Back after a long time

Message boards : Number crunching : Back after a long time
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335198 - Posted: 6 Feb 2013, 18:58:06 UTC

I just started up Boinc again , installed the new boinc version , have gotten about 20 cuda WU's , have a quad core amd cpu and a evga 460se that i used to run 2 GPUWu's on at a time as well as the 4 CPU Wu's , for some reason It is only running .04 cpu and 1 Gpu Wu at a time . Is that normal with this new version of Boinc ?

Thanks Floyd
ID: 1335198 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1335208 - Posted: 6 Feb 2013, 19:14:15 UTC - in response to Message 1335198.  

Depends on how you used to run 2 tasks on the GPU. Did you use an app_info.xml file in the past? Are you now? Did you set it up correctly to run 2 tasks at once?

But best ask that in the Number Crunching forum, as the people posting there won't be seen here. (Ask a moderator to move your thread, by clicking the red-x on your post).
ID: 1335208 · Report as offensive
David S
Volunteer tester
Avatar

Send message
Joined: 4 Oct 99
Posts: 18352
Credit: 27,761,924
RAC: 12
United States
Message 1335224 - Posted: 6 Feb 2013, 20:16:25 UTC

Okay, so now that you're in Number Crunching, here's part of an answer for you.

When a task runs on a GPU, it also needs a little bit of CPU time to "feed" the GPU. I'm not even sure what that means, but yes, it's normal and 0.04% is the number Seti always gives. It might not even be that much in reality.

Are you also saying it's only running one at a time when you thought it should be doing two? If so, there's a setting you need to change in one of your Boinc setup files, but I'll leave it to someone else to give you the details.

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.

ID: 1335224 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1335233 - Posted: 6 Feb 2013, 20:32:08 UTC

Hi Floye, welcome back - I missed you!

Looks like you did get some Scheduler contacts and got 98 gpu tasks. Your App details shows that you've installed the optimized apps, so you can run multiple instances by changing the "count" value to .5 (for 2) or .33 (for 3) in the 6.10 section of your app_info.xml file.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1335233 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335241 - Posted: 6 Feb 2013, 20:48:25 UTC - in response to Message 1335233.  

Thanks Fred , Glad to be back !
I had forgotten what to change and what the numbers should be , Got it doing 2 GPU WU's at a time now , Are we just getting Cuda Wu's or is there any Just CPU work out there ?
Got a quad core that is feeling left out for some reason... LOL
ID: 1335241 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1335246 - Posted: 6 Feb 2013, 20:57:12 UTC - in response to Message 1335241.  
Last modified: 6 Feb 2013, 20:59:20 UTC

You might want to update your Cuda app, x32f is really old and hardly any better than Stock, For a GTX460SE, try the x41zc_Cuda42 app, it'll be fastest:

http://jgopt.org

setiathome_CUDA: Found 1 CUDA device(s):
  Device 1: GeForce GTX 460 SE, 1024 MiB, regsPerBlock 32768
     computeCap 2.1, multiProcs 6 
     clockRate = 1440000 
setiathome_CUDA: CUDA Device 1 specified, checking...
   Device 1: GeForce GTX 460 SE is okay
SETI@home using CUDA accelerated device GeForce GTX 460 SE
Priority of process raised successfully
Priority of worker thread raised successfully
size 8 fft, is a freaky powerspectrum
size 16 fft, is a cufft plan
size 32 fft, is a cufft plan
size 64 fft, is a cufft plan
size 128 fft, is a cufft plan
size 256 fft, is a freaky powerspectrum
size 512 fft, is a freaky powerspectrum
size 1024 fft, is a freaky powerspectrum
size 2048 fft, is a cufft plan
size 4096 fft, is a cufft plan
size 8192 fft, is a cufft plan
size 16384 fft, is a cufft plan
size 32768 fft, is a cufft plan
size 65536 fft, is a cufft plan
size 131072 fft, is a cufft plan

 )       _   _  _)_ o  _  _ 
(__ (_( ) ) (_( (_  ( (_ (  
 not bad for a human...  _) 

Multibeam x32f Preview, Cuda 3.0

Work Unit Info:
...............
WU true angle range is :  0.424302


Claggy
ID: 1335246 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1335247 - Posted: 6 Feb 2013, 21:01:24 UTC

We're operating under "temporary" limits of 100 gpu and 100 cpu tasks per host. Scheduler is hard to reach, but when you are lucky, it will fill gpu first up to the limit, and then send cpu. The tasks are the same data, just wrapped as needed when the order is filled. If you want, you can turn off gpu fetch in your Seti@Home preferences - just untick "use Nvidia gpu". That will be passed back to BOINC on the next successful update, and BOINC will stop asking for gpu work. That's a little risky now with scheduler contacts so hard to get. Since you're near the gpu limit, I'd just wait.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1335247 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335255 - Posted: 6 Feb 2013, 21:21:59 UTC - in response to Message 1335246.  

Ok , I downloaded that file , got winzip to unzip it , do I just run the application file in it ?
Lunatics_x41zc_win32_cuda42
or do I need to manually insert something ?

You might want to update your Cuda app, x32f is really old and hardly any better than Stock, For a GTX460SE, try the x41zc_Cuda42 app, it'll be fastest:

http://jgopt.org

setiathome_CUDA: Found 1 CUDA device(s):
  Device 1: GeForce GTX 460 SE, 1024 MiB, regsPerBlock 32768
     computeCap 2.1, multiProcs 6 
     clockRate = 1440000 
setiathome_CUDA: CUDA Device 1 specified, checking...
   Device 1: GeForce GTX 460 SE is okay
SETI@home using CUDA accelerated device GeForce GTX 460 SE
Priority of process raised successfully
Priority of worker thread raised successfully
size 8 fft, is a freaky powerspectrum
size 16 fft, is a cufft plan
size 32 fft, is a cufft plan
size 64 fft, is a cufft plan
size 128 fft, is a cufft plan
size 256 fft, is a freaky powerspectrum
size 512 fft, is a freaky powerspectrum
size 1024 fft, is a freaky powerspectrum
size 2048 fft, is a cufft plan
size 4096 fft, is a cufft plan
size 8192 fft, is a cufft plan
size 16384 fft, is a cufft plan
size 32768 fft, is a cufft plan
size 65536 fft, is a cufft plan
size 131072 fft, is a cufft plan

 )       _   _  _)_ o  _  _ 
(__ (_( ) ) (_( (_  ( (_ (  
 not bad for a human...  _) 

Multibeam x32f Preview, Cuda 3.0

Work Unit Info:
...............
WU true angle range is :  0.424302


Claggy


ID: 1335255 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1335262 - Posted: 6 Feb 2013, 21:36:20 UTC - in response to Message 1335255.  
Last modified: 6 Feb 2013, 21:38:13 UTC

Ok , I downloaded that file , got winzip to unzip it , do I just run the application file in it ?
Lunatics_x41zc_win32_cuda42
or do I need to manually insert something ?


Copy all the files to the project directory, (I'm assuming you haven't got a recent Lunatics Installer) Open the MBCuda.aistub with Notepad, copy all the app_info info that lies between the <app_info> opening tag and the </app_info> closing tag,
Open your app_info with Notepad, then paste the info you copied from the MBCuda.aistub into your app_info making sure you overwrite the existing info for the Cuda app, Or PM me your app_info and i'll do it for you.

The other way of doing it (if you're installed the apps with a recent Lunatics Installer) is to copy the files to the project directory, then run aimerge.cmd, it'll merge all your .aistub's into one app_info.xml

Claggy
ID: 1335262 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335266 - Posted: 6 Feb 2013, 21:52:34 UTC - in response to Message 1335262.  
Last modified: 6 Feb 2013, 21:59:31 UTC

Edit--- I got a bunch of errors and had to restore the appinfo file back...
========================================================================
Ok , I think I did it right , Only problem is that just before I did That I was doing about 6 wu's at a time and now I'm ( maybe temporary ) only doing 2 gpu Wu's at a time again...


Copy all the files to the project directory, (I'm assuming you haven't got a recent Lunatics Installer) Open the MBCuda.aistub with Notepad, copy all the app_info info that lies between the <app_info> opening tag and the </app_info> closing tag,
Open your app_info with Notepad, then paste the info you copied from the MBCuda.aistub into your app_info making sure you overwrite the existing info for the Cuda app, Or PM me your app_info and i'll do it for you.

The other way of doing it (if you're installed the apps with a recent Lunatics Installer) is to copy the files to the project directory, then run aimerge.cmd, it'll merge all your .aistub's into one app_info.xml

Claggy[/quote]
ID: 1335266 · Report as offensive
Profile Fred E.
Volunteer tester

Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,140,697
RAC: 0
United States
Message 1335273 - Posted: 6 Feb 2013, 22:28:58 UTC

Edit--- I got a bunch of errors and had to restore the appinfo file back...
========================================================================
Ok , I think I did it right , Only problem is that just before I did That I was doing about 6 wu's at a time and now I'm ( maybe temporary ) only doing 2 gpu Wu's at a time again...

The errors shown on your task page are VLAR expirations - that's a no-fault error. You had some "lost" tasks from a lost scheduler response that BOINC didn't receive. When the next request was worked, scheduler filled the gpu request first from your lost tasks, but won't send VLAR's to a Nvidia gpu. Instead, they are timed out on the server.

By your comment of running only 2, do you currently have any cpu tasks on hand? The website Tasks page shows cpu tasks assigned, but they may be lost tasks (ghosts). If you have cpu tasks on hand but it won't crunch them, you may have inadvertently messed up the 6.03 section when you edited app_info.xml.

Might want to take up Claggy's offer for help on the app_info.xml file.
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.
ID: 1335273 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335276 - Posted: 6 Feb 2013, 22:39:55 UTC - in response to Message 1335273.  

I did send him the appinfo file by IM , i restored the orig app info file and have got some cpu tasks going now for a total of 6 wu's combined at a time.
from the event log it was showing a bunch of intel tasks or something erroring out. I'll try that new stuff when Claggy gets back to me...

Edit--- I got a bunch of errors and had to restore the appinfo file back...
========================================================================
Ok , I think I did it right , Only problem is that just before I did That I was doing about 6 wu's at a time and now I'm ( maybe temporary ) only doing 2 gpu Wu's at a time again...

The errors shown on your task page are VLAR expirations - that's a no-fault error. You had some "lost" tasks from a lost scheduler response that BOINC didn't receive. When the next request was worked, scheduler filled the gpu request first from your lost tasks, but won't send VLAR's to a Nvidia gpu. Instead, they are timed out on the server.

By your comment of running only 2, do you currently have any cpu tasks on hand? The website Tasks page shows cpu tasks assigned, but they may be lost tasks (ghosts). If you have cpu tasks on hand but it won't crunch them, you may have inadvertently messed up the 6.03 section when you edited app_info.xml.

Might want to take up Claggy's offer for help on the app_info.xml file.


ID: 1335276 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1335280 - Posted: 6 Feb 2013, 22:59:16 UTC - in response to Message 1335276.  

I did send him the appinfo file by IM , i restored the orig app info file and have got some cpu tasks going now for a total of 6 wu's combined at a time.
from the event log it was showing a bunch of intel tasks or something erroring out. I'll try that new stuff when Claggy gets back to me...

Edit--- I got a bunch of errors and had to restore the appinfo file back...
========================================================================
Ok , I think I did it right , Only problem is that just before I did That I was doing about 6 wu's at a time and now I'm ( maybe temporary ) only doing 2 gpu Wu's at a time again...

The errors shown on your task page are VLAR expirations - that's a no-fault error. You had some "lost" tasks from a lost scheduler response that BOINC didn't receive. When the next request was worked, scheduler filled the gpu request first from your lost tasks, but won't send VLAR's to a Nvidia gpu. Instead, they are timed out on the server.

By your comment of running only 2, do you currently have any cpu tasks on hand? The website Tasks page shows cpu tasks assigned, but they may be lost tasks (ghosts). If you have cpu tasks on hand but it won't crunch them, you may have inadvertently messed up the 6.03 section when you edited app_info.xml.

Might want to take up Claggy's offer for help on the app_info.xml file.


Those errors that you have if you look are listed in red text and are not your fault at all but an inhouse setup that times out VLAR from being sent to your GPU.

Now if those errors are list in black text then the problem is at your end.

Cheers.
ID: 1335280 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1335281 - Posted: 6 Feb 2013, 22:59:25 UTC - in response to Message 1335276.  

I did send him the appinfo file by IM , i restored the orig app info file and have got some cpu tasks going now for a total of 6 wu's combined at a time.
from the event log it was showing a bunch of intel tasks or something erroring out. I'll try that new stuff when Claggy gets back to me...

Edit--- I got a bunch of errors and had to restore the appinfo file back...
========================================================================
Ok , I think I did it right , Only problem is that just before I did That I was doing about 6 wu's at a time and now I'm ( maybe temporary ) only doing 2 gpu Wu's at a time again...

The errors shown on your task page are VLAR expirations - that's a no-fault error. You had some "lost" tasks from a lost scheduler response that BOINC didn't receive. When the next request was worked, scheduler filled the gpu request first from your lost tasks, but won't send VLAR's to a Nvidia gpu. Instead, they are timed out on the server.

By your comment of running only 2, do you currently have any cpu tasks on hand? The website Tasks page shows cpu tasks assigned, but they may be lost tasks (ghosts). If you have cpu tasks on hand but it won't crunch them, you may have inadvertently messed up the 6.03 section when you edited app_info.xml.

Might want to take up Claggy's offer for help on the app_info.xml file.


I haven't had any PM's from you.

Claggy
ID: 1335281 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335294 - Posted: 6 Feb 2013, 23:36:31 UTC - in response to Message 1335281.  

Ok , I clicked on your name on left side of post where it says send message and resent the file... hope it goes thru this time.
thanks

ID: 1335294 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1335296 - Posted: 6 Feb 2013, 23:46:00 UTC - in response to Message 1335294.  

Ok , I clicked on your name on left side of post where it says send message and resent the file... hope it goes thru this time.
thanks

PM sent,

Claggy
ID: 1335296 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335308 - Posted: 7 Feb 2013, 0:22:06 UTC - in response to Message 1335296.  

Hey Claggy , I replaced the app info (after saving a copy of the old file) , and was going to change the six 1's to all .5 's there seems to be a lot of intel (platform>windows_intelx86</platform>) in that appinfo file...

app_name>setiathome_enhanced</app_name>
<version_num>609</version_num>
<platform>windows_intelx86</platform>
<plan_class>cuda23</plan_class>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.040000</max_ncpus>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>

Maybe that was why I was seeing a lot of errors that mentioned intel in the event log earlier... all the places that i thought I was supposed to change the 1 to .5 to run 2 WU's at a time on my card seems to say intel... LOL , Just a thought...

Ok , I clicked on your name on left side of post where it says send message and resent the file... hope it goes thru this time.
thanks

PM sent,

Claggy


ID: 1335308 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1335313 - Posted: 7 Feb 2013, 0:37:09 UTC - in response to Message 1335308.  

The platform entries aren't anything to worry about, they are there just far backward compatibility across 32bit and 64bit OS's/Boinc versions, so any work marked up as windows_intelx86 or windows_x86_64 platforms are still done and don't get lost and have to be resent.

Claggy
ID: 1335313 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1335969 - Posted: 9 Feb 2013, 0:46:32 UTC - in response to Message 1335313.  

The platform entries aren't anything to worry about, they are there just far backward compatibility across 32bit and 64bit OS's/Boinc versions, so any work marked up as windows_intelx86 or windows_x86_64 platforms are still done and don't get lost and have to be resent.

Claggy


Hey Claggy

This seems to be working ok , I just have 2 (Completed, validation inconclusive)
since the server started sending the cuda_fermi WU's , Maybe that is a normal occurence with this type of WU ?

No Invalid's so far , just the 2 Completed, validation inconclusive WU's.
GPU is running about 10 C warmer , Increased speed of GPU fan to compensate , It is a pretty steady 54/55 C now.
Thanks Claggy for your help with this...
ID: 1335969 · Report as offensive
Profile Floyd
Avatar

Send message
Joined: 19 May 11
Posts: 524
Credit: 1,870,625
RAC: 0
United States
Message 1336041 - Posted: 9 Feb 2013, 5:33:07 UTC - in response to Message 1335969.  

EDIT - The two inconclusive WU's is Now 5 of them...
I really hope this is a common thing , instead of a glitch in my program...


Hey Claggy

This seems to be working ok , I just have 2 (Completed, validation inconclusive)
since the server started sending the cuda_fermi WU's , Maybe that is a normal occurence with this type of WU ?

No Invalid's so far , just the 2 Completed, validation inconclusive WU's.
GPU is running about 10 C warmer , Increased speed of GPU fan to compensate , It is a pretty steady 54/55 C now.
Thanks Claggy for your help with this...[/quote]

ID: 1336041 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Back after a long time


 
©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.