SETI@home v8.19 Windows GPU applications support thread

Message boards : Number crunching : SETI@home v8.19 Windows GPU applications support thread
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 9 · Next

AuthorMessage
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1823663 - Posted: 12 Oct 2016, 5:51:42 UTC

Please post issues with v8.19 Windows GPU SETI MB apps (if any) here.
Single thread for all flavours: ATi, NV, iGPU.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1823663 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1823716 - Posted: 12 Oct 2016, 13:16:00 UTC - in response to Message 1823663.  

Please post issues with v8.19 Windows GPU SETI MB apps (if any) here.
Single thread for all flavours: ATi, NV, iGPU.



. . Hi Raistmer,

. . As a bit of an amateur I am waiting for Lunatics Beta5 so I can get right into it. Since 4bit WUs offer better resolution (more/better information) has anyone talked about distributing them as well?

. . Thanks for all your hard work.

Stephen

.
ID: 1823716 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1823731 - Posted: 12 Oct 2016, 14:17:10 UTC
Last modified: 12 Oct 2016, 14:18:11 UTC

They run Ok on my Windows 10 PC with opencl_nvidia_SoG but they seem to use a good deal of CPU, up to 23%.Geforce GTX 750 OC.
Tullio
ID: 1823731 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1823748 - Posted: 12 Oct 2016, 15:50:33 UTC - in response to Message 1823716.  
Last modified: 12 Oct 2016, 15:50:43 UTC

It's not subject of this thread. AFAIK Richard already works on it.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1823748 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1823749 - Posted: 12 Oct 2016, 15:51:04 UTC - in response to Message 1823731.  

To decrease CPU usage one can use -use_sleep option.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1823749 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14644
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1823757 - Posted: 12 Oct 2016, 16:25:56 UTC - in response to Message 1823748.  

It's not subject of this thread. AFAIK Richard already works on it.

And has now posted about it in the appropriate (retitled) thread.
ID: 1823757 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1823876 - Posted: 13 Oct 2016, 2:04:38 UTC
Last modified: 13 Oct 2016, 2:23:03 UTC

I am back to v8 8.00, whose CPU usage is 7%.
Tullio
ID: 1823876 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1824145 - Posted: 14 Oct 2016, 5:56:07 UTC - in response to Message 1823876.  

v8 8.00

So you mean CUDA or what?
http://setiathome.berkeley.edu/apps.php
 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1824145 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1824171 - Posted: 14 Oct 2016, 10:24:01 UTC - in response to Message 1824145.  
Last modified: 14 Oct 2016, 10:27:32 UTC

Yes, Cuda 50 on the Windows 10 PC with GTX 750 OC. Instead it is v8 8.10 on the 64-Bit Linux box with HD 7770 and v8 8.05 on the 32-bit Linux laptop with no GPU.
Tullio
ID: 1824171 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1824421 - Posted: 15 Oct 2016, 9:20:35 UTC
Last modified: 15 Oct 2016, 9:40:41 UTC

Yesterday I started up a new machine ("woof-PC") running stock with 2 x GTX 750ti on an AMD FX-8350, 16gb of RAM, on a Gigabyte GA-990FXA-UD3 Rev 4.0 motherboard, 3 WUs/GPU. This morning I noticed that 3 WUs were running 7 or so hours and nowhere near completion (time remaining for one was ~18 hrs, and the other two > 3 days(!!)) and the machine was super-laggy. I aborted the WUs, shut down and restarted BOINC, and not so laggy anymore.

Here are the WUs:

10/15/2016 3:59:11 AM | SETI@home | task 21au09aa.14085.476.13.40.163_1 aborted by user
10/15/2016 3:59:44 AM | SETI@home | task 19ja09ad.31343.1708.15.42.156_1 aborted by user
10/15/2016 4:05:06 AM | SETI@home | task 19ja09ad.31343.1708.15.42.226_1 aborted by user

My command line is:

-sbs 512 - period_iterations_num 10 -hp


EDIT: looking at my woof-PC error WUs, I see they all ran 25,000+ secs, with maybe 17-20,000 secs CPU, essentially blocking 1/2 of my threads for doing work.

EDIT2: Also, looking in BoincTasks History, I see a few WUs that completed ok, but took 3-5 hours to run, where most run on the order of an hour or less.
What's up with that?

FWIW: no such problems on my 3 Intel machines.
ID: 1824421 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1824422 - Posted: 15 Oct 2016, 9:52:37 UTC - in response to Message 1824421.  
Last modified: 15 Oct 2016, 9:53:43 UTC

Check if any driver restart messages in system event log.
if yes:
return period_iterations_num to 50
decrease tasks per GPU to 2.

if not:
add -cpu_lock.

Or just revert to complete default (to be more on this thread subject) and report back if default settings require correction.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1824422 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34249
Credit: 79,922,639
RAC: 80
Germany
Message 1824433 - Posted: 15 Oct 2016, 12:55:35 UTC - in response to Message 1824422.  
Last modified: 15 Oct 2016, 13:02:03 UTC

Check if any driver restart messages in system event log.
if yes:
return period_iterations_num to 50
decrease tasks per GPU to 2.

if not:
add -cpu_lock.

Or just revert to complete default (to be more on this thread subject) and report back if default settings require correction.


On FX 8350 -cpu_lock is necessity running multiple instances.
Also i suggest to run only 2 instances on the 750TI.
With 2 GPU`s running 3 instances each 6 CPU cores are in use but only 4 FPU`available.
So the app will go into stall if no FPU ressources are available.


With each crime and every kindness we birth our future.
ID: 1824433 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1824446 - Posted: 15 Oct 2016, 14:45:48 UTC

I added -cpu_lock and set to 2 WUs/GPU, and that seems to have cured my problems (for now!); I should have remembered about the CPU having only 4 FPUs, but didn't.

Now WUs are well under an hour each (rather than 1+hours). We'll see if this holds up or not.

Thanks for the help!
ID: 1824446 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1824506 - Posted: 15 Oct 2016, 18:26:02 UTC
Last modified: 15 Oct 2016, 18:33:21 UTC

I've read the current documentation and gone back through previous and can't find the answer to what I want to do. In the CUDA app, through MBCUDA.CFG you could set the app priority using -bel -nor -abo and -hp. In the OpenCL app all I've been able to do is set -hp (high_priority) for the app. That results in too much system lag and responsiveness with my aggressive settings. I want to run in above_normal (-abo) for the r3528 app. I tried the -abo tag in the r3528 819 section <cmdline></cmdline> and my stderr.txt still reports tasks running in below normal priority. I also tried to set above normal priority in cc_config.xml using the <process_priority_special>3</process_priority_special> file option. It too did not have any effect on changing the app priority level to above normal while running nor did it change the reported level in stderr.txt.

Is it possible to change the app priority level via any of BOINCs' or SETIs' file configurations?

For the while I have managed to set the app running priority to above normal via a ProcessLasso rule. The app does run in the above normal priority but the app is unaware of its raised priority level because it still writes stderr.txt output of running in below normal priority.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1824506 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1824528 - Posted: 15 Oct 2016, 20:01:39 UTC - in response to Message 1824506.  

OpenCL builds have only 3 priority levels (third is real-time one).
ProcessLasso can be used (or similar app) to change to any desired level.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1824528 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1824534 - Posted: 15 Oct 2016, 20:18:31 UTC - in response to Message 1824528.  

OpenCL builds have only 3 priority levels (third is real-time one).
ProcessLasso can be used (or similar app) to change to any desired level.


https://bitsum.com/

Keith ,I run Process Lasso.

I also run these just to see how the system runs

BOINCtasks, and SIVX64
ID: 1824534 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1824535 - Posted: 15 Oct 2016, 20:22:20 UTC - in response to Message 1824421.  

Yesterday I started up a new machine ("woof-PC") running stock with 2 x GTX 750ti on an AMD FX-8350, 16gb of RAM, on a Gigabyte GA-990FXA-UD3 Rev 4.0 motherboard, 3 WUs/GPU. This morning I noticed that 3 WUs were running 7 or so hours and nowhere near completion (time remaining for one was ~18 hrs, and the other two > 3 days(!!)) and the machine was super-laggy. I aborted the WUs, shut down and restarted BOINC, and not so laggy anymore.

Here are the WUs:

10/15/2016 3:59:11 AM | SETI@home | task 21au09aa.14085.476.13.40.163_1 aborted by user
10/15/2016 3:59:44 AM | SETI@home | task 19ja09ad.31343.1708.15.42.156_1 aborted by user
10/15/2016 4:05:06 AM | SETI@home | task 19ja09ad.31343.1708.15.42.226_1 aborted by user

My command line is:

-sbs 512 - period_iterations_num 10 -hp


EDIT: looking at my woof-PC error WUs, I see they all ran 25,000+ secs, with maybe 17-20,000 secs CPU, essentially blocking 1/2 of my threads for doing work.

EDIT2: Also, looking in BoincTasks History, I see a few WUs that completed ok, but took 3-5 hours to run, where most run on the order of an hour or less.
What's up with that?

FWIW: no such problems on my 3 Intel machines.


Cruncher, that commandline was for the 980s..It has a lot more memory than the 750Ti does

-sbs is a function of how much memory the GPU has, not CPU memory. So for a 980, 980Ti, 10x0s, you can increase that number up to a max of 1024 but with lower RAM GPUs that value has to be lower.

As Raistmer said default for -period_iterations_num is 50 but with high end GPU you can decrease it down to as low as 1 (which I do for 980Ti and 1080, 1070s) But for a 750ti that is too low and best to keep it at default of 50.

Sorry, out of town so can't touch base much this weekend.

Zalster
ID: 1824535 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1824539 - Posted: 15 Oct 2016, 20:50:39 UTC

@Zalster: Thanks for the clarification(s). Hope you are having fun on your trip!

Is there a writeup for command line stuff? If there isn't, there really should be, comparable to the ones for the config xml files. Since the parameters are there for general use now, us SETIzens should have one. (If I've missed it, please let me know but flame me gently, please!).
ID: 1824539 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34249
Credit: 79,922,639
RAC: 80
Germany
Message 1824541 - Posted: 15 Oct 2016, 20:58:46 UTC - in response to Message 1824539.  

@Zalster: Thanks for the clarification(s). Hope you are having fun on your trip!

Is there a writeup for command line stuff? If there isn't, there really should be, comparable to the ones for the config xml files. Since the parameters are there for general use now, us SETIzens should have one. (If I've missed it, please let me know but flame me gently, please!).


I think i mentioned it once.
There should be a read me file in your project folder.
On the bottom i wrote some examples for the comand line options.


With each crime and every kindness we birth our future.
ID: 1824541 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1824544 - Posted: 15 Oct 2016, 21:04:22 UTC - in response to Message 1824528.  

OpenCL builds have only 3 priority levels (third is real-time one).
ProcessLasso can be used (or similar app) to change to any desired level.

Raistmer, if I understand correctly, then you are saying that there is no built-in method of controlling priority level with BOINCs or SETI's configuration files? The limitation is built-in to the OpenCL platform?

Using ProcessLasso is doing what I want for now. I just wanted to see if I could get the process priority level correctly written to stderr.txt.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1824544 · Report as offensive
1 · 2 · 3 · 4 . . . 9 · Next

Message boards : Number crunching : SETI@home v8.19 Windows GPU applications support thread


 
©2024 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.