Chunching has dropped to 1/3 or less

liderbug
liderbug
Joined: 13 Aug 10
Posts: 20
Credit: 37561962
RAC: 0
Topic 198666

(prev post didn't get all text)

My desktop: ASUS with an AMD 6 core @ 3.4 w raid. I was running FC-19 and Boinc (ver, I forget)
. I was getting right at 30,000 credits/day. Then I got this wild hair to upgrade to FC-22 non-raid. I reload Boinc via DNF logged in with MGR, I'm an existing.... and away we went ... at about 3,000 to 8,000 credits/day ... WT????
http://lidderdale.net/images/boinc.jpg
Boinc says "not our problem call your projects". On boincstats.com I'm liderbug. I've tried various compute configs - nothing changes.

Suggestions?
Thanks

Christian Beer
Christian Beer
Joined: 9 Feb 05
Posts: 595
Credit: 125616083
RAC: 337133

Chunching has dropped to 1/3 or less

The image you posted is for all your projects. Could you look up those images per project? You will then see which project is responsible for this drop. If they all drop then something on your new system is slowing tasks down.

Normally I would suggest to compare runtimes of Eisntein@Home tasks before and after the upgrade but this is not possible in your case because those old tasks are already purged from the system. If you are lucky you may find a file on you computer that has the timings for the old tasks. This is usually in the BOINC data directory (something like /var/lib/boinc on FC). There was a post in the forums a while ago that explained this file. I can't find it right now maybe someone else can point you to it.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2140
Credit: 2769874726
RAC: 934927

I think the file Christian is

I think the file Christian is referring to is the one called 'job_log_einstein.phys.uwm.edu.txt'. If the OS upgrade was performed in-place over the existing installation, it should still be present in the location stated, and should contain entries from both before and after the change in OS - they will provide the useful comparison. If the hard drive was reformatted without an external backup first, of course, that won't be possible.

A sample entry from my log reads

1467194835 ue 34886.100421 ct 647.388500 fe 590000000000000 nm PM0144_01111_314_0 et 30921.194592 es 0 I would break that down as

[pre]1467194835 // unix timestamp of task completion
ue 34886.100421 // original runtime estimate
ct 647.388500 // CPU time used
fe 590000000000000 // FLOPs estimate (task size)
nm PM0144_01111_314_0 // task name
et 30921.194592 // elapsed time (task run time)
es 0[/pre]

liderbug
liderbug
Joined: 13 Aug 10
Posts: 20
Credit: 37561962
RAC: 0

So here are the last few

So here are the last few lines of the current file. I don't understand what I'm looking at. IE. ct 58126 ?? should that be 158126? or 126??

466782195 ue 24783.827339 ct 58126.060000 fe 144000000000000 nm h1_0097.90_O1C01Cl2In3__O1AS20-100F_98.00Hz_624_1 et 58698.937601 es 0
1466864557 ue 24783.827339 ct 58024.290000 fe 144000000000000 nm h1_0097.90_O1C01Cl2In3__O1AS20-100F_98.00Hz_502_1 et 58561.320167 es 0
1466948354 ue 24783.827339 ct 58037.840000 fe 144000000000000 nm h1_0098.00_O1C01Cl2In3__O1AS20-100F_98.10Hz_412_1 et 58656.718024 es 0
1467032305 ue 24783.827339 ct 58118.120000 fe 144000000000000 nm h1_0098.00_O1C01Cl2In3__O1AS20-100F_98.10Hz_413_1 et 58764.781657 es 0

Also "Could you look up those images per project?" How? I've been all over the BAM page and the graph only shows "Combined".

Thanks

Christian Beer
Christian Beer
Joined: 9 Feb 05
Posts: 595
Credit: 125616083
RAC: 337133

You have the right file. You

You have the right file. You need to check from when the first (oldest) entry is from. For this you need to convert the first column from unixtime to normal time using something like http://www.unixtimestamp.com/. If this is before the FC 19->22 migration than you need to see if the runtime in the third column (the number after ct) is increasing at some point and you can then convert the timestamp to normaltime and see if this is around the date where you migrated. You can do this for the other projects on those computers too.
If the oldest entry in the file is the time you migrated than you can't compare runtime before and after and habve to look for other reasons. Maybe one of your projects ran out of work and another project that is not giving you this much Credits took it's place.

Btw: Here is the breackdown of projects for you: http://boincstats.com/en/stats/-1/user/detail/549393/projectList By clicking on the little image on the right hand side of each project line you can see details for this project. On this page there is a button "Charts" where you can see the chart for each individual project. For Einstein@home I see the same drop as can be seen in your BOINC combined image and it seems to be the most significant one but WCG and Asteroids Numbers are also lower after the migration.

What Einstein@home Apps did you run before the Update? If you used to run GPU apps before and now you are not (because the GPU is not recognized by BOINC e.g.) that would explain the drop in Credit. You should check your BOINC Manager log messages on all three machines.

liderbug
liderbug
Joined: 13 Aug 10
Posts: 20
Credit: 37561962
RAC: 0

As to what I was running

As to what I was running under FC19 I can't say as I had 3 500Gb drives raided. I kept getting a drive may be failing (see below) so I was looking for an excuse to upgrade from FC19 to 22 - and UNraid at the same time. Bottom line /home/chuck got backed up. /var/lib/boinc ... formated. So the first time listed would be the day I reloaded Boinc.
1464924909
Is equivalent to:
06/03/2016 @ 3:35am (UTC)

GPU, the web page listing Video cards doesn't list mine - but maybe 4 years ago it did and worked. Is it possible that I/it was using GPU and now not?

product: G92 [GeForce GTS 250]
vendor: NVIDIA Corporation

Is there a way to force GPU? The only thing I can find is Don't use GPU when computer is in use.

Drive-Failure: Seems there is a counter on drives that SMART reads how many times the drive has been started/stopped (or one of those counts) and at a given count it start saying "poss drive failure" when the drive isn't failing.

liderbug
liderbug
Joined: 13 Aug 10
Posts: 20
Credit: 37561962
RAC: 0

I'm thinking FC19 was using

I'm thinking FC19 was using CUDA/NVIDIA drivers. So I'm searching "boinc cuda" - "just load xorg-x11-drv-nvidia-libs" -> No package avail. OK Hey Google, ya go here "... for FC-10 ... " WT??? and after a while Isn't this the page I was on 1/2 hr ago???

liderbug
liderbug
Joined: 13 Aug 10
Posts: 20
Credit: 37561962
RAC: 0

SOLVED. (good news) (the

SOLVED. (good news)

(the bad) Not sure how I did it... but I have a Seti ... CUDA(60).
I downloaded
NVIDIA-Linux-x86_64-340.96.run
xorg-x11-drv-nvidia-libs-346...rpm
dnf install cuda
dnf install kernel-devel kernel-headers
dnf install libvdpau
dnf install xorg-x11-drv-nvidia
dnf install kmod kmod-libs kmod-devel libmikmod libmikmod-devel mikmod
dnf install $(nvidia-detect)
dnf install vdpauinfo libva-vdpau-driver libva-utils
dnf install plymouth-theme-solar

and somehow I got/get:
CUDA: NVIDIA GPU 0: GeForce GTS 250 (driver version 340.96, CUDA version 6.5, compute capability 1.1, 1023MB, 943MB available, 622 GFLOPS peak)
07-Jul-2016 07:02:29 [---] OpenCL: NVIDIA GPU 0: GeForce GTS 250 (driver version 340.96, device version OpenCL 1.0 CUDA, 1023MB, 943MB available, 622 GFLOPS peak)
07-Jul-2016 07:02:29 [---] [coproc] NVIDIA library reports 1 GPU

So it looks like I may be back - time will tell.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.