i7 Hyperthreading + GPU + Most Efficient Seti Workload

Message boards : Number crunching : i7 Hyperthreading + GPU + Most Efficient Seti Workload
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5

AuthorMessage
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1217511 - Posted: 13 Apr 2012, 3:57:58 UTC - in response to Message 1217508.  

I'm gonna let this I7 go for a day then set up the I3

I'm gonna take that 34 computer dude out with 4 computers.

Or die trying.

Wawasoo.
ID: 1217511 · Report as offensive
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1217514 - Posted: 13 Apr 2012, 4:02:05 UTC - in response to Message 1217511.  
Last modified: 13 Apr 2012, 4:06:31 UTC

Both my I7 & I3 are Cosair liquid cooled...and the I3 sever has a gazzion fanz...in the box.

If you don't have that, don't do this...cause I'm screaming.
ID: 1217514 · Report as offensive
Profile john3760
Avatar

Send message
Joined: 9 Feb 11
Posts: 334
Credit: 3,400,979
RAC: 0
United Kingdom
Message 1217518 - Posted: 13 Apr 2012, 4:08:27 UTC
Last modified: 13 Apr 2012, 4:10:38 UTC

First of all,make a new folder somewhere on your hard drive where you can easily
find it,and call it something like lunatics 0.40 app info files
Now locate your app info file in your BOINC folder.
In windows 7 it will be hidden.
You need to select show hidden files on your computer.
Then it should be something like
My computer/ drive c/ program data/ BOINC / projects/ SETI@HOME/

Scroll down past your work unit files till you come across one called app_info .
Right click on this and copy it.

Leave it alone for a while.

Find your new folder you made earlier and paste the copy of your app_info into it.
Paste it into your new folder 3 more times.They should automatically be called
app_info copy 1 etc.

Right click on app_info copy 1 and choose "open with notebook" or " edit"
Scroll down to the bottom of the page.
Slowly scroll back up until you see <COUNT> 1 <COUNT>
Replace the 1 with 0.5
Slowly scroll up and repeat this procedure every time you see the <COUNT> 1 <COUNT> line.
Now select FILE --- Save as,and select "all" file types.
It should now be saved in your new folder as app_info copy 1 as an XML file (not a text file).
Rename this file app_info X2.

Now go to app_info copy2.
Repeat the procedure above but use 0.3 to replace the 1.
Rename this file app_info X3

Goto copy 3 and repeat the procedure with 0.25 to replace the 1.
Rename this file app_info X4

You now have 4 app_info files in your new folder(out of harms way)
The original file will do 1 WU per GPU.
The X2 ,X3 and X4 files will do 2,3&4 WUs per GPU.

Stop all WUs that you are currently running.

Now copy one of these files ( I'd recommend X2 for the time being) and paste
it into your BOINC folder where you found the original app_info file.
It should appear next to the original file.
Delete the file called app_info ( don't worry, you have a copy of it in your new folder).
Rename the file "app_ info X2"by knocking the X2 bit off the end of it.
Resume your WUs ,and you should be doing 2 at a time.

Remember to leave your original file in your new folder alone,if things go T U you can
always replace it in your original BOINC folder.

john3760

edit Sorry it took me so long to write this ,that you have already done it !!!!!
ID: 1217518 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1217521 - Posted: 13 Apr 2012, 4:11:28 UTC - in response to Message 1217508.  

This thing is going through the roof. It's banging 3 Cudi Fermi Task a Minute.

Honestly....I ain't touched shxt. I just cut the count from one <1> to .33.

One use to to take 14 minutes, installed the sofware down to 8 minutes.....now 3 in one minute.

I'm a banging....bitch.

Wawasoo

Don't get to excited. Right now you are chewing up shorties, or VHAR tasks. VHAR=Very High Angle Range. Which you can see in the task result by looking at the number WU true angle range is :. In the case of that task it is 2.733399. Where a "normal" task would be around 0.42.

This thread has a good explanation of where the angle ranges fit in which group.
The easy thing to remember is VLAR=slower(lower number), VHAR=faster(higher number).


SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1217521 · Report as offensive
Profile perryjay
Volunteer tester
Avatar

Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 20,676,751
RAC: 0
United States
Message 1217658 - Posted: 13 Apr 2012, 15:45:07 UTC - in response to Message 1217514.  

Way to go Wawasoo, glad you got it going. One other thing that should be mentioned is the number of errors you are showing. Don't panic, they are VLARs that were sent to your GPU by mistake as resends. Since Nvidea GPUs don't handle VLARs very well it assigns them a very short due date so they time out quickly and sends them back in to be done over.


PROUD MEMBER OF Team Starfire World BOINC
ID: 1217658 · Report as offensive
Profile john3760
Avatar

Send message
Joined: 9 Feb 11
Posts: 334
Credit: 3,400,979
RAC: 0
United Kingdom
Message 1217786 - Posted: 13 Apr 2012, 20:02:33 UTC
Last modified: 13 Apr 2012, 20:20:29 UTC

The errors on his GT 240 aren't VLAR errors.
They seem to be 295 driver errors,as the GPU cannot be detected ( asleep perhaps? ).
The couple of GPU tasks that have completed have deferred to the CPU,hence the longer
run times.

@ wawasoo. You need to roll back the driver on this card,or disable the computer from
sleeping. 285.66 drivers will work.Also when you sort out the problem,I would only run
1 WU on that card for a while till you get it running smoothly.
At a later date try 2 WUs ,but 3 will probably be a bit too much for that card.

I'm sure somebody else will be able to pin the problem down a bit more precisely than
me and give you better advice,but I would definitely get rid of the 295 driver on that card.

Good luck on the taking down of 34 computer dude ;)

p.s You won't do it erroring out WUs .
p.p.s The -12 errors on your 560,can't be helped,don't worry about those.

john3760

Edit : It looks like he has woke his 240 up. WUs are starting to pend and validate.
ID: 1217786 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1217824 - Posted: 13 Apr 2012, 21:51:35 UTC - in response to Message 1217786.  
Last modified: 13 Apr 2012, 21:53:00 UTC

I was having problems with the newer drivers for my NVIDIA cards and had to roll them back. That worked for all but one system. I then went to the NVIDIA site and found the BETA drivers, 301 series and installed them and they are working great on my systems.

I was having all the noted problems, in various forms, now all is well in GPU land. I do have to say the one system crashed during the install, but when I rebooted it the drivers had been installed.

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1217824 · Report as offensive
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1217866 - Posted: 13 Apr 2012, 23:46:31 UTC - in response to Message 1217786.  

The errors on his GT 240 aren't VLAR errors.
They seem to be 295 driver errors,as the GPU cannot be detected ( asleep perhaps? ).
The couple of GPU tasks that have completed have deferred to the CPU,hence the longer
run times.

@ wawasoo. You need to roll back the driver on this card,or disable the computer from
sleeping. 285.66 drivers will work.Also when you sort out the problem,I would only run
1 WU on that card for a while till you get it running smoothly.
At a later date try 2 WUs ,but 3 will probably be a bit too much for that card.

I'm sure somebody else will be able to pin the problem down a bit more precisely than
me and give you better advice,but I would definitely get rid of the 295 driver on that card.

Good luck on the taking down of 34 computer dude ;)

p.s You won't do it erroring out WUs .
p.p.s The -12 errors on your 560,can't be helped,don't worry about those.

john3760

Edit : It looks like he has woke his 240 up. WUs are starting to pend and validate.


Yea it started throwing off erros on the screen saver I woke up and saw all the cuda apps had errored out. I think I got it now. I call that Pending file under Tasks, My Bank Account for future credit. The GT 240 had been keeping about 10, it's getting closer to 20 now after running correctly for 10 hours.

The I7 has increased it's account from around 300 to 460 after running 20 hours. Gonna be some big time points down the road on this machine.

The little AMD too was erroring out, but I think that one is now on the straight and narrow. That one took on average 19hours to complete a task which one core in my I7 could do in an hour. Got the first task out of that one in 2.5 hours, like 17 hours in processing increase. Fingers crossed that it can handle an AstroPulse on it's slim GPU. And it's got a bank account of 1, which is a record for it.

Now the I3 Server is next, it's bank of pending tasks has hovered at about 150, If I put the juice to that one I sould be well on my way north of 220 by tomorrow nite. But before I do I want to roll back that bad driver. What's the best way to roll it back?


ID: 1217866 · Report as offensive
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1218096 - Posted: 14 Apr 2012, 6:24:34 UTC - in response to Message 1217866.  
Last modified: 14 Apr 2012, 6:26:35 UTC

The I7 Bank Account it approaching 500 now.

Man did I have trouble with the I3 Server. It's a 64bit machine running WHS V1 which is 32bit. Blew up everything when i put the wrong Lunatic software on it.

64bit, Don't work.

But, I'm not switching to WHS 2011. I like my JOBD WHS V1 software, it's Idiot proof Server Software for the Home Computer Idoit.....and I'm living proof.

It's racking up nears 200 in it's Pending Bank account, about a third in 5 hours...even with me putting on the wrong software, it'll be over 220 Pending by the time I wake up. I'm not sure I've got the right CPU version running yet....still may need to do some more research.

That little AMD 1 processor has put up 2 tasks in the last few hours; 8. It use to do 2 every 3 days...so it's working. But if I can get this I7 up near 40,000 a day I'll build another one and lay that 35 computer dude in the Dust.

Any thoughts or Ideas would be appreciated. I've got a few old Cabinets, a brand new gts 450 sitting on the shelf, not fast enough for games...maybe I'll SLI up a beast from Tiger Direct parts. Any thoughts on that too would be appreciated.

RAC Numbers still Climbing....by the hour. This Post 17,234.

Wawasoo.
ID: 1218096 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13715
Credit: 208,696,464
RAC: 304
Australia
Message 1218108 - Posted: 14 Apr 2012, 6:37:43 UTC - in response to Message 1218096.  
Last modified: 14 Apr 2012, 6:40:42 UTC

It's a 64bit machine running WHS V1 which is 32bit. Blew up everything when i put the wrong Lunatic software on it.

64bit, Don't work.

If you're running a 32bit OS, then you can only run 32bit software.
If it's a 64bit OS then you can run 64bit software.

RAC Numbers still Climbing....by the hour. This Post 17,234.

Generally it takes 6-8 weeks for RAC to stabilise when you make major changes- ie a new machine, or add a GPU to a machine.
Going from the stock application to an optimised one it took about 4-6 weeks for the RAC to stabilise.
Of course server downtime & hiccups can easily add several (or more) weeks to that.
Grant
Darwin NT
ID: 1218108 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13715
Credit: 208,696,464
RAC: 304
Australia
Message 1218115 - Posted: 14 Apr 2012, 6:42:21 UTC - in response to Message 1218096.  

maybe I'll SLI up a beast from Tiger Direct parts. Any thoughts on that too would be appreciated.

SLI is of no benefit for crunching. It makes 2 video cards into one for games. For crunching you want multiple GPUs.

Grant
Darwin NT
ID: 1218115 · Report as offensive
Profile john3760
Avatar

Send message
Joined: 9 Feb 11
Posts: 334
Credit: 3,400,979
RAC: 0
United Kingdom
Message 1218118 - Posted: 14 Apr 2012, 6:44:49 UTC - in response to Message 1217866.  
Last modified: 14 Apr 2012, 6:50:06 UTC

Go to the nvidia website ,legacy drivers page. find the driver you need .do a clean install.

Or just don't let your computers sleep.,and switch off all screensavers.



john3760
ID: 1218118 · Report as offensive
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1218125 - Posted: 14 Apr 2012, 6:54:33 UTC - in response to Message 1218115.  

maybe I'll SLI up a beast from Tiger Direct parts. Any thoughts on that too would be appreciated.

SLI is of no benefit for crunching. It makes 2 video cards into one for games. For crunching you want multiple GPUs.



That went right over my head.

Wawasoo
ID: 1218125 · Report as offensive
Wawasoo

Send message
Joined: 3 Apr 99
Posts: 99
Credit: 12,215,833
RAC: 0
United States
Message 1218136 - Posted: 14 Apr 2012, 7:08:12 UTC - in response to Message 1216966.  
Last modified: 14 Apr 2012, 7:16:10 UTC

Hi,

I'm running my six physical core i73930K + one NV560TI-448 with
1) hyperthreading on (12 logical CPUs)
2) use 50% of the processors (6 tasks for 6 physical FPUs, no context-switching)
3) a single NV560TI-448 with 3 work units at a time (count=0.33) (3 logical CPUs serving GPU memory needs + needed CPU logic for GPU tasks)
4) 3 logical CPUs left for OS, GUI, etc.

I'm satisfied with the 24/7 Seti RAC around/above 40000. (HOSTID 5643864)

I've just started with the lunatics APV6 so I can't tell where the RAC is going to level out. I also had a complete reinstall of the OS/Seti a few days ago due to a failed attempt to upgrade to fedora 16.

-- petri 33 --


This is the guy I want to hear from, one I7 doing 41,000 RAC. Whatever he's running I want 2. And I'll take that guy out with 2 machines.

Please give me a step by step.

http://setiathome.berkeley.edu/show_user.php?userid=185350
ID: 1218136 · Report as offensive
Profile john3760
Avatar

Send message
Joined: 9 Feb 11
Posts: 334
Credit: 3,400,979
RAC: 0
United Kingdom
Message 1218157 - Posted: 14 Apr 2012, 8:06:39 UTC - in response to Message 1218125.  
Last modified: 14 Apr 2012, 8:10:45 UTC

If you run more than 1 video card in your computer don't use SLI if you wan't to use
them for crunching SETI.
Keeping the two cards separate allows BOINC to detect 2 GPUs( the processing bits
inside your graphics card).

john3760

p.s @ Wawasoo. You may want to start your own thread within Number Crunching,
as you are at the bottom of someone else's very long thread.I'm sure you will get more help
from very knowledgable people on the SETI message boards if you do so.
ID: 1218157 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5

Message boards : Number crunching : i7 Hyperthreading + GPU + Most Efficient Seti Workload


 
©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.