Linux CUDA 'Special' App finally available, featuring Low CPU use

Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 41 · 42 · 43 · 44 · 45 · 46 · 47 . . . 82 · Next

AuthorMessage
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3198
Credit: 64,269,958
RAC: 108,913
Australia
Message 1878503 - Posted: 16 Jul 2017, 5:52:19 UTC - in response to Message 1876744.  

. . @ TBar
. . GPU power consumption up about 10 to 15%, productivity up about 15%. GPU temp on GPU-0 now into the 60 C range even with fans at 90% in winter, keeping the room warm :). CPU use holding steady at 52%, I don't think I will bother crunching on the spare CPU core because normal AR Arecibo tasks now run in under 2 mins consistently (like clockwork). The app is finished in under 110 secs but BOINC/manager takes another 6 to 10 secs to close files and move the results to the upload Q. My humble little machine is churning out about 60 plus results per hour . . . Hoooeeeyyyy Bob!
. . The flies in the ointment are Guppis. Blc05 tasks are still taking about 5.2 to 5.4 mins :( Oh well nothing is perfect. :)

. . The 970s are doing significantly better than 1060s when the stops are pulled out. Now I will wonder how much improvement the 1060s might get from a Ryzen uograde ... 8^}


. . Oh well I was wrong yet again, the tasks that were completing in about 2 mins were not quite normal Arecibo tasks, their ARs were rather higher than average but not in the VHAR realm.

. . Having been able to look at the STDERR files for the results on some of them the numbers are more like this ...

     VHAR      AR > 1.1                1.3 mins
     QHAR      AR ~ 0.7 to 0.9         1.9 mins 
     NARA      AR ~ 0.42 to 0.44       2.8 mins
     LAR       AR ~ 0.36 to 0.38       3.2 mins

     BLC05     AR = very low           5.2 to 5.6 mins


. . It looks like I won't get to find out how the 1060s do under Ryzen, I seem to have bought a dud. ...

Stephen

:(
ID: 1878503 · Report as offensive     Reply Quote
Profile tazzduke
Volunteer tester

Send message
Joined: 15 Sep 07
Posts: 117
Credit: 17,652,674
RAC: 2,543
Australia
Message 1878518 - Posted: 16 Jul 2017, 9:25:24 UTC - in response to Message 1878503.  

Greetings Stephen

???? What did you just acquire that you seem to think is a dud ????

Regards
ID: 1878518 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3198
Credit: 64,269,958
RAC: 108,913
Australia
Message 1878530 - Posted: 16 Jul 2017, 13:59:19 UTC - in response to Message 1878518.  

Greetings Stephen

???? What did you just acquire that you seem to think is a dud ????

Regards


. . A new rig with an ASUS Prime X370 Pro mobo and a Ryzen7-1700 with Corair Vengeance 2666MHz ram.

. . Either I have bricked it (which I doubt because I have been super careful) or it is a dud. It will not run through the POST, gives no display and accepts no input. The lights come on and the fans and ODD spin up OK but then nothing ... :(

Stephen

:(
ID: 1878530 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1878553 - Posted: 16 Jul 2017, 16:21:45 UTC
Last modified: 16 Jul 2017, 16:38:16 UTC

I've been attempting to build a new version of BOINC that works in the Ubuntu 16.04.x Home folder. It's been mostly a Success with a couple Snafus. Has anyone running one of the newer versions of BOINC, around 7.6, noticed the view in the Tasks window has a tendency to jump to the bottom of the page during an Update? Sometimes it happens every few Updates, sometimes it takes over an hour or two before I'm looking at the bottom of the page. Other than that, and the time zone tags in the Event Log disappearing/reappearing, it seems to be working OK. Just wondering how many others have found the Tasks page has mysteriously moved to the bottom of the page. To give it a try, open it to the Tasks Tab and Sort it by Progress, with the Active Tasks at the Top of the Page.

I did compile a new build of zi3v in 16.04.2, but it doesn't seem to be much different than the current version. The OS display is a little different with BOINC 7.7.x, it names the version of Linux...for me anyway.
Operating System - Linux Ubuntu
Ubuntu 16.04.2 LTS [4.8.0-58-generic]
BOINC version - 7.7.0
ID: 1878553 · Report as offensive     Reply Quote
Profile tazzduke
Volunteer tester

Send message
Joined: 15 Sep 07
Posts: 117
Credit: 17,652,674
RAC: 2,543
Australia
Message 1878598 - Posted: 16 Jul 2017, 20:52:45 UTC - in response to Message 1878530.  

Hi Stephen

Sorry to hear, hope it works out for you.

Regards
ID: 1878598 · Report as offensive     Reply Quote
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1878678 - Posted: 17 Jul 2017, 4:29:30 UTC
Last modified: 17 Jul 2017, 4:30:24 UTC

Just a quick drop by to give a heads up, mostly directed at Petri, while sorting other stuff. Upcoming Cuda changes for [redacted] will likely require checking/modifications to all Warp Synchronous code. Both Baseline and Special.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1878678 · Report as offensive     Reply Quote
Gene Project Donor

Send message
Joined: 26 Apr 99
Posts: 99
Credit: 16,388,176
RAC: 27,777
United States
Message 1878691 - Posted: 17 Jul 2017, 5:21:20 UTC

@TBar re: tasks tab

Yes, I see the same effect in 7.6.33 (and it was the same in 7.6.22; and maybe before that...) My "workaround" is to sort tasks in REVERSE order, click second time on "progress" column, and then live with the running tasks filtering down and out of the bottom of the screen. There is also a similar effect in the "Transfers" tab, i.e. active transfers displayed at the top but the window selects the bottom portion of the list when the list size exceeds the window size.

Gene;
ID: 1878691 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1878743 - Posted: 17 Jul 2017, 14:55:49 UTC - in response to Message 1878691.  

Thanks for the reply. I remember the same problem, along with other problems, back in version 7.4.22. You have to go back to a 7.2.xx version to escape the bouncing page. The versions prior to 7.4 used wxWidgets 2.8, the newer versions use wxWidgets 3.0. Ubuntu versions above around 15.04 require wxWidgets 3. It's thought the problem is with wxWidgets. Compared to the current Linux Development version 7.4.22, both 7.6.33 and 7.7.x are much better as they appear to only suffer from the bouncing page and Event log time tags. Version 7.4.22 also has those problems along more 'serious' snafus. Hopefully we'll see something better than 7.4.22 posted as the Development version soon.
ID: 1878743 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1879072 - Posted: 19 Jul 2017, 17:36:22 UTC
Last modified: 19 Jul 2017, 17:39:30 UTC

I'm going to declare Mission Accomplished.

I Now have a version of BOINC 7.2.47 that;
1) Works in the Home Folder of 16.04.x where it Doesn't require Special Permissions
2) Built with Static wxWidgets 2.8.12 Libraries, meaning No LibWx Dependencies
3) Works with Ubuntu 12.04, 14.04, and yes, 16.04.x and above
4) Doesn't suffer from the Jumping Tasks Tab Page or the Disappearing Event Log Time tags which Every Higher Version I've tested Does Suffer from.
5) Very easy to install, if you Don't already have a BOINC Folder, Expand to your Home Folder and double click on boincmgr
6) If you have an existing BOINC Folder in your Home Folder, simply copy the five expanded files and paste them into your existing BOINC Folder

My three Linux machines are running it. First time there's been a worthy replacement for 7.2.33 in Ages.
Now running Linux is even Easier...

ID: 1879072 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3198
Credit: 64,269,958
RAC: 108,913
Australia
Message 1879110 - Posted: 19 Jul 2017, 22:41:02 UTC - in response to Message 1879072.  
Last modified: 19 Jul 2017, 22:41:44 UTC

I'm going to declare Mission Accomplished.
My three Linux machines are running it. First time there's been a worthy replacement for 7.2.33 in Ages.
Now running Linux is even Easier...


. . OK, when I get the new rig working I will try 16.04 LTS and this ...

Stephen

8^}
ID: 1879110 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1879131 - Posted: 20 Jul 2017, 0:31:28 UTC - in response to Message 1879110.  

I've been trying to decide what to do with 'it'. After reading the ReadMe here, https://github.com/BOINC/boinc/blob/master/README.md, I decided to just go ahead and post it at C.A. None of the code has been changed and according to the link you can pretty much give it to anyone. This version is much better than the current 7.4.22 Development version Berkeley has posted and works in the Newer Linux versions where the current Recommended version doesn't. Over the past week I've built and tested numerous versions and none of them had any problems other than the Jumping Tasks page and disappearing time tags. It's a mystery why the current 7.4.22 version preforms so badly. This version, 7.2.47, doesn't have those problems and should be tested on other machines. It's basically what you would get after downloading and running the .sh file from Berkeley. If you plan on using the run_manager and run_client files you need to first open them and change the word user to the name of your Home folder.

I can see possibilities, you could just insert the CUDA Special package into the BOINC package and make it extremely easy to be up and running in a couple minutes. But really, how hard is it to just do that manually?
;-)
ID: 1879131 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1879141 - Posted: 20 Jul 2017, 1:37:39 UTC

Oh, the files are here, http://www.arkayn.us/forum/index.php?topic=197.msg4515#msg4515
I had to split them into two parts to upload them...you need both parts.
ID: 1879141 · Report as offensive     Reply Quote
Profile tazzduke
Volunteer tester

Send message
Joined: 15 Sep 07
Posts: 117
Credit: 17,652,674
RAC: 2,543
Australia
Message 1879408 - Posted: 21 Jul 2017, 10:03:21 UTC

Greetings TBAR

On my two MINT 18.2 machines am running the repo version 7.6.33 and the Special Cuda App, my GTX 950 running it will soon overtake my GTX 960 running SOG on WIN 7 lol.

Yeah I have no hiccups with this version of BOINC, and with a couple of mouseclicks and one cmdline entry and away we go.

The bug about sorting tasks no the way its meant to be doesnt bother me, I just click the tab bar to sort and its happy.

Regards
ID: 1879408 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 4013
Credit: 221,539,081
RAC: 236,749
United States
Message 1879429 - Posted: 21 Jul 2017, 14:30:53 UTC - in response to Message 1879408.  
Last modified: 21 Jul 2017, 14:37:15 UTC

After the test I ran, the chances of the AsyncRPC Bug just being on My machines is just about Zero. Everyone has to have this Bug since it's in the BOINC Source code. It will affect different Platforms in different ways. Some may not have Jumping Pages, they might have things such as, failure to download tasks during RPC Events. I'm surprised Almost No One has mentioned this problem with BOINC suddenly moving the page to the bottom during RPC Events. It's such an obvious Bug. Imagine My surprise when Every version of BOINC I compiled above 7.2.47 had this Bug in Linux. It seems using BOINC 7.2.33 for so long was a very good thing. Now I have 7.2.47 that works in 16.04...I'm set for a few years ;-)
ID: 1879429 · Report as offensive     Reply Quote
Profile -= Vyper =-
Volunteer tester
Avatar

Send message
Joined: 5 Sep 99
Posts: 1365
Credit: 527,415,294
RAC: 150,148
Sweden
Message 1879431 - Posted: 21 Jul 2017, 14:58:18 UTC

Hi TBar..

Just wanted you to know that i've switched my Quad 750Ti to your application now. If you perhaps want a machine to monitor missbehaviour within the executable but i presume that wont be much of an issue.
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8053171

Thanks for keeping it up!

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group
ID: 1879431 · Report as offensive     Reply Quote
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1511
Credit: 313,738,387
RAC: 368,343
Finland
Message 1879905 - Posted: 23 Jul 2017, 13:17:53 UTC - in response to Message 1878678.  

Just a quick drop by to give a heads up, mostly directed at Petri, while sorting other stuff. Upcoming Cuda changes for [redacted] will likely require checking/modifications to all Warp Synchronous code. Both Baseline and Special.

Thanks. Will keep an eye on that.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1879905 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3198
Credit: 64,269,958
RAC: 108,913
Australia
Message 1880659 - Posted: 27 Jul 2017, 23:39:52 UTC

. . Hi TBar

. . I am most distressed ... I have an invalid task that is not a noise bomb ...

http://setiathome.berkeley.edu/workunit.php?wuid=2616796976

. . It looks OK to me though it completed slightly quicker than the usual Guppi task. 4.1 mins compared to 5.4, whether or not that is any pointer to why it might have been marked as invalid.

Stephen

??
ID: 1880659 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 9618
Credit: 123,972,430
RAC: 84,599
Australia
Message 1880666 - Posted: 28 Jul 2017, 0:25:46 UTC - in response to Message 1880659.  
Last modified: 28 Jul 2017, 0:29:21 UTC

. . I am most distressed ... I have an invalid task that is not a noise bomb ...
http://setiathome.berkeley.edu/workunit.php?wuid=2616796976
. . It looks OK to me though it completed slightly quicker than the usual Guppi task.

Their results
Spike count:    1
Autocorr count: 2
Pulse count:    6
Triplet count:  0
Gaussian count: 0


Your result
Spike count:    9
Autocorr count: 2
Pulse count:    4
Triplet count:  15
Gaussian count: 0



EdIt- that's a bit odd, it started on Device 1, then stopped & restated on Device 2.

In cudaAcc_initializeDevice(): Boinc passed DevPref 1
setiathome_CUDA: CUDA Device 1 specified, checking...
   Device 1: GeForce GTX 970 is okay
SETI@home using CUDA accelerated device GeForce GTX 970
Unroll autotune 13. Overriding Pulse find periods per launch. Parameter -pfp set to 13

setiathome v8 enhanced x41p_zi3v, Cuda 8.00 special
CUDA 8.0 Special version by petri33.


setiathome_CUDA: CUDA Device 2 specified, checking...
   Device 2: GeForce GTX 970 is okay
SETI@home using CUDA accelerated device GeForce GTX 970
Unroll autotune 13. Overriding Pulse find periods per launch. Parameter -pfp set to 13
Restarted at 72.01 percent, with setiathome enhanced x41p_zi3v, Cuda 8.00 special
**** Detected setiathome_enhanced_v7 task ****. Autocorrelations enabled, size 128k elements.

Grant
Darwin NT
ID: 1880666 · Report as offensive     Reply Quote
Profile Brent Norman Special Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2081
Credit: 174,340,974
RAC: 493,979
Canada
Message 1880667 - Posted: 28 Jul 2017, 0:29:15 UTC - in response to Message 1880666.  

EdIt- that's a bit odd, it stared on Device 1, then stopped & restated on Device 2.
That would be it. The app really doesn't like to be restarted from checkpoint, and can throw errors - even on the same card.
ID: 1880667 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 9618
Credit: 123,972,430
RAC: 84,599
Australia
Message 1880668 - Posted: 28 Jul 2017, 0:31:26 UTC - in response to Message 1880667.  
Last modified: 28 Jul 2017, 0:31:53 UTC

EdIt- that's a bit odd, it stared on Device 1, then stopped & restated on Device 2.
That would be it. The app really doesn't like to be restarted from checkpoint, and can throw errors - even on the same card.

Hmm, bit of house keeping required? Clear out everything before re-loading, initialising & re-starting the WU?

Odd that on the restart it thought it was doing a v7 WU.
Grant
Darwin NT
ID: 1880668 · Report as offensive     Reply Quote
Previous · 1 . . . 41 · 42 · 43 · 44 · 45 · 46 · 47 . . . 82 · Next

Message boards : Number crunching : Linux CUDA 'Special' App finally available, featuring Low CPU use


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.