Posts by äxl

1) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 99727)
Posted 21 Nov 2020 by äxl
Post:
So many "Error while computing". Should I detach this computer?
https://boinc.bakerlab.org/rosetta/results.php?userid=294942
2) Message boards : Number crunching : Never ending tasks and past tasks (Message 94839)
Posted 19 Apr 2020 by äxl
Post:
A small desktop fan blowing in to the system is your friend.
I needed this many years ago to keep a system's CPU below 80°c when working hard.
:-)

I measured voltage on the fan outlet to check if it's the mainboard. I accidentially short circuited the system and it crashed. I replugged the fan and now it runs at 100%. LOL (But: The fan is blowing outside. I didn't touch it physically.)
I can now safely run BOINC at 90% with both cores active.
I try to keep my temperature at around 60 °C. When I limit cores to one I can indeed reach 100% without going too far above 70 °C.
But I can also run BOINC on both cores and set usage safely to 70%. This is better, isn't it?
Not really.
Making use of "Use at most x% of CPU time" to reduce the load on a CPU actually puts more stress on the CPU as the constant starting & stopping actually puts quite a bit of thermal stress on it- it gets hot, then cool, then hot, then cool, then hot then cool. Expand, contract, expand, contact, expand, contract...

This CPU is almost 13 years old so it wouldn't be too bad if it broke IMO.
Also isn't 2x70 more than 1x100?
Why? Wouldn't give me a 1d WU give me a farther away deadline?
Nope.
The deadline is fixed, that is the period of time in which to return a Task. If you have it set to run for 24 hours, and then make use of "Use at most x% of CPU time" to keep your CPU cool, as you have found that increases the time it takes to finish the Task.
Hence why going with the default Target CPU runtime, making use of just the 1 core & setting "Use at most x% of CPU time" to 100% would be your best option- keep the temperatures down, get plenty of work done, and not run in to deadline problems.

Okay, I will lower preferred runtime in settings. But maybe I don't need to do it too much.

I am running this script BTW: https://gitlab.com/UMLAUTaxl/boinctemp/blob/master/boinctemp.sh
I guess I could rewrite it a bit to activate/deactivate cores between longer intervals instead of changing CPU usage every minute.
3) Message boards : Number crunching : Never ending tasks and past tasks (Message 93645)
Posted 6 Apr 2020 by äxl
Post:
What temperature is the CPU running at? As long it's 70°c or lower, it's not an issue. From memory, even with the stock heatsink & fan, even running Rosetta 24/7 shouldn't put it's temperature over 70°c as long as the heatsink & fan is clean, along with the inlets & outlets of your case and the case fan(s).

The other option would be let the Tasks run uninterrupted, but only use 1 Core of your CPU. More processing would get done, and you'd still keep the CPU cool.
Use at most  50 % of the CPUs
Use at most 100 % of CPU time

Thanks for reminding me to clean my (stock) heatsink and fan. I didn't know it would make such a big difference. :/
Unfortunately the case fan is broken so I always keep the case open. So I will have to clean it more often. xD

I try to keep my temperature at around 60 °C. When I limit cores to one I can indeed reach 100% without going too far above 70 °C.
But I can also run BOINC on both cores and set usage safely to 70%. This is better, isn't it?

Well, the task is running and it's checkpointing fine.
Is it right, what Mod.Sense worked out, that you've changed your preferred runtime to 24hrs rather than the default 8hrs?
Because if you've also set it to suspend running while in use, that's going to extend the runtime to exactly what you're seeing and sometimes you'll struggle to meet deadline.

Why? Wouldn't give me a 1d WU give me a farther away deadline?

The task has run successfully for 17hrs. Set a more appropriate preferred runtime (and the default 8hr suits you) and it should report in a time acceptable to you, before deadline.

You asked at the start if it's the taskproject or you. It's your amended settings.

True. xD

Now I've got a task running that hasn't been checkpointed since start. Is that bad?
CPU time			07:12:10
CPU time since checkpoint 	07:12:10
4) Message boards : Number crunching : Never ending tasks and past tasks (Message 93346)
Posted 4 Apr 2020 by äxl
Post:
Well, it's definitely running, but it's getting interrupted quite a lot (24hr difference between the two)
Do you have "Suspend when computer is in use" checked?
What's the time since the last checkpoint? Has it checkpointed at all?
I'm guessing this must be one of those 16hr (cpu time) tasks otherwise the watchdog would have cut in already
It looks like another dodgy task - it's not looking good

Don't worry about old tasks too much. They do seem to be aging them off quite quickly, I agree


Here's the full ouput:
Application 			Rosetta 4.07 
Name 				3az4ii6b_jhr_design1_COVID-19_SAVE_ALL_OUT_903430_1
State 				Running
Received 			Sun 29 Mar 2020 05:38:35 CEST
Report deadline 		Mon 06 Apr 2020 05:38:34 CEST
Estimated computation size 	80,000 GFLOPs
CPU time 			17:05:56
CPU time since checkpoint 	00:04:11
Elapsed time 			1d 19:53:47
Estimated time remaining 	19:00:07
Fraction done 			69.789%
Virtual memory size 		1.32 GB
Working set size 		1019.19 MB
Directory 			slots/1
Process ID 			13540
Progress rate 			1.440% per hour
Executable 			rosetta_4.07_i686-pc-linux-gnu


And yes, work is interruped quite a lot. But I've set it up that way cause I don't wanna fry my CPU.
5) Message boards : Number crunching : Never ending tasks and past tasks (Message 93272)
Posted 3 Apr 2020 by äxl
Post:
Weird. If you click on the task and select properties, what does it show for CPU time and Elapsed time?
I'm wondering if it's running at all?

Oh, cool. I didn't know this feature.
One of them says:
CPU time 15:29:54
Elapsed time 1d 15:25:10
Estimated time remaining 22:53:40
Fraction done 63.260%
Progress rate 1.440% per hour


The other one is at 68%. So everything's okay I guess.
Thanks!

(I still wonder where the past cancelled tasks went. Didn't task history used to be longer?)
6) Message boards : Number crunching : Never ending tasks and past tasks (Message 93255)
Posted 3 Apr 2020 by äxl
Post:
I had 2 tasks that were running for days and they still showed about 20h left. I cancelled those manually a few days ago when they hit the deadline.

Now I've got 2 tasks that run for 1d15h and they still show 19-23h left. Yesterday they showd 15h left ...
What is wrong? Is it my system?

I also can't find the past tasks that I cancelled. That is no longer ago than last week.

http://boinc.bakerlab.org/rosetta/results.php?userid=294942
7) Message boards : Number crunching : Rosetta and Android (Message 81073)
Posted 21 Jan 2017 by äxl
Post:
Please, read other thread
Rosetta Mini for Android.


TYVM for that. I wanted to rate your post positively but that didn't work out.
8) Message boards : Number crunching : Rosetta and Android (Message 81069)
Posted 20 Jan 2017 by äxl
Post:
I used to start WUs on my phone but since they're more needy than WorldCommunityGrid I only run these.
Now on my desktop PC I stopped Rosetta WUs because it seems to clutter requesting Lattice WUs. These are really rare and so it happens that my server runs out of WUs altogether. When I again allow new tasks from Rosetta the event log says:
Fri 20 Jan 2017 15:39:09 CET | rosetta@home | work fetch resumed by user
Fri 20 Jan 2017 15:39:09 CET | rosetta@home | update requested by user
Fri 20 Jan 2017 15:39:10 CET | rosetta@home | Sending scheduler request: Requested by user.
Fri 20 Jan 2017 15:39:10 CET | rosetta@home | Requesting new tasks for CPU
Fri 20 Jan 2017 15:39:13 CET | rosetta@home | Scheduler request completed: got 0 new tasks
Fri 20 Jan 2017 15:39:13 CET | rosetta@home | No work sent
Fri 20 Jan 2017 15:39:13 CET | rosetta@home | Rosetta Mini for Android is not available for your type of computer.

Why is that? On my AMD64 desktop PC!
My phone is not active here anymore:
Last time contacted server 27 Jul 2016 21:41:27 UTC
9) Message boards : Cafe Rosetta : Happy New Year! (Message 64685)
Posted 1 Jan 2010 by äxl
Post:
to everyone!






©2024 University of Washington
https://www.bakerlab.org