SETI@home v8.22 Windows GPU applications support thread

Message boards : Number crunching : SETI@home v8.22 Windows GPU applications support thread
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 7 · Next

AuthorMessage
Stephen "Heretic"Project Donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 2634
Credit: 48,253,849
RAC: 132,176
Australia
Message 1846396 - Posted: 4 Feb 2017, 4:26:24 UTC - in response to Message 1845540.  
Last modified: 4 Feb 2017, 4:28:44 UTC

It is using 1 core almost all of the time.....

boincmanager 7.6.33
NVIDIA GeForce GTX 1060 6GB (4095MB) driver: 376.33 OpenCL: 1.2
Win 10 home and an GenuineIntel Intel(R) Core(TM) i5-6400 CPU @ 2.70GHz [Family 6 Model 94 Stepping 3]
(4 processors)

I am running stock "SoG" app

Any idea......?


. . It is normal for SoG to fully utilise one CPU core for each task running on your GPU. If you want to change that turn "Sleep" on with the command -use sleep in your command line file, "mb_cmdline_win_x86_SSE3_OpenCl_NV_SoG.txt" if you have an Nvidia GPU.

Stephen

{PS} and as usual I am late to the party :).

.
ID: 1846396 · Report as offensive     Reply Quote
Profile TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 4 Oct 99
Posts: 499
Credit: 44,486,685
RAC: 38,572
Sweden
Message 1846875 - Posted: 5 Feb 2017, 20:35:36 UTC
Last modified: 5 Feb 2017, 20:37:30 UTC

More strange questions from me....:)

I am running MB8_win_x86_SSE2_OpenCL_ATi_HD5_SoG_r3584.exe from Mike's site.
ATI 5850 card...

I can't really see any difference in speed when using that SoG one or ATI_HD5 .r3500 version(i think it was from lunatics).

Am I doing anything wrong, missed something or is the apps about the same speed for ati 5850?

-------------------------------------------------

In boincmanager I see under /Tasks/program it says: lokal: setiathome_v8 8.22 (ati_opencl_sah)
In windows task manager it says i am running "MB8_win_x86_SSE2_OpenCL_ATi_HD5_SoG_r3584.exe"

And in app_config.xml the line:
<plan_class>opencl_ati_sah</plan_class>

If I look the same place for nvidia card it says: setiathome v8 8.22 opencl_nvidia_sog

But since I don't master in programming I have no clue if both is ok Nvidia and ati

Link to ati computor.... if needed.

https://setiathome.berkeley.edu/results.php?hostid=7358242

//TRuEQ
ID: 1846875 · Report as offensive     Reply Quote
Profile MikeProject Donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 30603
Credit: 57,610,184
RAC: 30,374
Germany
Message 1846917 - Posted: 5 Feb 2017, 23:20:21 UTC

no surprise to me.
With r_3500 you are using _period_iterations_num 3 and with r_3584 you are using no app args.
With each crime and every kindness we birth our future.
ID: 1846917 · Report as offensive     Reply Quote
Profile TRuEQ & TuVaLu
Volunteer tester
Avatar

Send message
Joined: 4 Oct 99
Posts: 499
Credit: 44,486,685
RAC: 38,572
Sweden
Message 1847003 - Posted: 6 Feb 2017, 10:25:23 UTC - in response to Message 1846917.  

no surprise to me.
With r_3500 you are using _period_iterations_num 3 and with r_3584 you are using no app args.



I added -period_iterations_num 5 -hp now....
See tomorrow how it looks.
Thanks...
ID: 1847003 · Report as offensive     Reply Quote
hsdecalcProject Donor

Send message
Joined: 1 Feb 15
Posts: 2
Credit: 1,190,520
RAC: 2,324
Germany
Message 1848347 - Posted: 12 Feb 2017, 15:26:06 UTC

I wonder that the task priority is put on "high" instead of "low". It's a bug or a feature?
Running Win 10 with setiathome-8.22_windows_intelx86__opencl_nvidia_SoG.exe
Inside cc_config.xml I use the param: <process_priority_special>1</process_priority_special>

stderr.txt:
Running on device number: 0
Sleep() & wait for event loops will be used in some places
Priority of worker thread raised successfully
Priority of process adjusted successfully, high priority class used
OpenCL platform detected: NVIDIA Corporation
BOINC assigns device 0
Info: BOINC provided OpenCL device ID used

...
ID: 1848347 · Report as offensive     Reply Quote
hsdecalcProject Donor

Send message
Joined: 1 Feb 15
Posts: 2
Credit: 1,190,520
RAC: 2,324
Germany
Message 1848366 - Posted: 12 Feb 2017, 17:13:31 UTC - in response to Message 1848347.  

task priority is put on "high" instead of "low". It's a bug or a feature?
...

Ok, found the feature: I remove line -hp in file "mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt".
ID: 1848366 · Report as offensive     Reply Quote
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 5809
Credit: 76,063,171
RAC: 51,199
Russia
Message 1848402 - Posted: 12 Feb 2017, 22:54:07 UTC - in response to Message 1848366.  

Always nice to support self-repairing :)
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1848402 · Report as offensive     Reply Quote
Garry

Send message
Joined: 7 Jul 02
Posts: 26
Credit: 409,758
RAC: 197
United States
Message 1848689 - Posted: 15 Feb 2017, 2:24:19 UTC - in response to Message 1848366.  

task priority is put on "high" instead of "low". It's a bug or a feature?
...


[This is a great stream to watch. I don't have to talk much; I can learn a lot. Thanks to all.]

In directory BOINC/slots/x (used by S@H), I see (line 4 for me): "Priority of worker thread raised successfully"

If it's a short explanation: What does this mean and what's the impact on my computer?

[The answer about "line -hp in file "mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt"." is ... maybe an answer to a different question? That set of files (one per application) has zero length on my system until I put something in it.]
ID: 1848689 · Report as offensive     Reply Quote
Profile Keith Myers
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 2433
Credit: 184,339,493
RAC: 359,223
United States
Message 1848700 - Posted: 15 Feb 2017, 3:05:23 UTC - in response to Message 1848689.  

task priority is put on "high" instead of "low". It's a bug or a feature?
...


[This is a great stream to watch. I don't have to talk much; I can learn a lot. Thanks to all.]

In directory BOINC/slots/x (used by S@H), I see (line 4 for me): "Priority of worker thread raised successfully"

If it's a short explanation: What does this mean and what's the impact on my computer?

[The answer about "line -hp in file "mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt"." is ... maybe an answer to a different question? That set of files (one per application) has zero length on my system until I put something in it.]

The SETI GPU/CPU tasks have an inherent System Idle Priority as written. They leave the CPU tasks alone which run at Idle System Priority, but bump up the GPU tasks by default to the next highest level above Idle, which is Below Normal System Priority so they get a little more timeslices since they need to be fed regularly by the CPU. You can alter the priority of the tasks either by calling out -hp in the "mb_cmdline-8.22 ~~platform~~.txt file which is delivered blank and ready for customization. Or you can call out a commandline parameter in a app_config file too as an alternative way to alter the task running priority. Another way to alter the stock priority level is with third-party apps like ProcessLasso to make any application run at whatever priority level you desire. The stock applications try to put the least amount of demand on the CPU/GPU resources to accommodate lesser hardware platforms. You just see the output of the GPU app stderr.txt showing the bump in process priority at each loading of a new task. The developers choose how they want their app to run. Some apps have considerable parameters that you can choose to alter the way the app runs on your system. Some apps are pretty locked down in how they can be run. When all else fails .... read the README.TXT files for each app to see what choices you have in running the task app.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1848700 · Report as offensive     Reply Quote
James Nelson
Volunteer tester
Avatar

Send message
Joined: 23 Mar 02
Posts: 381
Credit: 4,450,832
RAC: 1,033
United States
Message 1854409 - Posted: 10 Mar 2017, 16:48:01 UTC - in response to Message 1842497.  

is there an installer avaliable for this yet don't want to try manual and screw my setup up.
ID: 1854409 · Report as offensive     Reply Quote
Stephen "Heretic"Project Donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 2634
Credit: 48,253,849
RAC: 132,176
Australia
Message 1854685 - Posted: 11 Mar 2017, 8:07:07 UTC

. . Hi All,

. . I am toying around with Linux and it is a very strange land. Does the SoG or SaH app in Linux respond to the commandline file as with the Windows version? And what is the naming protocol? There is no blank template there to work off.

Stephen

??
ID: 1854685 · Report as offensive     Reply Quote
Profile MikeProject Donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 30603
Credit: 57,610,184
RAC: 30,374
Germany
Message 1854695 - Posted: 11 Mar 2017, 8:43:45 UTC - in response to Message 1854685.  
Last modified: 11 Mar 2017, 8:46:02 UTC

. . Hi All,

. . I am toying around with Linux and it is a very strange land. Does the SoG or SaH app in Linux respond to the commandline file as with the Windows version? And what is the naming protocol? There is no blank template there to work off.

Stephen

??


Good question Stephen.
I reprted this over at Lunatics but never got a good answer for it
I did some apha testing with different OpenCL apps on Mint 17.x.
To be honest i never got the app args working from comand line text file.
So i use either the cc_config.xml or appinfo.xml to adjust the comands.

Not using mb_cmdline.txt-file, using commandline options.
Number of period iterations for PulseFind set to 30
Running on device number: 0
Maximum single buffer size set to:384MB
SpikeFind FFT size threshold override set to:2048
TUNE: kernel 1 now has workgroup size of (64,1,4)
oclFFT global radix override set to:256
oclFFT local radix override set to:16
oclFFT max WG size override set to:256
oclFFT max local FFT size override set to:512
oclFFT number of local memory banks set to:32
oclFFT minimal memory coalesce width set to:32
Running on device number: 0
With each crime and every kindness we birth our future.
ID: 1854695 · Report as offensive     Reply Quote
Stephen "Heretic"Project Donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 2634
Credit: 48,253,849
RAC: 132,176
Australia
Message 1854717 - Posted: 11 Mar 2017, 9:52:30 UTC - in response to Message 1854695.  

. . Hi All,

. . I am toying around with Linux and it is a very strange land. Does the SoG or SaH app in Linux respond to the commandline file as with the Windows version? And what is the naming protocol? There is no blank template there to work off.

Stephen

??


Good question Stephen.
I reprted this over at Lunatics but never got a good answer for it
I did some apha testing with different OpenCL apps on Mint 17.x.
To be honest i never got the app args working from comand line text file.
So i use either the cc_config.xml or appinfo.xml to adjust the comands.

Not using mb_cmdline.txt-file, using commandline options.
Number of period iterations for PulseFind set to 30
Running on device number: 0
Maximum single buffer size set to:384MB
SpikeFind FFT size threshold override set to:2048
TUNE: kernel 1 now has workgroup size of (64,1,4)
oclFFT global radix override set to:256
oclFFT local radix override set to:16
oclFFT max WG size override set to:256
oclFFT max local FFT size override set to:512
oclFFT number of local memory banks set to:32
oclFFT minimal memory coalesce width set to:32
Running on device number: 0


. . Well I have heard scary things about playing with ap_info.xml so I might give that a miss but I am hoping cc_config.xml is much like ap_config.xml and fairly safe to use.

Stephen

:)
ID: 1854717 · Report as offensive     Reply Quote
Profile MikeProject Donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 30603
Credit: 57,610,184
RAC: 30,374
Germany
Message 1854783 - Posted: 11 Mar 2017, 16:31:10 UTC

You have to put the comand line into app_config.xml

</gpu_versions>
<cmdline>-device 0 -sbs 384 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32</cmdline>
</app>
With each crime and every kindness we birth our future.
ID: 1854783 · Report as offensive     Reply Quote
Stephen "Heretic"Project Donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 2634
Credit: 48,253,849
RAC: 132,176
Australia
Message 1854879 - Posted: 11 Mar 2017, 23:03:22 UTC - in response to Message 1854783.  

You have to put the comand line into app_config.xml

</gpu_versions>
<cmdline>-device 0 -sbs 384 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32</cmdline>
</app>


. . Thanks for that Mike.

Stephen

.
ID: 1854879 · Report as offensive     Reply Quote
Profile RueiKeProject Donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 270
Credit: 104,215,175
RAC: 235,985
Taiwan
Message 1858212 - Posted: 29 Mar 2017, 9:58:59 UTC

Recently I got 4 WU with computation errors using this app: SETI@home v8 v8.22 (opencl_ati5_SoG_nocal) windows_intelx86.
Here is one example: https://setiathome.berkeley.edu/result.php?resultid=5618949723
All on the same host, running default apps. Its with a FuryX GPU and gets consistently good results for SETI@home v8 v8.22 (opencl_ati5_nocal)
YouTube Channel: Rick's Performance Computing
ID: 1858212 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 8887
Credit: 115,101,087
RAC: 70,205
Australia
Message 1858220 - Posted: 29 Mar 2017, 10:50:50 UTC - in response to Message 1858212.  
Last modified: 29 Mar 2017, 10:52:19 UTC

Recently I got 4 WU with computation errors using this app: SETI@home v8 v8.22 (opencl_ati5_SoG_nocal) windows_intelx86.
Here is one example: https://setiathome.berkeley.edu/result.php?resultid=5618949723

I had a similar "too many restarts" error once. For some reason it got upset about the video driver, a reboot solved the issue.

EDIT- I notice it's a Win10 system- it hasn't updated the video driver by itself by any chance?
Grant
Darwin NT
ID: 1858220 · Report as offensive     Reply Quote
FloydHung

Send message
Joined: 22 Jul 99
Posts: 7
Credit: 1,054,014
RAC: 1,110
Taiwan
Message 1859553 - Posted: 5 Apr 2017, 9:14:35 UTC

All of my Intel GPU work failed to finish completely. One of the WU is located at https://setiathome.berkeley.edu/result.php?resultid=5620728108.
According to the error message, it seems it could not open a binary kernel file. Something wrong or missing in my computer?
ID: 1859553 · Report as offensive     Reply Quote
Kiska
Volunteer tester

Send message
Joined: 31 Mar 12
Posts: 240
Credit: 1,509,172
RAC: 94
Australia
Message 1859578 - Posted: 5 Apr 2017, 13:06:28 UTC - in response to Message 1859553.  

Have you whitelisted boinc and seti in Windows defender(assuming you haven't installed an antivirus)
ID: 1859578 · Report as offensive     Reply Quote
Kets_One

Send message
Joined: 5 Jun 03
Posts: 9
Credit: 972,549
RAC: 3
Netherlands
Message 1859586 - Posted: 5 Apr 2017, 13:54:18 UTC
Last modified: 5 Apr 2017, 14:19:11 UTC

Hey Guys,

Ive been running the optimized SETI 8.22 app from Mike's World website for a week now (without any other changes) and i dont see much improvement compared to the standards SETI apps.
Seeing the posts below on cmdline arguments, im wondering if some tweaking of app_config.xml would benefit the execution speed. BTW: currently i dont have a app_config.xml in my BOINC/SETI folder, so i have to start from scratch.

SETI WIKI provides explanation on different arguments in app_config.xml, but does not specify default values. Ive tried searching the forum, but there appears not to be a clearcut guide on what settings are good for my systems. Any guidance is appreciated!

My current configs:
Home:
Intel i7-4500U (SSE, AVX) (i dont use the CPU for SETI)
Nvidia GT 750M (i use 100% for SETI)
16GB system ram

Work:
Intel i3-3220 (SSE, AVX) (i use 50% of threads for SETI)
Nvidia Quadro K600 (i use 100% for SETI)
32 GB system ram

Thanks,
Kets
ID: 1859586 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 . . . 7 · Next

Message boards : Number crunching : SETI@home v8.22 Windows GPU applications support thread


 
©2017 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.