CUDA MB V12b rebuild supposed to work with Fermi GPUs

Message boards : Number crunching : CUDA MB V12b rebuild supposed to work with Fermi GPUs
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · Next

AuthorMessage
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999011 - Posted: 26 May 2010, 23:27:46 UTC - in response to Message 999009.  

Hi Steve
Your welcome, temps seem fine at 70 although i am seriously considering watercooling once its up and running. 10000 in 1 week phew you ought to be proud of that!
I dare say i will get there i am determined its just a matter of time.
Have a good one Steve

ID: 999011 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6662
Credit: 121,090,076
RAC: 0
United States
Message 999014 - Posted: 26 May 2010, 23:46:59 UTC - in response to Message 999011.  

Hi Steve
Your welcome, temps seem fine at 70 although i am seriously considering watercooling once its up and running. 10000 in 1 week phew you ought to be proud of that!
I dare say i will get there i am determined its just a matter of time.
Have a good one Steve



You are right, I had to practice a lot of patience, and it payed off. I know you will get your running. It is definitly possible. The others will most likely be asking you different things. Over and over the people on these boards have gone way beyond to help someone out. I started mine off with BOINC 6.10.43, and last night upgraded to 6.10.56, and everything is very smooth. You'll get there.

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 999014 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 999032 - Posted: 27 May 2010, 1:43:13 UTC

Good day Gents,
While it is true that there is no optimized application as of this moment a card that is producing invalid results will not have a chance to heat up to the temps that might cause errors. I personally would recommend watercooling without a doubt because of Steve's findings. Mine running full bore hit around 72C in an AC controled datacenter (it is 63F in there are all times - summer is going to be nice)

I'm writing this on my phone so it is a bit of a challenge to view the previous comments and to view the person that was having the problems. Maybe he could PM me with the quoted App_Info.xml file so I can take a look what might be going on in there.

My experience with the new with new client has been very good and it didn't require much modification to get operational for my particular needs.

Have the other components of the system been verified to be working properly? You had said with my previous files that it was working but things can go haywire. I would look at your tasks but that is a bit difficult right now.

If you do PM them to me when I get back home later tonight I can review them and possibly resolve the issue. I do have a very early morning tomorrow (5am leaving for Minneapolis) so it might be a very quick response. Vyper being in a different part of the world might read this post too.

One way or the other we will get it up and running.

Todd
ID: 999032 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 999061 - Posted: 27 May 2010, 3:52:07 UTC - in response to Message 999032.  
Last modified: 27 May 2010, 4:04:46 UTC

Good day Gents,
While it is true that there is no optimized application as of this moment ...


Well not completely true anymore. Just to keep you a little updated: We of course have been poking around in the background to see what these new chips can do, and have started to make some good progress considering that it's extra challenging passing builds among those of us experienced enough for detailed pre-alpha tests with the relevant hardware & those making development optimisation decisions. While, as far as I'm concerned, there is a long way to go before something publicly releasable, our builds are already somewhat faster than stock beta apps at this time.

Jason
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 999061 · Report as offensive
Profile MadMaC
Volunteer tester
Avatar

Send message
Joined: 4 Apr 01
Posts: 201
Credit: 47,158,217
RAC: 0
United Kingdom
Message 999066 - Posted: 27 May 2010, 5:33:36 UTC

Toxic, just to check, have you tried running your 480 gpu fans at 100% for a while, just in case it is a temperature issue?
Also, try running rescheduler in TEST mode only, just to see if you have any VLAR or VHAR's slowing things down.
Does anyone know if rescheduler works with the 6.10 app?

Is it essential to watercool these cards? If it is, then that is my mind made up and I will be buying 295's as I dont want to go down the watercooling route yet.
Anyone running these on air cooling and having no issues..
ID: 999066 · Report as offensive
Profile Todd Hebert
Volunteer tester
Avatar

Send message
Joined: 16 Jun 00
Posts: 648
Credit: 228,292,957
RAC: 0
United States
Message 999069 - Posted: 27 May 2010, 5:48:49 UTC

My three GTX-480's are running fine on air cooling with 100% fan speed. Yes they make a ton of noise though, but no issues with stability - I also have them overclocked slightly. The temps range from 68C - 72C depending on which card you are looking at. Obviously the card in the center is a bit warmer because of placement. I also have a significant amount of air flow in my CoolerMaster Stacker 830 - but I'm sure that another high end case would provide enough. Something like the HAF932 or Corsair Obsidian 800D would do nicely.

The rescheduler app works fine with the latest client - although sometimes when you do run the transformation it does disconnect the client from the service - but this happened before.

Jason - glad to hear that futher development is progressing along - if you need a tester just let me know :)
Todd
Team GPU Users
ID: 999069 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999106 - Posted: 27 May 2010, 11:53:07 UTC - in response to Message 999066.  

Hi MadMac
I am going to step away for a few hours, recharge my batteries and come back to it fresh and start again.
If you all dont mind before i start i will post exactly what i am using version numbers and a short step by step snapshot along with the app_info when i am done to see if anyone can tell me what i am doing wrong.
Thanks all
ID: 999106 · Report as offensive
Profile MadMaC
Volunteer tester
Avatar

Send message
Joined: 4 Apr 01
Posts: 201
Credit: 47,158,217
RAC: 0
United Kingdom
Message 999118 - Posted: 27 May 2010, 13:04:46 UTC - in response to Message 999106.  

One possible is to completely uninstall boinc, remove the 2 program directories from <Program Files> and <ProgramData> and then re=install using stock apps only and see how you get on??

When having trouble, strip everything back to basics to eliminate potential problems
ID: 999118 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999142 - Posted: 27 May 2010, 16:04:58 UTC - in response to Message 999118.  

Ok round 14 day 5 ;-)
i7 930
w7 64 bit
gtx 480
Boinc 6.10.56
Lunatics optimized 0.36
Todds files
Ran cpu-z processor shows MMX.SSE (1,2,3,3S,4.1,4.2) When it comes to installing opt .36 which multibeam cpu edition is best to use? Have unchecked astropulse to try and make it simpler when it comes to edit app_info. On page 2 of .36 install which cuda mb v and when it comes to cuda library 2.2 or 2.3?
Once i have installed Boinc and .36 drop Todds 4 files into data folder and edit his app_info according to cpu edition.
Any thoughts and pointers would be welcome, if its preferable to use another os or a particular nvidia driver (or the new beta even) please let me know and i will proceed...caffeine on tap and all 3 brain cells fully functional.
Thanks All
ID: 999142 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6662
Credit: 121,090,076
RAC: 0
United States
Message 999156 - Posted: 27 May 2010, 16:50:36 UTC - in response to Message 996735.  

My three GTX-480's are working ok - I got the files from the CUDA SDK and some from Rastimer. All of my results are completing and being validated - had 42k RAC and rising until the server crash over the weekend. Had 1000 WU's that were waiting to be sent in 24 hours - finally uploaded this morning.

Here is my host http://setiathome.berkeley.edu/show_host_detail.php?hostid=5220505

The files that I am using and an example app_info.xml file can be downloaded here http://rapidshare.com/files/388464735/setiathome.berkeley.edu.rar- your results may vary so please keep that in mind. Sorry for using Rapidshare - it was the quickest way for me to upload these files before I had to head out to a client site.
Todd




Ok round 14 day 5 ;-)
i7 930
w7 64 bit
gtx 480
Boinc 6.10.56
Lunatics optimized 0.36
Todds files
Ran cpu-z processor shows MMX.SSE (1,2,3,3S,4.1,4.2) When it comes to installing opt .36 which multibeam cpu edition is best to use? Have unchecked astropulse to try and make it simpler when it comes to edit app_info. On page 2 of .36 install which cuda mb v and when it comes to cuda library 2.2 or 2.3?
Once i have installed Boinc and .36 drop Todds 4 files into data folder and edit his app_info according to cpu edition.
Any thoughts and pointers would be welcome, if its preferable to use another os or a particular nvidia driver (or the new beta even) please let me know and i will proceed...caffeine on tap and all 3 brain cells fully functional.
Thanks All


What I did to get mine working was not the best way, but it worked. I had depleted my cache to nothing, and went with a fresh install of BOINC 6.10.43. I attached to the project, and set it to not accept work. I downloaded and unzipped all of Todds files from the above link. I coppied every one of the files that were in his Program Data folder directly to the Seti@Home folder in Program Data/BOINC/Projects. I coppied everything including the app_info file, and didn't edit a thing. I set BOINC to accept new work, and took off like a rocket. I know this was not the right way to do it, but it worked beautifully!

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 999156 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999157 - Posted: 27 May 2010, 16:59:51 UTC - in response to Message 999156.  

Thanks for the reply Steve, am going to try the "direct approach" like you did, will get back to you and let you know what happens.
Cheers ;-)
ID: 999157 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6662
Credit: 121,090,076
RAC: 0
United States
Message 999162 - Posted: 27 May 2010, 17:11:41 UTC - in response to Message 999157.  

Thanks for the reply Steve, am going to try the "direct approach" like you did, will get back to you and let you know what happens.
Cheers ;-)


One other thing. I did exit BOINC before I coppied everything into it. There were two folders in Todd's download. Program Data, and Program Files. I tried to copy the program files as well, but it kept stopping me saying another program was using them. I don't know which program because BOINC was shut off. Anyway, I gave up on the Program Files folder, and restarted BOINC, then set it to accept work. Unfortunately I am not skilled enough to troubleshoot beyond what I just told you, but I intend to be in the near future. I love learning, and I am an advanced LabVIEW programmer, that is just learning C.

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 999162 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999163 - Posted: 27 May 2010, 17:16:34 UTC

I don't know wether to laugh or cry, your direct approach Steve seems to have worked ok.
My cache had ran dry so followed your directions, i let boinc download some tasks not too many just in case.
Sure enough it fired up right away, just one worry the regular wu's came down as 46 hours estimated and the cuda tasks showed 35 hours!
The first cuda task of 35hrs has done 25% in 5 mins, i am extremely greatfull its working but may i ask you for a little more advice on those wu timings.

Cheers
ID: 999163 · 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 999175 - Posted: 27 May 2010, 18:53:49 UTC - in response to Message 999163.  

Hi ToxicTBag,
Someone might correct me if I'm wrong but you might try adding your flops counts in your app-info. The instructions are in this thread http://setiathome.berkeley.edu/forum_thread.php?id=53593

That should get your times a bit closer to reality. I'm pretty sure they should work ok with the new Fermi's.


PROUD MEMBER OF Team Starfire World BOINC
ID: 999175 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6662
Credit: 121,090,076
RAC: 0
United States
Message 999176 - Posted: 27 May 2010, 19:13:16 UTC

I just found this book that I will be ordering when I get home that may help those who want to understand more about what is going on under the hood of these apps.

http://www.amazon.com/Beginning-XML-2008-Professional-Experts/dp/1430209976/ref=sr_1_2?ie=UTF8&s=books&qid=1274987360&sr=1-2

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 999176 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999178 - Posted: 27 May 2010, 19:19:43 UTC

Have downloaded more wu's and timings seem to be a little closer to reality, it still seems to be working ok. I dont want to tempt fate just yet and mess with the app_info incase i get it wrong. Thanks for the input perryjay, maybe if you get time you would not mind takeing a look at the pending/validated on my rig to see if they look ok to you?
That looks like a good read Steve, even if it seems a bit beyond me i think i will give it a try.
Thanks All
ID: 999178 · 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 999205 - Posted: 27 May 2010, 21:21:34 UTC - in response to Message 999178.  

They look great to me especially compared to my E5400 and 9500GT. Have you started using the rescheduler? I noticed you had a couple of VLARs from yesterday. They definitely hit you on the run time on GPUs. You're running slower than Steve but he's got a 980 so I guess that makes the difference.

As far as the flops count, without it you will notice a pretty high fluctuation after you run a few shorties then do a normal work unit but I know what you mean about not wanting to mess with the app-info. I put it off as long as I could but it finally bugged me enough I talked myself into it.


PROUD MEMBER OF Team Starfire World BOINC
ID: 999205 · Report as offensive
Profile SciManStev Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Jun 99
Posts: 6662
Credit: 121,090,076
RAC: 0
United States
Message 999207 - Posted: 27 May 2010, 21:29:38 UTC

I'm just glad you got it crunching. I take no credit for anything to do with the apps, I just coppied what others had already done. I really didn't think it would work, but my rig was ready, and I was ready to experiment. I am determined to learn what I need to know, to perhaps make some contribution in the future. I do have my shader and memory over clocked, and my 980 is running at 4.2339 Ghz. At least I know that no matter how hard I drive these 480's, heat will never be a problem. If Furmark could only get them up to 56°C and 58°C, then I have no worries. I intend to experiment with Rivatuner and see if I can increase the clock speed. EVGA Precision tool has the clock speed greyed out, and I can't adjust it.

Perryjay, thank you for checking his work. I was greatful when you took a look at mine and assured me it was OK.

Steve
Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website
ID: 999207 · Report as offensive
ToxicTBag

Send message
Joined: 5 Feb 10
Posts: 101
Credit: 57,197,902
RAC: 0
United Kingdom
Message 999219 - Posted: 27 May 2010, 22:27:40 UTC

Hi Perryjay
Thanks for taking the time to check the work. I am just gaining an understanding of vlars etc and with some coaching from MadMac finally feel confident enough to use rescheduler.I have just ran it on the 480 rig and it seems to have worked ok, there might be a few vlars still to show but i think i have the idea right.
Earlier my heart sank when i realised it started crunching gpu tasks really quickly then slowed right down, as this was what was happenning on my previous attempts over the last 5 days. I realise now of course they do tend to start off quite slow and seem to speed up around the 50% complete level so what looks like an eternity to begin with usually comes down to around 10-11 minutes.
Thank you to you to SciManStev, i couldn't actually believe that after 5 days of trying that i looked at your message and thought "no it can't be that simple" but it was and thank you for the encouragement to keep trying too.
I know there's a long way to go with these fermi chips and were probably looking at a while to see them fully optimized. As much as it was a pain in the butt to get it working i enjoyed the learning experience, there's nothing quite as satisfying as learning something you did not allready know.
I will be keeping a close eye on the rig and look forward to learning more in time.
I want to let it "settle in" for a few days but will keep my word, create a new user and donate the output to gpu users group, sorry i wont be able to bring all hardware with me OcUk is my home team and am there for a long time.
Cheers All ;-))
ID: 999219 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 999297 - Posted: 28 May 2010, 10:48:35 UTC - in response to Message 998735.  

[...]
I tried the autostart with ReSchedule tool, but after every reboot of my PC I got a strange error in it and it didn't worked (WinXP Home 32bit).
I guess it need a delayed execution after reboot in my OS. But I don't know how I could do it. :-#

So all the time I make a reboot the first activity is always to click on the icon on the WinDesktop. :-D


Sutaru,
One possibility is to use a windows script file and place that in your startup folder.

You would need to create a file e.g. reschedule.vbs which contains something like :-


Dim WshShell, oExec
Set WshShell = CreateObject("WScript.Shell")

WScript.Sleep 20000

Set oExec = WshShell.exec("c:\reschedule.exe")


The sleep command is in milliseconds so sleep 20000 delays for 20 seconds etc and then reschedule is executed.

Hope this helps.


John.


Thanks for the hint!

But.. could you (or someone else) explain it how to do it?
Step by step for PC dummies like me.. ;-)

ID: 999297 · Report as offensive
Previous · 1 . . . 7 · 8 · 9 · 10 · 11 · Next

Message boards : Number crunching : CUDA MB V12b rebuild supposed to work with Fermi GPUs


 
©2025 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.