| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Microbiome Immunity Project
Thread: sigificant credit drop - only for me or did someone else see this? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 117
|
|
| Author |
|
|
Magiceye04
Cruncher Joined: Jul 5, 2008 Post Count: 31 Status: Offline Project Badges: 
|
Hi!
----------------------------------------at the moment i only crunch Microbiome workunits. Until 2017-10-11 i got around 9000credits per day on my Ryzen1700 (15 threads running in parallel @Ubuntu17.04). At 2017-10-12 it was cold in my home, so I increased the clock speed from 3200 to 3500 MHz. But the credits were lower from that day on. Only around 6000cr/d. I thought there might something be wrong with the overclocking (but i got no errors), so i decided to set the CPU back to stock settings yesterday. But now i am around 4500 cr/d. So the question is: did someone else see this credit drop some days ago? Then this would be the normal behaviour ot the project, otherwise i need to check my PC, but its perfectly cooled and powered and running stable at all. Best Regards Magiceye  clickme [Edit 7 times, last edit by magiceye04 at Oct 17, 2017 9:22:08 PM] |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7222 Status: Offline Project Badges: 
|
The only observation I can make is the change in duration of the MIP units. They had started in the 4 to 5 hour range, then dropped to the one hour range and now back to the 3 to 3.5 hour range. I don't know if others have experienced this, but I have seen more points per hour on the shorter units than on the longer units. So, on my systems at least, if the units have gotten longer, my points earned would have decreased. I will do some further checking to see if I can statistically validate this theory.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1664 Status: Offline Project Badges:
|
Hi magiceye64,
----------------------------------------I reported continuous and recurrent performance problems with MIP1 since the project launch; see https://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=548472 and following. My last assumption is that running simultaneously more than 2 threads could have a significant negative impact on the system performance. Please take my assumption carefully, since it is only an assumption. Yves |
||
|
|
sphospital
Cruncher Joined: Apr 30, 2014 Post Count: 2 Status: Offline Project Badges:
|
Yes.
----------------------------------------I have seen a 16 thread AMD Ryzen drop to 40% of original output, when all or most threads are on MIP1 jobs. Removed MIP1 project and results went back up to normal. I can backup Yves and your observations. [edit] All CPUs on this PC would still run at 100% on all threads with MIP1 jobs but temperature would drop dramatically to just a few clicks over idle. The tasks would run for over 5 hours. (longer than a Core Duo) [Edit 1 times, last edit by sphospital at Oct 18, 2017 12:21:51 AM] |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7222 Status: Offline Project Badges:
|
OK, I think my theory was wrong. Here is a spreadsheet showing average time per work unit and the the points (both WCG and Boinc) per hour of time. I am not seeing a correlation between length of the work unit and the average points granted. In fact the points granted per unit of time appear to have gotten a little bit better lately.
---------------------------------------- Cheers
Sgt. Joe
----------------------------------------*Minnesota Crunchers* [Edit 1 times, last edit by Sgt.Joe at Oct 18, 2017 3:10:51 AM] |
||
|
|
rod4x4
Cruncher Joined: Apr 29, 2014 Post Count: 12 Status: Offline Project Badges:
|
Hi Sgt.Joe
I see what you are trying to do with this spreadsheet, but it does not adequately isolate whether the figures returned were from hosts doing a mix of projects, or dedicated to MIP1. If this spreadsheet could isolate those hosts, it would reflect a marked drop in credit for CPU cycle for all hosts or at least a subset of host configurations that are affected. I have withdrawn from MIP1 (temporarily until fixed, this is a great project to support) and observed an immediate jump in CPU temperatures (safe figures, indicating the CPU is actually working) and credit has also jumped dramatically. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7222 Status: Offline Project Badges:
|
I see what you are trying to do with this spreadsheet, but it does not adequately isolate whether the figures returned were from hosts doing a mix of projects, or dedicated to MIP1. If this spreadsheet could isolate those hosts, it would reflect a marked drop in credit for CPU cycle for all hosts or at least a subset of host configurations that are affected. Yes, I agree, but I have no way getting those figures. I had thought I maybe had a simple solution, but that is not the case. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Suppose your CPU has 8 threads, but that CPU has only 4 FPU's and all 8 MIP jobs want to suck at that one particular FPU function. What do you get, and how does one fix this?
|
||
|
|
Magiceye04
Cruncher Joined: Jul 5, 2008 Post Count: 31 Status: Offline Project Badges:
|
I can also see, that the run time of the WUs is increased.
So instead of 250 WUs per day its now less than 100. I don't complain about the credits itself* - i just wanted to know if my PC has a problem, which seems not to be existing. Thank you for confirming. *its funny, that my Odroid C2 with 4 ARM-cores@1,5GHz (5W) gets more credits with universe@home than the Ryzen with Microbiome (110W) - but this are completely different projects. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7222 Status: Offline Project Badges:
|
Suppose your CPU has 8 threads, but that CPU has only 4 FPU's and all 8 MIP jobs want to suck at that one particular FPU function. What do you get, and how does one fix this? What you get is a bottleneck which will slow down things a bit. You used to be able to get a separate FPU in the days of the 386/486, but I don't think they are available anymore. So, I guess you can't fix it. At least I don't know how. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|