| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Microbiome Immunity Project
Thread: WU Characteristics |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 100
|
|
| Author |
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3295 Status: Offline Project Badges: 
|
I think it runs warmer! Usually the cores are at 59-62º with VINA but most were at 66º while running MIP.
----------------------------------------I saw the battery drain from 54% to 37% in about 2 hours while running MIP only. Now, it's running FAAH/SCC since last night and the battery is at 43%. AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W AMD Ryzen 7 7730U 8C/16T 3.0 GHz [Edit 2 times, last edit by Falconet at Aug 27, 2017 2:58:34 PM] |
||
|
|
phytell
Cruncher Joined: Sep 8, 2014 Post Count: 37 Status: Offline |
I've only got info on windows 7/10, but:
runtimes: short (around SCC, maybe a little shorter). Betas took a bit longer, so maybe we'll see longer running WUs later. ram: higher (I saw 200mb used). Disk: 28.5m or so in the slots directory per unit. As far as I can tell, its generally constant, not like CEP2 where you'd start out at one level and have it skyrocket as the unit progressed. I don't have any info on amount of disk activity, unfortunately. Transfer: about 75mb initially. I haven't been able to measure incoming WUs at all, but outgoing is pretty small (boinc reports 35kb). The disk and ram requirements are a little higher than FAAH or SCC, but they're still not bad. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges: 
|
Do we miss some MIP1 results in the statistics?
----------------------------------------The current situation of my Linux host (i7, Ubuntu Mate) crunching for MIP1 is really not OK: its average daily contribution (credits) drops down from 11'800 to I cannot assume that MIP1 is only granting 50% of the contribution work (if it would be the case, a correction should be done in the science application). My concern is about maybe some statistics mistakes. I would really appreciate that other owners of Linux based hosts could provide feedback. Cheers, Yves --- Edited for correction; see https://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=548912 ---------------------------------------- [Edit 1 times, last edit by KerSamson at Sep 2, 2017 4:55:39 AM] |
||
|
|
Halo Jones
Cruncher Joined: Mar 29, 2015 Post Count: 31 Status: Offline Project Badges: 
|
Do we miss some MIP1 results in the statistics? ...average daily contribution (credits) drops down from 11'800 to 6'000 points , without running dry of WU... ...I would really appreciate that other owners of Linux based hosts could provide feedback. Cheers, Yves Not so dramatic drop here. My Linux Mint laptop with A10 CPU crunching MIP seems to make around 12.000 points per day shown on WCG stats page. This is only a bit less than crunching SSC. Boinc manager's graph shows slow drop from 1.700 points (SSC) to 1.600. points (MIP) and may be still dropping slowly. [Edit 1 times, last edit by Halo Jones at Aug 30, 2017 11:12:22 AM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Do we miss some MIP1 results in the statistics? The current situation of my Linux host (i7, Ubuntu Mate) crunching for MIP1 is really not OK: its average daily contribution (credits) drops down from 11'800 to 6'000 points , without running dry of WU. I cannot assume that MIP1 is only granting 50% of the contribution work (if it would be the case, a correction should be done in the science application). My concern is about maybe some statistics mistakes. I would really appreciate that other owners of Linux based hosts could provide feedback. Cheers, Yves KerSamson, why do you keep on beating the dead horsie? WCG is zero interested in allowing/figuring consistency in GFLOP measurement to assess performance. Contrary to my good judgement, you made me look, and I shouldn't have, but to take the Coq-o-Vin blanc, the credit cook keeps on using bad vinegar instead, of the poor 2015 harvest, Cock-o-vinaigrette, inedible. My hunting tool has an advanced pivot slicer/plotter, and when filtering the most curious observation is that on my W10 there's is no decline, the linear regression trend line is practically flat, but on Linux there is, gradually going from about 20 per hour, to 14-15 per hour. Oh, the W10 machine runs a app_config controlled mix of SCC1(3)/ZIKA(3)/MIP1(2) concurrent, where latter's credit/hour is well higher than of the ZIKA/SCC1, but that bend may have some other warp because VINA on Windows runs half as fast then on Linux i.e. the Credit_New mess persists, unwavering! My last word on the matter (something I always say on this maligned topic). [Edit 1 times, last edit by SekeRob* at Aug 31, 2017 12:29:56 AM] |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges:
|
I give up !
----------------------------------------Back to OET1 ! Yves |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges:
|
Hi,
----------------------------------------I'm back from travel and I have to add some corrections to my previous statements, Finally, within one week, the mentioned Linux host dropped its daily average from 11'800 to 9'300 before I switched it back to OET1. It is about -21%. However, after 7 days, the trend was still not stable and it was still dropping down. My mistake regarding my statement about -50% has been caused by (still on-going) update troubles with boincstats. The reported figures are based on boinctask. Cheers, Yves ---------------------------------------- [Edit 1 times, last edit by KerSamson at Sep 2, 2017 4:57:41 AM] |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges:
|
I don't know if somebody will take care, but I made a strange observation regarding MIP and Linux, see below:
---------------------------------------- From Friday 2017-08-25 (first Friday on the figure), the host (i7-6770K, Ubuntu Mate 16.04 x64) computed MIP only. At the same time, the core average temperature started to drop down from about 62°C to 45°C within a week. On Friday 2017-09-01 (second Friday on the figure), the host started again to compute OET1 only and the core temperature came back to 62°C. The CPU load remained at 100% during this time.  It looks like that, on Linux, as more MIP WUs will be crunched, as slower the machine is running even showing 100% CPU load. This continuous temperature decreasing fits perfectly with the decreasing of granted points. I come back to my usual argument: I'm still considering that the granted points reflect the system performance and that they can be used for monitoring purpose. With MIP on Linux, the problem is not a CPU benchmark problem but a real performance and behaviour problem. I would appreciate if somebody by the scientists or by the WCG tech team could investigate this issue. Since I do monitor a couple of my hosts with LibreNMS, I have a sounded overview about the host behaviours. Cheers, Yves |
||
|
|
smeyer55
Senior Cruncher Joined: Feb 15, 2009 Post Count: 303 Status: Offline Project Badges: 
|
I noticed in the last few days the work units have slowly climbed from ~1.5 hours to 4-7 hours on all of my machines.
Either these are some bigger jobs or they are playing around with the work unit sizes. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges:
|
Hi smeyer55,
----------------------------------------in the case I reported, I discovered that the CPU (even with a 100% load) was performing less, otherwise the CPU temperature would not have dropped down from 62°C to 45°C under the same environmental conditions (constant room temperature). Obviously, if the CPU is performing less good, the WU duration increases. Honestly, I did not expect such a behaviour and I was not looking for this. It was a fortuitous discovery. I would be very interested to know if other Linux machines do show a similar behaviour with MIP. Cheers, Yves |
||
|
|
|