confidence check, please?

Everything about the project RNA World
Nachricht
Autor
julianop
Idle-Sammler
Idle-Sammler
Beiträge: 4
Registriert: 19.10.2015 16:01

confidence check, please?

#1 Ungelesener Beitrag von julianop » 26.03.2016 01:08

I have a monster unit, currently at 130d 07:52:xx, with 42d 17:36:xx remaining.

Unit name is cmsvm2_GA-p[e20-30MB_Lin64f]_1_Oryzias-latipes-(Japanese-medaka)_DG000006.lin.EMBL_RF00028_Intron_gpI_1349111823_70000_32
and it's running on cmSearch VM (VirtualBox) 1.0.2.15 (vbox64) on Win7/64 Ultimate on Intel i7 920@2.67GHz with 9.00GB RAM.
(I wish I hadn't needed to type that in by hand!)

In the past 7 days the time remaining has only gone down 2 days, and as I watch the remaining (estimated) only clicks down 1 second every 3 or 4 seconds of real time.

Should I consider that things are going "normally" ?

ChristianB
Vereinsvorstand
Vereinsvorstand
Beiträge: 1915
Registriert: 23.02.2010 22:12

Re: confidence check, please?

#2 Ungelesener Beitrag von ChristianB » 26.03.2016 11:27

Don't watch time remaining. Watch the progress done. If this is still progressing (or stuck at 98.765%) everything is normally. Since we can't predict the runtime beforehand the remaining time is useless for the VM tasks.

Jacob Klein
Oberfalter
Oberfalter
Beiträge: 481
Registriert: 26.07.2013 15:41

Re: confidence check, please?

#3 Ungelesener Beitrag von Jacob Klein » 26.03.2016 13:38

Yeah, it is normal. Exactly as Christian said.

Additionally, you might read through the thread I posted about how, for RNA World tasks, it's nearly impossible to get a good estimate without looking at a wingman's completed result. See here: viewtopic.php?f=75&t=16160

Some of my estimates of my tasks, indicate that they could take 2+ years to complete! If you have an Intel CPU with hyperthreading (which it looks like you do), and you want the task to complete faster (possibly up-to-twice-as-fast), then I recommend setting "Use at most x% CPUs" to 50% instead of 100%, when you have an RNA World task. That will distribute work across cores, instead of using hyperthreading to place 2 tasks/threads on the same core. Though it also means that less work gets done overall, so it's your choice.

PS: If you go to your RNA World web account, and click Tasks, you can see all the info for you tasks, so you can easily copy/paste it, including task names if you click "show names". And if you click your computer number, it'll give you info on your computer so you can easily copy/paste it. Of course, it would have been helpful to us to see your task or work unit - perhaps you can paste a URL link to the task or work unit? :)

Edit: Hmm.. I found that info. Will calculate some estimates for you shortly...

Regards,
Jacob

Jacob Klein
Oberfalter
Oberfalter
Beiträge: 481
Registriert: 26.07.2013 15:41

Re: confidence check, please?

#4 Ungelesener Beitrag von Jacob Klein » 26.03.2016 14:52

I spent about an hour, calculating your estimates, based on how long it took the Wingman and comparisons versus your CPU.

They're grouped by "how the Wingman was potentially running", and then within each group, there are 3 estimates based on YOUR setting --
1) you running (all threads) using "use at most 100% CPUs"
2) you running (half threads) using "use at most 50% CPUs"
3) you running just a single task.

Notice how your setting influences the performance within each group!

Anyways, it seems likely that this task will probably take you around 200 days CPU time, but potentially up to 370 days if you are running (all threads) and wingman was running just a single task. I still recommend "Use at most 50% CPUs" on your Intel Hyperthreaded CPU while an RNA World task is running.

There's your "Vote of Confidence!" :) Keep that task crunching!!

Code: Alles auswählen

julianop's CPU:
Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
Cores / Threads: 4 / 8
Multiplier and Turbo: 20, 1/1/1/2
Passmark score (all threads, total): 4,992
Passmark score (all threads, per thread): 624 (4,992 / 8)
Passmark score (single thread): 1,161
Passmark score (half threads, per thread): 1,108 (1,161 * 21 / 22)
Measured FLOPS: 2822.17 million ops/sec

Wingman's CPU:
Intel(R) Core(TM) i5 CPU 661 @ 3.33GHz
Cores / Threads: 2 / 4
Multiplier and Turbo: 25, 1/2
Passmark score (all threads, total): 3,198
Passmark score (all threads, per thread): 800 (3,198 / 4)
Passmark score (single thread): 1,326
Passmark score (half threads, per thread): 1,277 (1,326 * 26 / 27)
Measured FLOPS: 3242.75 million ops/sec

               Name: cmsvm2_GA-p[e20-30MB_Lin64f]_1_Oryzias-latipes-(Japanese-medaka)_DG000006.lin.EMBL_RF00028_Intron_gpI_1349111823_70000
           Web link: http://www.rnaworld.de/rnaworld/workunit.php?wuid=6330948
       Web workunit: 6330948
     Wingman status: Completed: 4 Jul 2014, 8:41:18 -- Intel(R) Core(TM) i5 CPU 661 @ 3.33GHz -- CPU Time: 15,025,240 -- 173d 21:40:40

================================================================================================
Estimates for julianop to complete:
================================================================================================

Wingman running (single thread):
  369d 13:03:55 -- 31928635 s -- (1,326 / 624) * time
  208d 2:51:10  -- 17981470 s -- (1,326 / 1,108) * time
  198d 14:50:10 -- 17160610 s -- (1,326 / 1,161) * time

Wingman running (half threads, per thread):
  355d 21:19:28 -- 30748768 s -- (1,277 / 624) * time
  200d 10:16:36 -- 17316996 s -- (1,277 / 1,108) * time
  191d 6:41:10  -- 16526470 s -- (1,277 / 1,161) * time

Wingman running (all threads):
  222d 22:52:08 -- 19263128 s -- (800 / 624) * time
  125d 13:29:09 -- 10848549 s -- (800 / 1,108) * time
  119d 19:55:09 -- 10353309 s -- (800 / 1,161) * time

FLOPS:
  199d 19:40:09 -- 17264409 s -- (3242.75 / 2822.17) * time


julianop
Idle-Sammler
Idle-Sammler
Beiträge: 4
Registriert: 19.10.2015 16:01

Re: confidence check, please?

#5 Ungelesener Beitrag von julianop » 29.03.2016 14:47

Thanks very much for your work, Jacob :-)

I'll "carry on crunching" !

Jacob Klein
Oberfalter
Oberfalter
Beiträge: 481
Registriert: 26.07.2013 15:41

Re: confidence check, please?

#6 Ungelesener Beitrag von Jacob Klein » 29.03.2016 14:57

:) Just to confirm ... Task Manager still shows that the task is regularly using CPU (probably 12.5% on your system, which is 1 thread), right? Seeing CPU usage in Task Manager, is the primary way to know that an RNA World VM task is still crunching normally, especially when it progresses so far that its progress only shows 98.765%, even though it is still doing work and should still finish just fine.

Silverdrake
Mikrocruncher
Mikrocruncher
Beiträge: 17
Registriert: 21.04.2013 22:06

Re: confidence check, please?

#7 Ungelesener Beitrag von Silverdrake » 06.04.2016 03:01

I also have a long-running task that is decrementing Remaining by 1 second per 5 seconds elapsed, with 1007 hours estimated remaining. It is running on this computer. Do AMD CPUs have the hyperthreading and would decreasing it to 50% CPUs help speed this monster up?

I'm still running BOINC 7.6.9 because when I upgraded to 7.6.22, the WU showed a VM error. I recovered it by rolling the install back to 7.6.9. When the VBox wrapper update was announced, I asked in that thread if it would affect this WU and never received a reply.

A potential problem shows when checking the Workunit listing: "errors Too many errors (may have bug)"

So in addition to not knowing if resettng the CPU % used would help, I don't even know if this WU is still valid.

ChristianB
Vereinsvorstand
Vereinsvorstand
Beiträge: 1915
Registriert: 23.02.2010 22:12

Re: confidence check, please?

#8 Ungelesener Beitrag von ChristianB » 06.04.2016 07:38

If you don't have problems I would advise against an upgrade of the BOINC Client. There is no data that would support a speedup when Hyperthreading is disabled in case of the RNA World cmsearch application. Furthermore this is a BIOS setting it has nothing to do with the CPU % used you can set in the Manager/Preferences.

Please ignore the "Too many errors (may have bug)" message. Your task will get accepted either way and I just have to reset it later.

Antworten

Zurück zu „RNA World Discussions (english)“