| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: OpenZika
Thread: 90%++ invalid results ??? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 15
|
|
| Author |
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
Receiving 90%++ invalid results (OpenZIKA) on Mac mini / 10.7.5, i5, 16GB RAM, Intel HD Graphics 3000 (no openCL i suppose).
----------------------------------------This machine should be reasonably fast enough (any, which helps is, i suppose). A second computer (Macbook Pro 10.9.4, HD Graphics 4000) is not getting any invalid results. Is the server sending me packages meant for otehr types of computers eventually? Best and thank you in advance, jayrope (Last edit changed 50%++ to 90%++ in title, as this turned out only over time watching the results.) [Edit 7 times, last edit by jayrope at Sep 1, 2016 8:51:32 PM] |
||
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
Here's a quick excerpt from the list of progress:
ZIKA_ 000056080_ x2jly_ DV4_ NS3hlcs_ ssRNAdelMn_ chB_ 0163_ 0-- sehrneumaus-2.l You see how many invalid there are... |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Smells like your Mac Mini is producing a slightly different result from half of your verifying copy crunchers i.e. there being a homogeneity issue withing the Mac group, the particular OS-X version a possible (not the first time). That or your mini is broken in some way, overheating, bad memory, so maybe in need of some thorough testing and cleaning.
----------------------------------------Edit: The GPU is completely irrelevant, not at all used at WCG but to show the graphics candy (run in a separate process, and only when you opt to show it) [Edit 1 times, last edit by SekeRob* at Aug 13, 2016 1:24:22 AM] |
||
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
Yes, the computers use two different OS versions.
----------------------------------------I tested the hardware, it checks out well. Who might need to hear about an incongruity in their software? Boinc? They only seem to offer one software for all OSX versions 10.4 or later. [Edit 1 times, last edit by jayrope at Aug 17, 2016 9:06:10 AM] |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7574 Status: Offline Project Badges: 
|
I believe that i5 is a dual core with hyperthreading. If you are running it with the HT on, try turning that off and see if that gets rid of the invalids. If you are not using HT, try only using one core (no HT) and see if the problem persists. You say the hardware checks out fine, but sometimes running under full load in that tiny case for long periods can cause unforeseen problems.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1670 Status: Offline Project Badges:
|
I experience problem as well with a lot of invalid results on a specific host (Athlon II x4, LinuxMint 17.2 x64). The host is normally OK (no particular trouble) but, from time to time, with Vina-based project, there are a lot of invalid results.
----------------------------------------I have a second identical host (same CPU model, same OS). I only assume that there is some differences between the both CPU versions. Cheers, Yves |
||
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
Sgt.Joe, thanx for the suggestions. -
I am running this Mac mini only at 50% load, as otherwise it makes a lot of fan noise. Turning off hyperthreading didn't change anything unfortunately, the results still check out at around 90% invalid. I'll have the machine work on another project now and see, how that work checks out. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1670 Status: Offline Project Badges:
|
Since the release of the first project based on Vina, I experience unexplainable invalid results on a significantly higher rate than for projects for example based on autodoc.
----------------------------------------Initially, I supposed that it was related to a specific AMD CPU (Athlon II x4), later I notice that such invalid results happen on other CPU. In 99.9% of the cases, it happens for Linux-based hosts. I do not have any idea about the possible causes excepted that maybe Vina performs so many rounding with intermediate results that, at the end, the final results could differ between hosts. (It is the only one speculation I keep over the years). In few words, Vina-based projects are particularly sensible in terms of computing platform - CPU, OS - and they cause a higher rate of invalid results than other projects. Cheers, Yves |
||
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
KerSamson, but my mac is not Linux-based.
|
||
|
|
jayrope
Cruncher Joined: Jul 23, 2016 Post Count: 46 Status: Offline |
So.
----------------------------------------Upon further checking i found out, that Mapping Cancer Markers WUs check out 100% valid on my Mac mini 10.7.5, whereas ZIKA leads to over 90% WUs invalid results. In other words, it is not my system config. How can i let the ZIKA Wu creators know? [Edit 1 times, last edit by jayrope at Sep 1, 2016 8:41:32 PM] |
||
|
|
|