equation and combine that with Rosetta, which in some instances can also hit 1 gb per core if not more. Windows might be hitting the virtual memory up. One other thing to consider is Ram speed, which can have an impact on Ryzen processors, though from what I can tell it shouldn't be bringing CPU usage per task down to that level of so called inefficiency. My 1800x is still running ddr4 2400 though I'll be replacing that with a little higher speed - maybe 2666 or 3000.

If not that, one more thing to try is using an app config to limit work from Rosetta to see if, at the very least, CPU usage per core picks back up. That will tell you if things are going into swap or not or at least something isn't able to keep up.

Thanks and hope that helps a little.


Well never thought about memory speed being an issue. Always figured it was the processor.
Memory speed is 1066mhz (but the CPU-Z says something about 2133 and when I look at MSI Command center DRAM it talks about 2133) and 24,526mb capacity. With all the stuff I have running including monitor programs I am using 91-94% of the memory. I don't know how to OC memory, I have always left that to the programs to control automatically when I crank up the CPU speed.
5) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91215)
Posted 4 Oct 2019 by Profile Greg_BE
Post:

So how do I get Rosie to act like those tasks?


Rosetta does not control the actual dispatch of CPU time. The operating system does this. But we know that the Rosetta application will demand as much CPU as it can get (based on the BOINC settings, and the resource availability in your machine). My guess would be that memory used by the other applications is less, and that R@h is often encountering the situation where the memory pages it requires to run are swapped out, and it has to wait for the operating system to bring them back in to physical memory before it can actually continue running.


So in short, I can run it the way it is or dump a project or two that use a lot of cpu time?

Note: I have taken one core away from LHC (they have the option to let you say how many cores you want to use), will see if that does anything. If not that, then I drop one of their applications that I see sucks up a ton of CPU time.
6) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91211)
Posted 4 Oct 2019 by Profile Greg_BE
Post:
so...a 4.07 task has been running 11hrs and 24 minutes and has only 2 hrs 22 minutes cpu time.
BUT...it is only using 20.9% CPU usage, but what is weird is the cpu usage climbs each time BOINC tasks updates. So by the time you read this, it is using 30-31%.
Now a different 4.07 tasks is 9hrs elapsed time with 3:31 cpu time and 36% cpu usage.

The only consistant cpu to elapsed time task is a non Rosie task from both LHC and Einstein. Both are within 1hr of their cpu time.

So how do I get Rosie to act like those tasks?
Why are they using 94,98 and 107% and Rosie is only 20-40%?
All this is based on what BOINC tasks is reporting.
7) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91210)
Posted 4 Oct 2019 by Profile Greg_BE
Post:
https://1drv.ms/u/s!AsKhVK_EPRXJgpNHq6GczXavstTj_w?e=YSHQGs

Goto this location to see the image. I can't get it to load.
8) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91208)
Posted 4 Oct 2019 by Profile Greg_BE
Post:
This is an extract from BOINC Tasks, so tell me what I am looking at in terms of Rosie vs the rest. Rosie is set for 120%, but I don't know if that gives it extra processing priority or what?
The system is a 8 core 16 thread AMD Ryzen 7 2700 chip. I am OC with automatic control to 39.25 GHZ with 24 GB of memory. I rarely go past 80% usage if even that when running ATLAS simultaneously with Rosie or other programs.

Image problems..will correct in a few hours.
9) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91207)
Posted 4 Oct 2019 by Profile Greg_BE
Post:
hey Neal,

Very cool! Much easier way to see whats going on.
Thanks for the recommendation.
10) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91206)
Posted 3 Oct 2019 by Profile Greg_BE
Post:
You have now taken my original point. The question of "why is it taking 18 hours to get 6 hours of CPU time?"

The resource shares is really just deciding when to run each project and how much work to request. It is TRYING to run R@h for those 18hrs (or whatever exact number), but during that period of time only actually was able to get 6 hours of CPU time. It basically means that the BOINC Manager is showing the task in a "running" state for 18 hours, but if you were watching the Window tasks manager during that period of time, it would show the task only getting 33% of one core of one CPU.

Do you see BOINC Manager showing tasks "waiting for memory"?

You mentioned vbox, is BOINC running in a vbox? If so, how many cores and how much memory is available in the vbox?

You also mentioned "Rosie is using a combined total of about 6%", I was unclear exactly what you meant by this. When the R@h task is running, Windows task manager should show it using close to 100% of one CPU core (at least once passed the first minute or so of initial start-up time for a task, during which time it will not be using 100% of CPU yet).


Never known Rosie to use VBOX.
Boinc never says waiting for memory, I just put in another stick to deal with ATLAS tasks from LHC
The 6% was the total CPU time I was seeing in windows task manager.
I checked the graphics of one of the 4.07 units. 18 hrs run time, 2+ hrs CPU time.
Lots of pauses in the searching. It's like it is stalling.
I'll read more about what you said my tomorrow. its past midnight here now.
11) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91205)
Posted 3 Oct 2019 by Profile Greg_BE
Post:
Greg_BE:

Try un-checking "Suspend when computer is in use" in "Computing preferences..." and see if that has any effect. I got better results after disabling that.



That has never been checked.
12) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91197)
Posted 2 Oct 2019 by Profile Greg_BE
Post:
So check this out: https://boinc.bakerlab.org/rosetta/result.php?resultid=1096093226
Run time 18 hours 53 min 37 sec
CPU time 5 hours 59 min 25 sec

But project is set for 100% as are all the others.

So....do I increase resource share to get Rosie back up to about the same level as the other projects that are 100% or what?

See all the details about how BOINC is set up below.

Resource share now 120%
13) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91195)
Posted 1 Oct 2019 by Profile Greg_BE
Post:
leave non gpu tasks in memory is selected...any other ideas?

Just fyi, the clock is running normal lately, but the run time are super long as you see.

I'm 18 hrs into the most current task and it shows 17 hrs to completion.
Very odd. Unless I should bump the resources % up a few notches?
14) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91193)
Posted 1 Oct 2019 by Profile Greg_BE
Post:
Note: Vbox is taking 41% when its running
15) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91190)
Posted 1 Oct 2019 by Profile Greg_BE
Post:
Hey Mod!

No..the only other thing running is FAH to keep a graphics card busy.
Looking at Task Manager I see its using 44% CPU time.
Rosie is using a combined total of about 6%
Universe is taking about 10%

those are my big users.

That's the only outside the BOINC program running, otherwise everything is default.
All BOINC projects are 100% equal.
There are no modifying scripts for Rosie.
16) Message boards : Number crunching : Rosetta 4.0+ (Message 91178)
Posted 30 Sep 2019 by Profile Greg_BE
Post:
Well no wonder 4.07 takes so long. There is lots of wasted time.
I had a look at one task running on my system, the searching box ran blank for 4 cycles before I closed it.
That's a lot of wasted time.
It's no wonder it takes 15 hrs to run on my system!

Is this a bug or is it just not finding anything to chart?
Seems weird because I thought ALL TASKS ALWAYS had something in the searching box.
17) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91177)
Posted 29 Sep 2019 by Profile Greg_BE
Post:
I can't run tasks using 4.07.
They get stuck and the status level does not move and the time remaining clock keeps counting up instead of down. So for now, anything 4.07 is being aborted.

Why do they contain the name "intel" in the the application?
What is the difference between these and the 3.78 that uses "windows"?

Another question, why doesn't my account view page allow us to select which version we want to run? So that I could opt out of 4.07 and get other tasks from other applications?



OK...the Intel is answered.
But does anyone have an idea why some of the 4.07 remaining time count up and not down after reaching 50%? Also what is the typical run time of a 4.07 task? I have BOINC set on 6hr run times before switching tasks and I see most 4.07 tasks running 12-16 hrs. Well that's elapsed time. So the task has run 2 of my cycles so far and not completed, which I find odd. Any ideas?
18) Message boards : Number crunching : Tasks using 4.07 - time remaning clock counting up (Message 91167)
Posted 27 Sep 2019 by Profile Greg_BE
Post:
I can't run tasks using 4.07.
They get stuck and the status level does not move and the time remaining clock keeps counting up instead of down. So for now, anything 4.07 is being aborted.

Why do they contain the name "intel" in the the application?
What is the difference between these and the 3.78 that uses "windows"?

Another question, why doesn't my account view page allow us to select which version we want to run? So that I could opt out of 4.07 and get other tasks from other applications?
19) Message boards : Number crunching : Rosetta 4.0+ (Message 91166)
Posted 27 Sep 2019 by Profile Greg_BE
Post:
I am aborting any and all 4.07 tasks. They don't work on my pc.
Instead of counting down, they reach around 40% and count up.
Useless for me to waste my cpu time on these.

Can someone explain the name of the application version name: Rosetta v4.07
windows_intelx86?

Is this an version oriented towards INTEL chips or what?
I have a 50% or there about rate of success with these applications.

BTW...someone should tell the webmaster who writes the code for the user section, to give us the option to select which versions we want to run. How can I opt out of 4.07 Intel?
20) Message boards : Number crunching : Rosetta 4.0+ (Message 91033)
Posted 16 Aug 2019 by Profile Greg_BE
Post:
Incorrect function.
(0x1) - exit code 1 (0x1)</message>
<stderr_txt>
command: projects/boinc.bakerlab.org_rosetta/rosetta_4.07_windows_intelx86.exe -beta -frag3 00001.200.3mers -frag9 00001.200.9mers -abinitio::increase_cycles 10 -mute all -abinitio::fastrelax -relax::default_repeats 5 -abinitio::rsd_wt_helix 0.5 -abinitio::rsd_wt_loop 0.5 -abinitio::use_filters false -ex1 -ex2aro -in:file:boinc_wu_zip hugh_8_17_seq_24_fragments_fold_data.zip -out:file:silent default.out -silent_gz -mute all -in:file:fasta 00001.fasta -out:file:silent_struct_type binary -nstruct 10000 -cpu_run_time 28800 -watchdog -boinc:max_nstruct 600 -checkpoint_interval 120 -database minirosetta_database -in::file::zip minirosetta_database.zip -boinc::watchdog -run::rng mt19937 -constant_seed -jran 2044900

ERROR: ERROR: FragmentIO: could not open file 00001.200.9mers
ERROR:: Exit from: ..\..\..\src\core\fragment\FragmentIO.cc line: 233
BOINC:: Error reading and gzipping output datafile: default.out

No different than the rest of you...really makes me wonder whats happening in la la land.


Next 20



©2020 University of Washington
http://www.bakerlab.org