Linux x64 Cuda Multibeam (x41g)

Message boards : Number crunching : Linux x64 Cuda Multibeam (x41g)
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile ivan
Volunteer tester
Avatar

Send message
Joined: 5 Mar 01
Posts: 783
Credit: 348,560,338
RAC: 223
United Kingdom
Message 1182780 - Posted: 4 Jan 2012, 17:40:29 UTC

Darn! I got the sources compiled with CUDA 4.0 (a bit extra faffing to get it to use gcc 4.4 on the SLC5 machine... Also had to manually tweak the makefile as the older bash doesn't support substitution in environment variables.). First two WUs on the SLC6 machine finished in less than 30 seconds, but it turns out they were -9 overflows and verified OK. No other problem there except that I got a file-name wrong by one character today and it burned through about 15 WUs with "file not found" before I got to the suspend button.

Things not so rosy with the GTX 460 in the SLC5 server box. Temperatures started at 43 C, and the job seemed to run OK but when the temp hit 100 C after about 2 minutes and 15% I suspended GPU work. :-( Looks like the forced airflow and standard heatsink aren't enough to cope with the heat generated. I have a vortexx fan/sink that I bought when the OEM fan disintegrated but Palit don't use the nVidia reference design and it wouldn't fit the intended way; turned by 180-deg the offset fan fouled motherboard components -- it might just fit on the server's riser card if there's a spare +12V from the power supply.
ID: 1182780 · Report as offensive
Profile ivan
Volunteer tester
Avatar

Send message
Joined: 5 Mar 01
Posts: 783
Credit: 348,560,338
RAC: 223
United Kingdom
Message 1182844 - Posted: 4 Jan 2012, 22:19:36 UTC

OK, after a bit of experimenting I put this in my cc_config.xml file, and BOINC then activated the C1060 rather than the GTX 460:
<ignore_cuda_dev>1</ignore_cuda_dev>
Eventually, a result was verified with a wingman so I turned "Allow new tasks" back on.

Note all this, despite my brain-fades earlier, was with CUDA 4.0. I tried 4.1 on the SLC5 machine, but its libraries were too old. Next, I'll install 4.1 on the SLC6 machine, and see if there are any code adjustments needed. For 4.0, none were.

Oh, the C1060 was idling at 55 C and 33% fan; under seti@home load it is (or was -- I seem to have caught a couple of shorties and run out of GPU WUs!) running at 79 C and 49% fan. Ah, here we go again, just landed a batch of 17 WUs...
ID: 1182844 · Report as offensive
Profile ivan
Volunteer tester
Avatar

Send message
Joined: 5 Mar 01
Posts: 783
Credit: 348,560,338
RAC: 223
United Kingdom
Message 1183004 - Posted: 5 Jan 2012, 18:38:53 UTC - in response to Message 1182844.  

Oh, the C1060 was idling at 55 C and 33% fan; under seti@home load it is (or was -- I seem to have caught a couple of shorties and run out of GPU WUs!) running at 79 C and 49% fan. Ah, here we go again, just landed a batch of 17 WUs...

Ahh... It helps if the forced air fans are running rather than idling! When I set up the server I changed the fan control setting in the BIOS from the default None (fans at 100%) to High Performance, to keep the noise down. Looks like that means the four centre fans are then controlled from the CPU temps, blowing through a shroud over the CPUs, while the two outer pairs, which suck across the MB and blow down the PCI-E bays, are controlled as "case fans" -- presumably from sensors somewhere on the MB. Because the CPU fans were working fairly hard I didn't realise the "case fans" were just idling until I took the cover off while it was running. So I went into the BIOS and changed back to the default.

Before, the C1060 (which has a built-in fan) idled at 33% fan/55 C and went to 45%/79 C running SETI@home. The GTX 460 idled at 45%/43 C and went to 100%/100+ C with seti (the broken fan is removed so the %-age is what its temperature controller thinks it should be running at).

Now, they idle at <30%/43 C and 33%/33 C, and run at 35%/70 C and 98%/89 C resp. :-) There's scope for putting a deflector to direct more air over the 460's finning as it only covers half of the double-width PCI-E bay that the fans blow down.
ID: 1183004 · Report as offensive
Jeroen
Volunteer tester

Send message
Joined: 28 Mar 01
Posts: 13
Credit: 8,155,940
RAC: 0
United States
Message 1183063 - Posted: 5 Jan 2012, 23:44:43 UTC - in response to Message 1181309.  

This new build is working great on my Linux systems. Performance is vastly improved on my GTX 295 and similar via my GTX 580 compared to the previous version. Thanks.

Linux x64 build of Cuda Multibeam x41g
Available now in the Lunatics, Multibeam for Linux Downloads section
, mirrored at Arkayn's Site


ID: 1183063 · Report as offensive
Profile aaronh
Volunteer tester
Avatar

Send message
Joined: 27 Oct 99
Posts: 169
Credit: 1,442,686
RAC: 0
United States
Message 1183449 - Posted: 7 Jan 2012, 8:47:57 UTC - in response to Message 1182844.  

Note all this, despite my brain-fades earlier, was with CUDA 4.0. I tried 4.1 on the SLC5 machine, but its libraries were too old. Next, I'll install 4.1 on the SLC6 machine, and see if there are any code adjustments needed. For 4.0, none were.

It was actually found that CUDA 4.0 was slower for most (all?) scenarios, and although things were fixed to work with 4.0, we stayed with using 3.2.

However, 4.1rc2 might be showing promise, with that new fandangled llvm compiler...
ID: 1183449 · Report as offensive
Team kizb

Send message
Joined: 8 Mar 01
Posts: 219
Credit: 3,709,162
RAC: 0
Germany
Message 1183559 - Posted: 7 Jan 2012, 20:12:45 UTC - in response to Message 1181623.  
Last modified: 7 Jan 2012, 20:14:27 UTC

...We're upgrading all our GPU systems to CUDA 4.1 at the moment, so it'd be nice to see the project move ahead to that level (some minor changes to APIs, etc. mean that problems are to be expected)...


Hi Ivan. Glad to hear you're back in the game :) Yeah development Cuda 4.1rc2 builds for Windows already exist under NDA, either building from the current X-branch svn head against 4.1rc2 SDK, or pestering Aaron for a test Linux build should work, assuming you're registered with nVidia so can get the SDK legally.

Jason

Sorry about the lack of communications, 2011 wasn't a year of great motivation following on from job and health matters in 2010; my job situation wasn't settled until March, but latest info is that I appear to be funded up until about a year from my nominal retirement in five year's time. My colleague and office-mate is registered for the 4-series CUDA (I'd registered years ago, not sure if that still works) so I just use his downloads to get the SDK. Battling a memory allocation problem in my cufft plans in the holographic movies at the moment; something's changed and I need to start looking at the memory map again (4Kx4K holograms are now failing on 512 MB cards, but still work on a C2070).


I wish I had a clue what your talking about, way over my head!
My Computers:
â–ˆ Blue Offline
â–ˆ Green Offline
â–ˆ Red Offline
ID: 1183559 · Report as offensive
Profile ivan
Volunteer tester
Avatar

Send message
Joined: 5 Mar 01
Posts: 783
Credit: 348,560,338
RAC: 223
United Kingdom
Message 1183765 - Posted: 8 Jan 2012, 13:18:42 UTC - in response to Message 1183559.  

I wish I had a clue what your talking about, way over my head!

I'm a physicist, you're not supposed to understand me! ;-)
ID: 1183765 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1183877 - Posted: 8 Jan 2012, 21:56:27 UTC - in response to Message 1183765.  

I wish I had a clue what your talking about, way over my head!

I'm a physicist, you're not supposed to understand me! ;-)

Ok, Yer a physicist, I understood Ya.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1183877 · Report as offensive
S@NL - John van Gorsel
Volunteer tester
Avatar

Send message
Joined: 5 Jul 99
Posts: 193
Credit: 139,673,078
RAC: 0
Netherlands
Message 1184030 - Posted: 9 Jan 2012, 13:37:00 UTC

I have tried it on this pc and it seems to work fine. Due to the lack of work I rescheduled a CPU task (AR=2.7) and it finished in 2m40 where the previous application took 6-7 minutes for a similar task. I had an identical pc running WinXP and that one finished a similar task in 2m30

This pc uses a stock GTX260 and driver 275.09.07.


Seti@Netherlands website
ID: 1184030 · Report as offensive
Profile ivan
Volunteer tester
Avatar

Send message
Joined: 5 Mar 01
Posts: 783
Credit: 348,560,338
RAC: 223
United Kingdom
Message 1185045 - Posted: 13 Jan 2012, 22:02:19 UTC - in response to Message 1183004.  

Before, the C1060 (which has a built-in fan) idled at 33% fan/55 C and went to 45%/79 C running SETI@home. The GTX 460 idled at 45%/43 C and went to 100%/100+ C with seti (the broken fan is removed so the %-age is what its temperature controller thinks it should be running at).

Now, they idle at <30%/43 C and 33%/33 C, and run at 35%/70 C and 98%/89 C resp. :-) There's scope for putting a deflector to direct more air over the 460's finning as it only covers half of the double-width PCI-E bay that the fans blow down.

Well, I finally fitted a deflector today, lovingly crafted from some material our technician supplied -- the side of a Wheat Krispies packet! Just a simple ramp up to the heatsink, continuing along to keep all the air flowing over the card, and a ramp down again. Unfortunately my ASCII-art impression is mangled by the display software...

Anyway, the upshot was that the idle temp for the 460 dropped to 25 C, and at full SETI it's a whole 20 C cooler at 69 C (I had to stop using the 460 earlier during a shortie storm -- those calculations pushed it up to 97 C!).

ID: 1185045 · Report as offensive
Profile Crunch3r
Volunteer tester
Avatar

Send message
Joined: 15 Apr 99
Posts: 1546
Credit: 3,438,823
RAC: 0
Germany
Message 1185064 - Posted: 13 Jan 2012, 23:06:34 UTC - in response to Message 1182272.  

unfortunately i was unable to find a link to the source code to compile it myself on an old SuSE 11.2 (and it's not listed there -> http://lunatics.kwsn.net/index.php?module=Downloads;catd=2 either which is a bit odd since it's GPLed software)



Why would you point the bone like that?


What? The source code wasn't available at lunatics as it's supposed to be... maybe that's why !?!


When you were given full svn access from the start.


Uhmm... that was like 2 years ago... do you honestly think i remember that ?!?
Nor does it help anyone who tries to compile a app specific to their OS library deps... But that is kind of unwanted since requesting access to the source is answered with banning people from accessing a hidden SVN code base !!!!

Then you know full well the requirements are to make the source for release builds available on request.


So this is your reply to my "request" on getting the source code...that's interesting...blocking public access. Is that really the way to go ? I don't think so since all changes to the lunatics code haven't been public in the first place since you got involved in it... except for Raistmers releases, that for some reason were published including source... go figure...


If you require updated sources & forgot how to use the SVN, or would prefer a different method, just ask... and don't ever try beat me over the head with GPL again, I actually read it.


How about a public SVN in the first place and stop being so paranoid about code ...

Join BOINC United now!
ID: 1185064 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1185072 - Posted: 13 Jan 2012, 23:26:53 UTC - in response to Message 1185064.  

So this is your reply to my "request" on getting the source code...

It would probably have gone down better if you hadn't been so cavalier over the (L)GPL in the past: available to the team only. Really?

And if you hadn't been so negative about SETI in more recent postings on other project web sites. A shame your expertise wasn't applied here after our collaboration over the 2.4 V(ista) apps.
ID: 1185072 · Report as offensive
Profile Crunch3r
Volunteer tester
Avatar

Send message
Joined: 15 Apr 99
Posts: 1546
Credit: 3,438,823
RAC: 0
Germany
Message 1185078 - Posted: 13 Jan 2012, 23:36:51 UTC - in response to Message 1185072.  
Last modified: 13 Jan 2012, 23:46:16 UTC


It would probably have gone down better if you hadn't been so cavalier over the (L)GPL in the past: available to the team only. Really?


Ahh... you like digging around in the past(5 years old post)... why not mention that i contributed the source code to boinc that enabled ATI GPUs support ????


I guess that you forgot about that one... try searching on TRAC you find the entry where that was mentioned... i'm sure you can find that one...


And if you hadn't been so negative about SETI in more recent postings on other project web sites. A shame your expertise wasn't applied here after our collaboration over the 2.4 V(ista) apps.


Well... my 'expertise' was applied to milkyway@home using my hand code sse2-sse4.1 code that is now part of the OFFICIAL code base which by some unknown reason is publicly available there -> http://github.com/Milkyway-at-home/milkywayathome_client

Do a search and you'll find the entry when my code was integrated into the stock source code... if not, ask Matt Arsenault over at MilkyWay@Home via PM. I'm sure he can point you to the right place to search for it.
If that is to hard todo, i'll gladly find that one for you and post it here... in public....

Join BOINC United now!
ID: 1185078 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1185083 - Posted: 13 Jan 2012, 23:44:16 UTC - in response to Message 1185078.  

So this is your reply to my "request" on getting the source code...

It would probably have gone down better if you hadn't been so cavalier over the (L)GPL in the past: available to the team only. Really?

Ah... you like digging around in the past... why not mention that i contributed the source code to boinc that enabled ATI GPUs support ????

I guess that you forgot about that one... try searching on TRAC you find the entry wher that was mentioned... i'm sure...

And if you hadn't been so negative about SETI in more recent postings on other project web sites. A shame your expertise wasn't applied here after our collaboration over the 2.4 V(ista) apps.

Well... my 'expertise' was applied to milkyway@home using my hand code sse2-sse4.1 code that is now part of the OFFICIAL code base which by some unknown reason is publicly available there -> http://github.com/Milkyway-at-home/milkywayathome_client

I'm glad for you. Collaboration and respect is the name of the game, as is clear from the opening post in this thread.
ID: 1185083 · Report as offensive
Profile Crunch3r
Volunteer tester
Avatar

Send message
Joined: 15 Apr 99
Posts: 1546
Credit: 3,438,823
RAC: 0
Germany
Message 1185086 - Posted: 13 Jan 2012, 23:54:44 UTC - in response to Message 1185083.  
Last modified: 14 Jan 2012, 0:16:45 UTC


I'm glad for you. Collaboration and respect is the name of the game, as is clear from the opening post in this thread.


Yes, Collaboration is the key and unlike here, Full source code(optimized or not) is available over at MilkyWay@Home. That is a key to helping out the project in the first place!

Milkyway opened up the source code and while using most of my personal code in the stock app now, enabling it to increase processing speed by a factor of at least 2.5.

Over at MilkyWay there's no question about where to find the code since all of it (including optimized code(GPLed)) is public... unfortunately this is not the case for SETI and the lunatics code.

On the contrary, one get's banned from accessing a hidden code repository... you honestly think that's the way to go ?

Besides all that.. i'm also the one that helped out PrimeGrid codeing a new LLR Wrapper (LLR v6.09, evidence can be found over at the PrimeGrid forum)... and of course there are a bunch of other projects as well that you don't even know about where i got involved (at least more than 6 projects requested and got help)...but it seems, you don't care about that, which is sad...

Instead you choose to dig out a 5+ years old post to support your claims,which is of course obsolete. You're trying to play dirty here without giving people the recent facts ?

Join BOINC United now!
ID: 1185086 · 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 1185091 - Posted: 14 Jan 2012, 0:18:29 UTC - in response to Message 1185064.  

What? The source code wasn't available at lunatics as it's supposed to be... maybe that's why !?!


Read the GPL. You obviously have not. It is, as always has been, available on request. If you think it has to be posted on a website, then you are mistaken, though I have now done so especially for you, since you aren't a team player & repository access wasn't good enough.

Perhaps being courteous when you want something would be a better approach, for now you burn your bridges with me & are now on ignore.

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: 1185091 · Report as offensive
Profile Crunch3r
Volunteer tester
Avatar

Send message
Joined: 15 Apr 99
Posts: 1546
Credit: 3,438,823
RAC: 0
Germany
Message 1185097 - Posted: 14 Jan 2012, 0:34:05 UTC - in response to Message 1185091.  
Last modified: 14 Jan 2012, 0:36:39 UTC


Read the GPL. You obviously have not. It is, as always has been, available on request. If you think it has to be posted on a website, then you are mistaken,


Yes, but that requires at least an address/e-mail where to request a download link to the source. None of that is available.... Not even in any of the readme files included in the installer !


Perhaps being courteous when you want something would be a better approach, for now you burn your bridges with me & are now on ignore.
Jason


Courteous, would be mentioning at least an email address to mail to getting the source. In my initial post i wasn't even poking at the fact that none of that was possible but giving a hint that compiling on a not so recent linux would be the way to go.
Obviously you've taken that as a personal insult... which sort of baffled me,since it was only 'directed' at the one compiling the app in the first place to include older linux distros.

However, having a look at your recent replies and you feeling the need to taking personal some posts that weren't even directed at you in the first place... it's kind of curious why you feel the need to band people from hidden 'public' SVNs...

Join BOINC United now!
ID: 1185097 · 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 1185100 - Posted: 14 Jan 2012, 0:47:19 UTC - in response to Message 1185097.  
Last modified: 14 Jan 2012, 0:47:49 UTC

However, having a look at your recent replies and you feeling the need to taking personal some posts that weren't even directed at you in the first place... it's kind of curious why you feel the need to band people from hidden 'public' SVNs...


The SVN is private, not public. Sources for release builds available on request as always. I'm not surprised you can't remember how it works, or where the website is, since you never bother to visit until there's something you want. Big surprise you post here about something being unavailable without even asking on the right forum.

Yes, bandying about misinterpretations of legal documents ARE personal attacks, so wise up & adjust the attitude. You have a problem with asking politely & prefer to point the finger. Go look in the mirror.

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: 1185100 · Report as offensive
Profile Crunch3r
Volunteer tester
Avatar

Send message
Joined: 15 Apr 99
Posts: 1546
Credit: 3,438,823
RAC: 0
Germany
Message 1185107 - Posted: 14 Jan 2012, 1:08:26 UTC - in response to Message 1185100.  
Last modified: 14 Jan 2012, 1:16:18 UTC

However, having a look at your recent replies and you feeling the need to taking personal some posts that weren't even directed at you in the first place... it's kind of curious why you feel the need to band people from hidden 'public' SVNs...

The SVN is private, not public. Sources for release builds available on request as always.
[quote]

So where is the public address to request access to it ? Please point me to one that hasn't been 'public' since 'this' started...

I'm not surprised you can't remember how it works, or where the website is, since you never bother to visit until there's something you want.


So calling me stupid being unable to figure out a way to access a hidden 'public' SVN is your way to go... Besides that please point out exactly what i wanted to that YOU had... There's nothing you personally could offer that would interest me in any way.(This is where YOU start personal ATTACKS
)

Big surprise you post here about something being unavailable without even asking on the right forum.



Yes, bandying about misinterpretations of legal documents ARE personal attacks, so wise up & adjust the attitude. You have a problem with asking politely & prefer to point the finger. Go look in the mirror.

Jason


Wow... that's just one of the reasons why i choose to getting rid of the whole SAH garbage.. Just try reading your post again and think about why you feel personally attacked in the first place without my post being direct at yourself in the first place ... It's kind of disgusting.

Join BOINC United now!
ID: 1185107 · 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 1185110 - Posted: 14 Jan 2012, 1:16:08 UTC

Oh well, best of luck with your collaborations then. I certainly hope you have less communication difficulties at the projects you believe in helping more. It's funny, I never had a problem with you before, but you made one now. Let's call it a personality clash & move on then.

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: 1185110 · Report as offensive
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : Linux x64 Cuda Multibeam (x41g)


 
©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.