Completed, validation inconclusive
log in

Advanced search

Message boards : Number crunching : Completed, validation inconclusive

1 · 2 · 3 · Next
Author Message
fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 47 - Posted: 11 Jul 2009, 17:07:09 UTC

I am getting a bunch of these.

http://boinc.thesonntags.com/collatz/workunit.php?wuid=6098 is an example. Oddly, my machine finished the unit in 28 seconds, the other machine, a slightly more powerful card, took 750 seconds.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 48 - Posted: 11 Jul 2009, 17:39:19 UTC - in response to Message 47.

I am getting a bunch of these.

http://boinc.thesonntags.com/collatz/workunit.php?wuid=6098 is an example. Oddly, my machine finished the unit in 28 seconds, the other machine, a slightly more powerful card, took 750 seconds.


What drivers version you have?
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 49 - Posted: 11 Jul 2009, 17:48:06 UTC - in response to Message 48.
Last modified: 11 Jul 2009, 17:50:10 UTC

I am getting a bunch of these.

http://boinc.thesonntags.com/collatz/workunit.php?wuid=6098 is an example. Oddly, my machine finished the unit in 28 seconds, the other machine, a slightly more powerful card, took 750 seconds.


What drivers version you have?


Uhmm... you're returning incomplete results...

<core_client_version>6.3.21</core_client_version>
<![CDATA[
<stderr_txt>
Beginng processing
worker: trying boinc_get_init_data()...
2361183359331862292840

2361183359334009776488

Success in SetCUDABlockingSync for device 0
collatz_1247248134_1696
2361183359331862292840
2361183359334009776488




worker: CPU time: 28.046875 seconds, Wall clock time: 31.656252 seconds
called boinc_finish

</stderr_txt>
]]>

<core_client_version>6.6.28</core_client_version>
<![CDATA[
<stderr_txt>
Beginng processing
worker: trying boinc_get_init_data()...
2361183359331862292840

2361183359334009776488

Success in SetCUDABlockingSync for device 0
collatz_1247248134_1696
2361183359331862292840
2361183359334009776488
2361183359332471936274
1630
1276726917655

worker: CPU time: 557.968750 seconds, Wall clock time: 1103.390625 seconds
called boinc_finish

</stderr_txt>
]]>

Your BOINC client is very old... Could be that... Try to uptate it...
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 50 - Posted: 11 Jul 2009, 17:51:59 UTC

ForceWare 177.35

aqua@home didn't like that machine either. Guess I will leave it on gpugrid until I have time to do major maintenance on it.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 51 - Posted: 11 Jul 2009, 17:57:27 UTC - in response to Message 50.
Last modified: 11 Jul 2009, 18:13:59 UTC

ForceWare 177.35

aqua@home didn't like that machine either. Guess I will leave it on gpugrid until I have time to do major maintenance on it.


The version that works fine with CUDA is 185.50 and the lattest is 186.18.

Your's are very very older...:) With a lot of problems and bugs...

Try to update it too...;)

NVIDIA drivers


Best regards.
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 53 - Posted: 11 Jul 2009, 18:23:35 UTC

Yes, I need to upgrade both boinc and the drivers on that machine. I settled on that combination many months ago as it was stable and worked with GPUGRID. Finding the perfect match of forceware and boinc without having to tweak it daily is difficult. It took me several hours of reading the forums to find the combination that had the least number of complaints against it.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 55 - Posted: 11 Jul 2009, 18:42:59 UTC - in response to Message 53.
Last modified: 11 Jul 2009, 18:45:17 UTC

Yes, I need to upgrade both boinc and the drivers on that machine. I settled on that combination many months ago as it was stable and worked with GPUGRID. Finding the perfect match of forceware and boinc without having to tweak it daily is difficult. It took me several hours of reading the forums to find the combination that had the least number of complaints against it.


If you upgrade to the lattest drivers (185.50 or 186.18, both works fine), you will amazing with the improvement with speed in anyone of CUDA projects... If you decides to do it, please, let me know the results...;)
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 57 - Posted: 11 Jul 2009, 19:23:35 UTC - in response to Message 55.

Did you mean 185.85? The only .50 I see is 182.50. There is some debate on the gpugrid forums whether 185.85 is good or not. I still have the old install files for the ancient driver I am running, so I think I will give 185.85 a try.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 58 - Posted: 11 Jul 2009, 19:28:49 UTC - in response to Message 57.
Last modified: 11 Jul 2009, 19:33:34 UTC

Did you mean 185.85? The only .50 I see is 182.50. There is some debate on the gpugrid forums whether 185.85 is good or not. I still have the old install files for the ancient driver I am running, so I think I will give 185.85 a try.


Sorry... Yes. It's 185.85. 182.50 have problems too.(see NVIDIA site about differences and improvements between 182 and 185/186 versions)

Try with 185.85 if you want. With me worked fine, but I'm using now 186.18.;)
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 60 - Posted: 11 Jul 2009, 20:31:17 UTC - in response to Message 58.

I switched to 185.85 and it completed a work unit in about 9 minutes, more typical of what I have seen before. The downside is this machine is a LOT more slugish, particularly when gaming.

What is more interesting is that one of my work units got a similarly broken wingman, and as such, we both got credit for it.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 61 - Posted: 11 Jul 2009, 20:36:52 UTC - in response to Message 60.

I switched to 185.85 and it completed a work unit in about 9 minutes, more typical of what I have seen before. The downside is this machine is a LOT more slugish, particularly when gaming.

What is more interesting is that one of my work units got a similarly broken wingman, and as such, we both got credit for it.


Tell him the solution...;)
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

fractal
Send message
Joined: 11 Jul 09
Posts: 14
Credit: 1,001,340,489
RAC: 0
Message 62 - Posted: 11 Jul 2009, 20:53:37 UTC

That was a hint, perhaps too subtle, that slicker might want to detect invalid results...

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 79 - Posted: 12 Jul 2009, 13:20:32 UTC - in response to Message 60.

I switched to 185.85 and it completed a work unit in about 9 minutes, more typical of what I have seen before. The downside is this machine is a LOT more slugish, particularly when gaming.

What is more interesting is that one of my work units got a similarly broken wingman, and as such, we both got credit for it.


That's not good. Now that the ATI GPU app is working okay, I guess I need to go back and take another look at the CUDA app and then manually recreate the WUs for any where this happened.

Profile STE\/E
Avatar
Send message
Joined: 12 Jul 09
Posts: 581
Credit: 761,710,729
RAC: 0
Message 100 - Posted: 12 Jul 2009, 20:09:29 UTC - in response to Message 79.
Last modified: 12 Jul 2009, 20:24:03 UTC

Now that the ATI GPU app is working okay


What ATI GPU Application & Driver is working okay Slicker, I tried some on a 4870 with 8.12 Drivers this morning and they all Failed within 10 Seconds ??? If they don't run with 8.12 Drivers then their in conflict with the Milkyway Project as that's about all that runs for most guys over there.

Also how long are we going to have to hold on to all the Finished Wu's we already have run with the Projects Application ... ???

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 102 - Posted: 12 Jul 2009, 21:15:55 UTC - in response to Message 100.

Now that the ATI GPU app is working okay


What ATI GPU Application & Driver is working okay Slicker, I tried some on a 4870 with 8.12 Drivers this morning and they all Failed within 10 Seconds ??? If they don't run with 8.12 Drivers then their in conflict with the Milkyway Project as that's about all that runs for most guys over there.

Also how long are we going to have to hold on to all the Finished Wu's we already have run with the Projects Application ... ???



1. Catalyst 8.12 works just fine. That's what I'm running.

2. MW and Collatz don't play well together. Since ATI isn't supported by BOINC, they step on each other and have no way of letting each other know that the other is using the GPU at the moment. Since BOINC only supports CUDA, the scheduler thinks of ATI apps as CPU apps and if you have 8 cpus, it runs 8 of them. MW can tell if other MW apps are running and Collatz can tell if other Collatz apps are running, but there isn't a good way (yet) for them to both run simultaneously. With Gipsel's help, we may come up with something at some point, but for now, you probably have to choose between MW and Collatz if using ATI apps.

3. If it fails immediately after it writes out "supporting double precision", then the next version should fix that.

4. As far as how long... you should be able to return and report results, but the work generator is set to generate only 1 WU per minute. That should let people finish their WUs and report them while I figure out whether the rogue apps are returning valid results or not.

Profile STE\/E
Avatar
Send message
Joined: 12 Jul 09
Posts: 581
Credit: 761,710,729
RAC: 0
Message 103 - Posted: 12 Jul 2009, 21:51:11 UTC
Last modified: 12 Jul 2009, 22:16:01 UTC

Okay & Thanks Slicker, I'll try Suspending Milkyway on the ATI Cards when there's more work to get and see if they run okay that way ...

PS: They work on my i7 with a 4870 but are going to take well over an hour to do, probably in the 1:40:00 Range, much slower than the NVIDIA Cards ...

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 104 - Posted: 12 Jul 2009, 22:35:26 UTC - in response to Message 103.

Okay & Thanks Slicker, I'll try Suspending Milkyway on the ATI Cards when there's more work to get and see if they run okay that way ...

PS: They work on my i7 with a 4870 but are going to take well over an hour to do, probably in the 1:40:00 Range, much slower than the NVIDIA Cards ...


1:40 is probably the elapsed time. Boinc starts 8 of them but only 1 can run on the kernel at a time. The app allows 2 to compete for the GPU at a time and puts the other 6 on hold. Boinc doesn't know about that so it keeps counting the elapsed seconds. When one of them finishes, one of the 6 that was on hold begins processing. Boinc, of course, start another one so there are always 2 running and 6 that show running buit really aren't.

The actual processing time should be 0.125 * 1:40. If you look at the "GPU time" in the output, you'll see that they finish much faster than 1:40.

For proof:
Set ncpus to 1 in the cc_config.xml, set avg_cpus to 1.0 and set -n1 (e.g. "dash N one") for the cmdline parameter in the app_info.xml

That will make your machine think it is a single core and only 1 WU will run. It should finish in around 6 minutes. Then try again with -n2 and avg_cpus = 0.5 and they should both finish in around 12 minutes. Increasing the number of cpus (ncpus set to 2) will start 4 but only 2 will run and after the first two finish in 12 minutes, the other two will show double that or more.

Profile STE\/E
Avatar
Send message
Joined: 12 Jul 09
Posts: 581
Credit: 761,710,729
RAC: 0
Message 106 - Posted: 13 Jul 2009, 0:31:11 UTC
Last modified: 13 Jul 2009, 0:44:50 UTC

It doesn't seem to matter what I put in the Settings, either 6 or 7 Wu's all start @ once, all show the same amount of time Progression & all show the same amount of Percentage Progression. And all figure to finish in 1 hr 40 Minutes running time going by the amount of Time Progression v Percentage Progression.

I should say I have a 4870 & a 4850 X2 in the i7 Box, that could/may make a difference I suppose. But all my ATI Box's have Dual Cards & a mixture of X2's so they would all act the same way I assume. HT is enabled on the i7 too if that makes a difference ... ???

Also what does the -k128 Setting supposed to do in the Application ... Thanks

Rob.B
Send message
Joined: 30 Jul 09
Posts: 28
Credit: 11,805,038
RAC: 0
Message 396 - Posted: 4 Aug 2009, 18:31:31 UTC

Sorry to ask a dumb question but what causes the "Validation Inconclusive" status to be flagged? Additionally how is it resolved?

http://boinc.thesonntags.com/collatz/workunit.php?wuid=124074

Rob.B

Liuqyn
Send message
Joined: 8 Jul 09
Posts: 26
Credit: 164,516,656
RAC: 0
Message 407 - Posted: 5 Aug 2009, 16:38:52 UTC - in response to Message 396.

first there are no dumb questions. as for an answer, I believe it means that the different hosts didn't agree on the result at which point the server has to send the wu out to another host.

1 · 2 · 3 · Next
Post to thread

Message boards : Number crunching : Completed, validation inconclusive


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.