Remaining time is running upwards!?


log in

Advanced search

Questions and Answers : GPU applications : Remaining time is running upwards!?

Author Message
Profile champ
Volunteer tester
Avatar
Send message
Joined: 12 Mar 03
Posts: 3642
Credit: 1,489,147
RAC: 0
Germany
Message 841078 - Posted: 17 Dec 2008, 21:14:48 UTC

Have successful installed the new version of BOINC. CUDA was found, and it look like that it is running. But since i am running CUDA, the remaining time is running upwards. Have i done something wrong?

17.12.2008 21:57:48||Starting BOINC client version 6.4.5 for windows_intelx86
17.12.2008 21:57:48||log flags: task, file_xfer, sched_ops
17.12.2008 21:57:48||Libraries: libcurl/7.19.0 OpenSSL/0.9.8i zlib/1.2.3
17.12.2008 21:57:48|SETI@home|Found app_info.xml; using anonymous platform
17.12.2008 21:57:48||Processor: 2 GenuineIntel Intel(R) Pentium(R) Dual CPU E2180 @ 2.00GHz [x86 Family 6 Model 15 Stepping 13]
17.12.2008 21:57:48||Processor features: fpu tsc pae nx sse sse2 mmx
17.12.2008 21:57:48||OS: Microsoft Windows XP: Home x86 Editon, Service Pack 3, (05.01.2600.00)
17.12.2008 21:57:48||Memory: 1022.10 MB physical, 2.40 GB virtual
17.12.2008 21:57:48||Disk: 232.88 GB total, 210.49 GB free
17.12.2008 21:57:48||Local time is UTC +1 hours
17.12.2008 21:57:48||CUDA devices found
17.12.2008 21:57:48||Coprocessor: GeForce 8500 GT (1)
17.12.2008 21:57:48||Version change (6.2.14 -> 6.4.5)
17.12.2008 21:57:48|SETI@home|URL: http://setiathome.berkeley.edu/; Computer ID: 4309594; location: work; project prefs: work
17.12.2008 21:57:48||Computer location: work
17.12.2008 21:57:48||General prefs: using separate prefs for work
17.12.2008 21:57:48||Preferences limit memory usage when active to 511.05MB
17.12.2008 21:57:48||Preferences limit memory usage when idle to 919.89MB
17.12.2008 21:57:48||Preferences limit disk usage to 100.00GB
17.12.2008 21:57:48||Running CPU benchmarks
17.12.2008 21:57:48||Suspending computation - running CPU benchmarks
17.12.2008 21:58:19||Benchmark results:
17.12.2008 21:58:19|| Number of CPUs: 2
17.12.2008 21:58:19|| 1895 floating point MIPS (Whetstone) per CPU
17.12.2008 21:58:19|| 4067 integer MIPS (Dhrystone) per CPU
17.12.2008 21:58:20|SETI@home|Starting 13no08ae.24387.16023.8.8.57_1
17.12.2008 21:58:20|SETI@home|Starting task 13no08ae.24387.16023.8.8.57_1 using setiathome_enhanced version 528
17.12.2008 21:58:20|SETI@home|Starting 22no08ae.12987.10706.12.8.247_1
17.12.2008 21:58:20|SETI@home|Starting task 22no08ae.12987.10706.12.8.247_1 using setiathome_enhanced version 528

____________

Profile Sergej O. S.
Volunteer tester
Send message
Joined: 29 Oct 08
Posts: 123
Credit: 44,886
RAC: 0
Russia
Message 841084 - Posted: 17 Dec 2008, 21:25:11 UTC - in response to Message 841078.

You're not running CUDA. Current CUDA is version 6.06.

17/12/2008 17:48:25|SETI@home Beta Test|Starting task 03no08aa.27103.277504.4.11.137_0 using setiathome_enhanced version 605


____________

Profile champ
Volunteer tester
Avatar
Send message
Joined: 12 Mar 03
Posts: 3642
Credit: 1,489,147
RAC: 0
Germany
Message 841092 - Posted: 17 Dec 2008, 21:36:11 UTC

I thought 6.06 is the Beta version.

But what have i done wrong?
____________

Profile Sergej O. S.
Volunteer tester
Send message
Joined: 29 Oct 08
Posts: 123
Credit: 44,886
RAC: 0
Russia
Message 841098 - Posted: 17 Dec 2008, 21:48:58 UTC - in response to Message 841092.

Right, my bad.
Anyway, you need 6.05 to run CUDA. Detach from project and then attach again using email adress and password. Detach is need to delete your old programs and get the new 6.05, but it also will delete WUs, too.
____________

Profile champ
Volunteer tester
Avatar
Send message
Joined: 12 Mar 03
Posts: 3642
Credit: 1,489,147
RAC: 0
Germany
Message 841115 - Posted: 17 Dec 2008, 22:14:05 UTC

Thx for that help Sergej. Now i have CUDA WU´s, but they are still running upwards.
____________

Profile Sergej O. S.
Volunteer tester
Send message
Joined: 29 Oct 08
Posts: 123
Credit: 44,886
RAC: 0
Russia
Message 841125 - Posted: 17 Dec 2008, 22:28:26 UTC - in response to Message 841115.
Last modified: 17 Dec 2008, 22:49:33 UTC

Yes, where initial time to completition is 00:00:00 time is running up, but some tasks initially report time to completition of ex. 32 hours and time running down fast. I don't know how to fix it.
CUDA task takes on my 9800 GT 5-25 minutes, BTW, and if time go wrong, at some point it will go normally.
____________

Profile champ
Volunteer tester
Avatar
Send message
Joined: 12 Mar 03
Posts: 3642
Credit: 1,489,147
RAC: 0
Germany
Message 841145 - Posted: 17 Dec 2008, 22:52:25 UTC

Well, it does not look good for me. I have now one hour crunch time and 12% completed. If this is normal and will continue, i will go back to opt. version 5.28. There i can use both CPU´s on my Core Duo. I am using a NVIDIA GForce 8500 GT. Looks to me, this one is to slow for CUDA.
____________

Profile BigWaveSurfer
Send message
Joined: 29 Nov 01
Posts: 169
Credit: 8,656,241
RAC: 6,887
United States
Message 841178 - Posted: 17 Dec 2008, 23:51:38 UTC

My system is running through non-ap work units at just under 2 minutes (14.70 credit)and has all 0's from the est. time. Once the WU starts it shows me how much time it has been working and how much time to finish....

However.... My AP WU says 15973:35:11 to finish right now!!!! It is dropping quickly but I am a little worried. It is at 0.670% finished with 28 minutes processed.

From what I can tell right now the GPU is woking on the 'normal' WU and zipping through that, but the CPU is running the AP work unit. It would take me about 60 hours to run through a AP WU running the previous version of Seti with my processor set to run at 75%.

I am wondering if the GPU will take over crunching the AP WU when it gets to it or if it will skip over it and let the CPU continue working on it. Right now I set my system to 100% usage.
____________

Profile Euan Holton
Avatar
Send message
Joined: 4 Sep 99
Posts: 65
Credit: 13,996,509
RAC: 7,041
United Kingdom
Message 842213 - Posted: 19 Dec 2008, 21:43:59 UTC - in response to Message 841178.

I am wondering if the GPU will take over crunching the AP WU when it gets to it or if it will skip over it and let the CPU continue working on it. Right now I set my system to 100% usage.

Astropulse is CPU only at the moment. Also, there is an issue with a server-side variable called Task Duration Correction Factor (DCF) which can be set to a remarkably high value when you start processing CUDA units. This causes the BOINC client to seriously overestimate the processing time required for Astropulse units and can potentially cause BOINC to go into 'panic' mode, stopping you from getting more SETI workunits until the Astropulse WUs clear and the BOINC server starts adjusting the DCF to more sane values.
____________

Profile BMaytum
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 101
Credit: 3,972,683
RAC: 4,059
United States
Message 842246 - Posted: 19 Dec 2008, 22:13:42 UTC - in response to Message 842213.

Yes I saw the bizarre DCF effect Euan described when I got 1 CUDA WU and 2 AstroPulse WUs from Seti@home Beta Test project. The Cuda WU finished in about 2 CPU minutes, but Boinc (v6.4.5) said the "Completion Time" estimate for the 2 AP WUs was ~5000 hours each! The completion time estimate dropped relatively quickly as the AP WUs were crunched (in High Priority mode), but their combined Competion Time was so high I could not download any more WUs to crunch until they were done.
____________
Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2014 x64

Profile BigWaveSurfer
Send message
Joined: 29 Nov 01
Posts: 169
Credit: 8,656,241
RAC: 6,887
United States
Message 842352 - Posted: 20 Dec 2008, 3:56:25 UTC - in response to Message 842213.

I am wondering if the GPU will take over crunching the AP WU when it gets to it or if it will skip over it and let the CPU continue working on it. Right now I set my system to 100% usage.

Astropulse is CPU only at the moment. Also, there is an issue with a server-side variable called Task Duration Correction Factor (DCF) which can be set to a remarkably high value when you start processing CUDA units. This causes the BOINC client to seriously overestimate the processing time required for Astropulse units and can potentially cause BOINC to go into 'panic' mode, stopping you from getting more SETI workunits until the Astropulse WUs clear and the BOINC server starts adjusting the DCF to more sane values.



Thanks, and yes it is saying 1894:24:10 till the AP WU is done, I am at 52% @ 36 hours in reality though. You are right, BOINC will NOT download any CUDA units to crunch until AP is done it seems, which really is not good! Oh well, hopefully they will fix it soon.
____________

Profile BMaytum
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 101
Credit: 3,972,683
RAC: 4,059
United States
Message 842609 - Posted: 20 Dec 2008, 18:17:08 UTC

Yesterday I had a non-AP WU that terminated with a Compute Error, whilst it was running the time To Complete was Increasing (not decreasing) AND the Progress% was stuck at 0.000%. That WU was crunched using 6.06-Cuda. See
http://setiathome.berkeley.edu/forum_thread.php?id=50793
____________
Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2014 x64

Questions and Answers : GPU applications : Remaining time is running upwards!?

Copyright © 2014 University of California