New AstroPulse applications for GPUs.


log in

Advanced search

Message boards : News : New AstroPulse applications for GPUs.

Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · 9 . . . 10 · Next
Author Message
Profile arkaynProject donor
Volunteer tester
Avatar
Send message
Joined: 14 May 99
Posts: 3748
Credit: 48,777,915
RAC: 1,076
United States
Message 1277057 - Posted: 29 Aug 2012, 12:58:23 UTC - in response to Message 1277011.

Hi

Firstly it is best to post questions like this in the Number Crunching Forum.

I can only see your desktop machine attached to your account, however you should not worry as Astropulse work units should not normally be sent to Nvidia cards, this was a problem at the server end and has now been rectifed.

A GFX470 should not have problems with regular work units, laptops on the other hand do heat up quite easily. I have mine on a cooling pad, works a treat.

Bernie


Actually they have released a stock app for for both ATI/AMD and Nvidia. So yes you will be getting AP work unless you change your preferences.

http://setiathome.berkeley.edu/prefs.php?subset=project
____________

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2894
Credit: 6,640,316
RAC: 7,914
Bulgaria
Message 1277084 - Posted: 29 Aug 2012, 14:02:48 UTC - in response to Message 1276949.
Last modified: 29 Aug 2012, 14:50:28 UTC

Is there any way to throttle the load put on the GPU?

Yes, this program (TThrottle) will let you choose the Max temperature you want:
http://efmer.eu/boinc/
http://www.efmer.eu/boinc/download.html

After installing and running TThrottle type the Max temperature for CPU and GPU and check 'Auto Active'
This will Throttle processes running under BOINC when these Max temperatures are reached and keep them at that level.

Choose/type 'reasonable' Max temperatures - not too high, not too low
(good is 15-20°C under the Max allowed temperatures for CPU and GPU
e.g.:
if CPU is specified for 70°C Max allowed working temperature - type 55°C in the CPU box
if GPU is specified for 90°C Max allowed working temperature - type 70°C in the GPU box
)





____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

CommanderKeen
Send message
Joined: 25 Aug 12
Posts: 2
Credit: 31,508
RAC: 0
Germany
Message 1277288 - Posted: 29 Aug 2012, 21:01:10 UTC - in response to Message 1277084.

Is there any way to throttle the load put on the GPU?

Yes, this program (TThrottle) will let you choose the Max temperature you want:
http://efmer.eu/boinc/
http://www.efmer.eu/boinc/download.html

[...]


Sir,
thank you - this is exactly what I was looking for as its acting on temp and now keeping all cold and silent. ♥♥♥

Cheers!

Profile BilBg
Volunteer tester
Avatar
Send message
Joined: 27 May 07
Posts: 2894
Credit: 6,640,316
RAC: 7,914
Bulgaria
Message 1277559 - Posted: 30 Aug 2012, 14:17:53 UTC - in response to Message 1277288.


:)
If you have money - consider 'thank you' to the author of TThrottle:
Profile: S@NL - Fred - www.efmer.eu/boinc
http://setiathome.berkeley.edu/view_profile.php?userid=8906489


[Donate] at the bottom of page:
http://efmer.eu/boinc/

"Why a donation?
The program is absolutely free, but we like to improve TThrottle and supply you with signed drivers.
If you like our program, help us to continue our work.
If you want to give us a donation the minimum amount is $1.
"


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

RSCBeard
Send message
Joined: 9 Oct 03
Posts: 3
Credit: 617,100
RAC: 1,051
United States
Message 1278354 - Posted: 31 Aug 2012, 17:38:23 UTC - in response to Message 1276629.

I'm using a Toshiba which evidently has issues with drivers. The verification tool will not even load.

Is this your laptop? (Toshiba Tecra® R850-S8530):
http://us.toshiba.com/computers/laptops/tecra/r850/r850-s8530/

If it is - look here (there is 'AMD Display Driver(06-29-2011; 131.17M)' - but probably it is with 'high CPU usage bug' for OpenCL):
http://www.csd.toshiba.com/cgi-bin/tais/support/jsp/modelContent.jsp?ct=DL&os=&category=&moid=2998365&rpn=PT520U&modelFilter=R850-S8530&selCategory=2756709&selFamily=1073768664

Or try to find your laptop here (and select 'Display' for listing the Display Drivers):
http://www.csd.toshiba.com/cgi-bin/tais/support/jsp/home.jsp?nav=Download




This would load a driver of the same verison I'm running. The laptop is relatively new - May of this year and is running Win 7 (64bit).


____________

Marc Elvy (UWEO)
Send message
Joined: 13 Aug 99
Posts: 1
Credit: 22,203,069
RAC: 5,174
United States
Message 1280258 - Posted: 5 Sep 2012, 13:59:49 UTC

Need more AP... need more AP... need more AP! :) I have two ATI 5850's set to Xfire in my i7 950 box. They mow through the AP units in short order. Sometimes the GPU's are fully loaded 80% or better. On some AP units they run at between 12% ~ 20% or so, kind of odd. 90% of the time they are waiting to get their silicone on more units. Still both GPU's get AP WU, the i7 950 never gets AP WU... well done team!
____________

Eric KorpelaProject donor
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar
Send message
Joined: 3 Apr 99
Posts: 1127
Credit: 10,725,651
RAC: 18,560
United States
Message 1281031 - Posted: 7 Sep 2012, 15:17:33 UTC - in response to Message 1280258.

Since there can only be about 3% as many Astropulse workunits as SETI@home workunits, it's had to catch very many. Astropulse also splits really fast compared to SETI@home. Check the server status page to see if any is actively being split. Once data files are listed as "done" it's less likely that you will get any.
____________

JohnDKProject donor
Volunteer tester
Avatar
Send message
Joined: 28 May 00
Posts: 899
Credit: 49,256,139
RAC: 45,649
Denmark
Message 1281062 - Posted: 7 Sep 2012, 16:15:15 UTC

Isn't it possible to cut down on the AP splitters so they don't finish long before the MB's are done?

Profile Sunny129
Avatar
Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1281069 - Posted: 7 Sep 2012, 16:37:22 UTC - in response to Message 1281062.

Isn't it possible to cut down on the AP splitters so they don't finish long before the MB's are done?

what would that solve though? the project server only receives splittable AP data when Arecibo (or wherever it comes from) has data to send. and that data comprises only a minority of all SETI data that must be crunched (which is why there is approx. 1 AP task for every 20 or so Multibeam tasks). if the AP splitters are intentionally slowed down, we'll see AP work less often than we do now...
____________

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8832
Credit: 53,658,006
RAC: 49,082
United Kingdom
Message 1281083 - Posted: 7 Sep 2012, 17:04:10 UTC - in response to Message 1281069.

Isn't it possible to cut down on the AP splitters so they don't finish long before the MB's are done?

what would that solve though? the project server only receives splittable AP data when Arecibo (or wherever it comes from) has data to send. and that data comprises only a minority of all SETI data that must be crunched (which is why there is approx. 1 AP task for every 20 or so Multibeam tasks). if the AP splitters are intentionally slowed down, we'll see AP work less often than we do now...

The potential advantage of slowing down the AP splitters - perhaps by reducing the number of instances running at the same time - would be to even out the traffic flow on the download link. At the moment, it gets grossly oversaturated when AP is being split, and can be underutilised (though less often, nowadays) between AP batches.

Grant (SSSF)
Send message
Joined: 19 Aug 99
Posts: 5955
Credit: 62,515,954
RAC: 40,777
Australia
Message 1281209 - Posted: 7 Sep 2012, 22:56:27 UTC - in response to Message 1281083.

.... and can be underutilised (though less often, nowadays) between AP batches.

I'm pretty sure the only times the network traffic isn't maxed out these days is when they system is down for maintenance, or there's a problem.

____________
Grant
Darwin NT.

Tsukasa
Send message
Joined: 9 Sep 12
Posts: 6
Credit: 1,774,468
RAC: 0
France
Message 1282874 - Posted: 12 Sep 2012, 13:56:18 UTC
Last modified: 12 Sep 2012, 14:15:03 UTC

Hey~

I'm relatively new on BOINC (joined a few days ago), and I have a problem with OpenCL tasks.

I only got 1 OpenCL task (90 tasks in total for this computer) through BOINC. The server doesn't return any other OpenCL task for me, even if I force the refresh. The log event says "Scheduler request completed : got 0 new tasks".

It's frustrating, since my GPU computes 16 times faster than my CPU for MilkyWay@Home (05:47 for GPU versus 01:37:23 for CPU ). I own an AMD Radeon HD 7800 which support OpenCL 1.2, with Catalyst 12.8 (had nothing too for 12.6).

Any idea ?

Windows Version : 7 Professional x64
BOINC Manager version : 7.0.28

EDIT : nvm, just discovered the server status page. So no Astropulse task for the moment. Too bad :(

Pollux_P3D
Volunteer tester
Send message
Joined: 3 Feb 01
Posts: 4
Credit: 4,118,321
RAC: 164
Germany
Message 1282898 - Posted: 12 Sep 2012, 16:04:15 UTC - in response to Message 1282874.
Last modified: 12 Sep 2012, 16:13:58 UTC

Huhu Tsukasa
Astropulse there are not many.
With the Lunatics Installer (Lunatics_Win64_v.040_setup) your Ati can also calculate Multi Beams.

Have a Nice Day
Pollux



1 - Windows Seti@Home apps
http://lunatics.kwsn.net/index.php?module=Downloads;catd=9
____________

Tsukasa
Send message
Joined: 9 Sep 12
Posts: 6
Credit: 1,774,468
RAC: 0
France
Message 1283064 - Posted: 13 Sep 2012, 1:18:31 UTC
Last modified: 13 Sep 2012, 1:21:45 UTC

WTF o_o All the remaining time of my tasks are speeding up. I was at 3+ hours on a task (which is now computed with my GPU thanks to Lunatics), it ended in less than 3 minutes (fastest got computed in 02:10). It computes more than 60 times faster on my GPU, and at least 2 times faster on my CPU (one of the tasks is going at least 10 times faster). Tasks computed by the GPU are computed a lot faster than Astropulse tasks.

Something seems wrong. We are far away from the estimated 8x computing speed acceleration for the GPU. Even the CPU is computing at an insane rate. I guess I'll have to keep an eye on the "Invalid" page.

Anyway, thanks for the download link for this software. I'm impressed.

Tsukasa
Send message
Joined: 9 Sep 12
Posts: 6
Credit: 1,774,468
RAC: 0
France
Message 1283198 - Posted: 13 Sep 2012, 11:28:31 UTC - in response to Message 1283064.

They are all valid. Impressive.

Christoph
Volunteer tester
Send message
Joined: 21 Apr 03
Posts: 76
Credit: 262,351
RAC: 0
Germany
Message 1283341 - Posted: 13 Sep 2012, 20:19:24 UTC - in response to Message 1283198.
Last modified: 13 Sep 2012, 20:19:50 UTC

The Lunatics crew is cooperating very close with the project developers.
All their apps are tested in the SETI@home Beta project first to get the kings out.

You should watch the tempretures. Have a look for TThrottle for that. http://www.efmer.eu/boinc/
____________
Christoph

Tsukasa
Send message
Joined: 9 Sep 12
Posts: 6
Credit: 1,774,468
RAC: 0
France
Message 1283351 - Posted: 13 Sep 2012, 21:15:41 UTC - in response to Message 1283341.

Temperatures are OK
- Max 77°C for CPU cores
- 62°C for GPU

I've seen worse :)

I've set the following rules :
- > 80°C for CPU, throttle
- > 75°C for GPU, throttle

Profile John Neale
Volunteer tester
Avatar
Send message
Joined: 16 Mar 00
Posts: 533
Credit: 3,267,680
RAC: 1,687
South Africa
Message 1283616 - Posted: 14 Sep 2012, 14:26:38 UTC - in response to Message 1283064.

WTF o_o All the remaining time of my tasks are speeding up. I was at 3+ hours on a task (which is now computed with my GPU thanks to Lunatics), it ended in less than 3 minutes (fastest got computed in 02:10). It computes more than 60 times faster on my GPU, and at least 2 times faster on my CPU (one of the tasks is going at least 10 times faster). Tasks computed by the GPU are computed a lot faster than Astropulse tasks.

Something seems wrong. We are far away from the estimated 8x computing speed acceleration for the GPU. Even the CPU is computing at an insane rate. I guess I'll have to keep an eye on the "Invalid" page.

Anyway, thanks for the download link for this software. I'm impressed.


For this type of analysis, you need to compare apples with apples. Not all tasks have the same runtimes. Runtimes vary according to the angle range of the telescope at the time of recording. In addition, some tasks might exit early if they're very noisy - usually as a result of terrestrial interference.
____________

Tsukasa
Send message
Joined: 9 Sep 12
Posts: 6
Credit: 1,774,468
RAC: 0
France
Message 1283691 - Posted: 14 Sep 2012, 18:14:49 UTC - in response to Message 1283616.
Last modified: 14 Sep 2012, 18:16:07 UTC

WTF o_o All the remaining time of my tasks are speeding up. I was at 3+ hours on a task (which is now computed with my GPU thanks to Lunatics), it ended in less than 3 minutes (fastest got computed in 02:10). It computes more than 60 times faster on my GPU, and at least 2 times faster on my CPU (one of the tasks is going at least 10 times faster). Tasks computed by the GPU are computed a lot faster than Astropulse tasks.

Something seems wrong. We are far away from the estimated 8x computing speed acceleration for the GPU. Even the CPU is computing at an insane rate. I guess I'll have to keep an eye on the "Invalid" page.

Anyway, thanks for the download link for this software. I'm impressed.


For this type of analysis, you need to compare apples with apples. Not all tasks have the same runtimes. Runtimes vary according to the angle range of the telescope at the time of recording. In addition, some tasks might exit early if they're very noisy - usually as a result of terrestrial interference.

I agree that runtimes vary.

But I was talking about before and after Lunatics installation on my ATI computer. I had some tasks which just started when I installed it, and the remaining time changed. It went down. A lot. Tasks are computing faster. Even when computed by CPU, their computation time are 100% faster at least. And with GPU, it computes sometimes more than 6000% faster. That's why I'm impressed.

Mike
Send message
Joined: 26 Jan 06
Posts: 3
Credit: 1,089,961
RAC: 171
Message 1289813 - Posted: 30 Sep 2012, 19:33:59 UTC

I have been having problems lately with GPU processing. AstroPulse v6 6.04 (ATI_opencl_100) has been causing error messages about my AMD video driver not responding and a restart being successful. It does this about every few minutes (all input locks up and the screen goes blank then slowly comes back. I am running 3 monitors but not displaying the SETI processing graph on any screen. Usually it only involves 2 monitors (connected to the graphics card - the other is connected via USB). Also the elapsed time resets to zero during these episodes. The other thing is that I do not have my local preferences set to use the GPU and Boinc is running one more task than my local preferences allow. It only does this when the above mentioned astropulse program is available. I will revisit my web specified options to see if I can turn off using the GPU. My video card is an AMD Radeon HD 6570. At first I thought I was having either video hardware or software problems but quickly narrowed it down to astropulse being the culprit.
____________

Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · 9 . . . 10 · Next

Message boards : News : New AstroPulse applications for GPUs.

Copyright © 2014 University of California