+- +-
Say hello if visiting :) by Gecko
11 Jan 2023, 07:43:05 pm

Seti is down again by Mike
09 Aug 2017, 10:02:44 am

Some considerations regarding OpenCL MultiBeam app tuning from algorithm view by Raistmer
11 Dec 2016, 06:30:56 am

Loading APU to the limit: performance considerations by Mike
05 Nov 2016, 06:49:26 am

Better sleep on Windows - new round by Raistmer
26 Aug 2016, 02:02:31 pm

Author Topic: SETI MB CUDA for Linux  (Read 387571 times)

Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #780 on: 10 Jul 2011, 07:41:20 pm »
What are you trying to do with X? Do you have 1 or 2 monitors? Your xorg.conf is like you want to run two X screens in one X server, while your xorg.0.log says that you have only one monitor connected and then it proceeds to unload the nvidia kernel module  :-\

(II) UnloadModule: "nvidia"

Running
lsmod | grep nvidia
what does it give you?

Also, What does
ls /dev/nv*
give you?

I edited your xorg.conf commenting out the double device, screen and monitor sections and changing some other minor stuff.

I see at mandriva's site that they have released 2011 RC1. Why don't you try something newer?
« Last Edit: 10 Jul 2011, 08:16:59 pm by sunu »

Offline Terror Australis

  • Squire
  • *
  • Posts: 31
Re: SETI MB CUDA for Linux
« Reply #781 on: 10 Jul 2011, 09:52:29 pm »
Hi Sunu
A quick explanation and time line,
Built machine and installed OS and BOINC - latest available updates applied - Started machine with only one monitor connected -  When BOINC running noticed only one GPU was being used - searched web for answers - one answer was to make 2nd card think it had a monitor connected - Even with a 2nd monitor physically connected 2nd card still not crunching - tried several more of the "fixes" found on the web - still no joy - checked hardware by installing different OS - AOK - help sought from Lunatics.
SETI is down atm and the box is out of work, I'll try your xorg file when the project is back up and report back then.

I find this really intriguing, BOINC recognises both cards and runs 2 instances of the crunching app - it's just that both run on the card in PCIE slot 1 - Using one card only, in either slot it, works properly - Even setting the BIOS to boot using PCIE slot 2 makes no difference. What I haven't done is try a different project to see if that works, I'll try that tonight.

One positive thing to come out of this, my vi skills have definitely improved  :)

Thanks for your help
T.A.

Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #782 on: 10 Jul 2011, 10:22:26 pm »
one answer was to make 2nd card think it had a monitor connected
This is/was for windows. Doesn't matter in linux. Unless you want to run multiple X screens or servers.



I'll try your xorg file when the project is back up and report back then.

You still owe me many things:
- A proper ps... when seti is running (supposedly or not) in two GPUs.
- A proper nvidia-smi output with and without seti running, current and new xorg.conf.
- Those two above lsmod... and ls... with your current and new xorg.conf.
- A xorg.0.log with the new xorg.conf.
« Last Edit: 10 Jul 2011, 10:27:16 pm by sunu »

Offline Terror Australis

  • Squire
  • *
  • Posts: 31
Re: SETI MB CUDA for Linux
« Reply #783 on: 10 Jul 2011, 11:02:56 pm »
Is this with only one GPU?
No - both were installed

Quote
1) Why do you run as root?
I don't - I just happened to be root in the console when I ran those commands

Quote
2) There is no "--device x" flag on the running app. Is this because of small terminal window? Make it bigger and recheck.
Pardon my ignorance but where should that line be ?

Quote
3) This isn't what I would expect from nvidia-smi. Sample ouput:....
That's all that came up. (Is this a clue to the problem ?)


Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #784 on: 11 Jul 2011, 07:11:13 am »
Is this with only one GPU?
No - both were installed
It shows only one instance running.

2) There is no "--device x" flag on the running app. Is this because of small terminal window? Make it bigger and recheck.
Pardon my ignorance but where should that line be ?
Maximize your terminal window and rerun it. Alternatively rerun it with > ps.txt at the end and post the file.


3) This isn't what I would expect from nvidia-smi. Sample ouput:....
That's all that came up. (Is this a clue to the problem ?)
No, it is a clue that you need to update kernel, driver, distro, everything  ;D


Also, how is boinc installed? From repositories or from berkeley? In root or your home directory? I always keep everything boinc/seti related in my home directory and far away from root.
« Last Edit: 11 Jul 2011, 07:39:48 am by sunu »

Offline Terror Australis

  • Squire
  • *
  • Posts: 31
Re: SETI MB CUDA for Linux
« Reply #785 on: 11 Jul 2011, 11:49:29 am »
Just a quick one. I didn't have time today to go through all your suggestions but I did have time to sign up to Milky Way. Both cards ran on this project exactly as they should have. I've gone back to thinking this is a problem with the SAH app and the way it handles multiple instances on multiple devices.

I'm going back to square one tomorrow and will incorporate all your suggestions and see how it goes, if nothing else I'll have a tidier install.

BTW the correct command for nvidia-smi is: nvidia-smi -a

I'll post everything when I get a chance

T.A.

Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #786 on: 11 Jul 2011, 11:56:57 am »
BTW the correct command for nvidia-smi is: nvidia-smi -a

It must have been with your driver version. As I said I don't remember how nvidia-smi worked with that driver version. From nvidia-smi manual:
Quote
The  -a,  -s and -g arguments are now deprecated in favor of -q and -i, respectively. However, the old arguments still work for this release.

EDIT: I just checked that app and it doesn't obey --device x flags. So no matter what we do, I don't think we can make it utilize other than the first GPU.
« Last Edit: 11 Jul 2011, 12:32:48 pm by sunu »

Offline riofl

  • Knight o' The Round Table
  • ***
  • Posts: 240
Re: SETI MB CUDA for Linux
« Reply #787 on: 24 Feb 2012, 02:11:16 pm »
Been inactive for a while. boincing fried both my vid cards. both have bad memory now and cause periodic lockups in my system.. been saving for a new card. was looking at the nvidia 590 but i am wondering if i should wait for the 690.. i assume the 690 will be dual gpu like the 590. thing is i need support for 4 monitors and the 590 does that which is nice. hope the 690 does too...

so... any thoughts about 590 vs 690?

Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #788 on: 24 Feb 2012, 06:53:56 pm »
Nobody really knows what 690 will be like. It is also unknown when exactly it will be out, some say by the 3rd quarter. I would still wait for it.

As for your 4 monitors I think it would be better to go to a dedicated card for them. That is, use a separate card for display purposes and dedicate the 690 for cuda computing. Look at http://www.nvidia.com/object/desktop-nvs.html

This is what I'm thinking to do in the future. I'm also waiting for the 690 to replace my old 295+285 setup and use the 690 for cuda and a lowly card, like a 9600gt, for display (I don't have big requirements like your 4 monitor setup), if my motherboard cooperates (if it can run a display from the secondary PCIE slot).
« Last Edit: 24 Feb 2012, 07:23:51 pm by sunu »

Offline riofl

  • Knight o' The Round Table
  • ***
  • Posts: 240
Re: SETI MB CUDA for Linux
« Reply #789 on: 24 Feb 2012, 11:11:19 pm »
that nvs is interesting.. extremely low power too.. might be a bit of a mess making x work with it though. it only supports
2560x1600 per screen however i have 2 screens defined as 3840x1200 spanning across 2 monitors each screen, then i have those 2 screens 'twinview' enabled so i can move the mouse from one to the other screen. i would have preferred twinview all the way across but i could not find a way to do it so each screen was well defined and unique. didnt work so i had to go to the above solution. then again maybe that will work since each screen spans 2 ports. too tired to think properly. but this looks like a very nice setup.. much better than sharing my desktop with boinc.


Offline Jason G

  • Construction Fraggle
  • Knight who says 'Ni!'
  • *****
  • Posts: 8980
Re: SETI MB CUDA for Linux
« Reply #790 on: 24 Feb 2012, 11:31:52 pm »
echoing Sunu's comments regarding the 690 directly, it looks like a good time to sit and wait if you can hold out & don't need a new card right now.  It's looking like an uncommon situation development wise at the moment, where  nv driver & dev support staff appear to be in hiding off doing something else (for example the latest windows WHQL drivers aren't very well polished & there is currently zero publicly visible progress on reported issues).   With the company having recently reported something like $4 billion earnings, I'd say either the staff are hard at work on that next gen to get it 'right'... or possibly abducted by aliens.

Anyway, given that, by my understanding, Linux drivers tend to be simpler to develop than Windows ones, probably you won't see much of the expected teething problems over here in Linux land, but might be in for a quiet wait.  Windows display driver model based drivers, have already been buckling under multiple card setups for some time with latency & traffic issues [and most recently basic power management...  ::) ], so it's fairly logical that the situation will be worse for us in the short term & nv will have to find ways around that.

Jason
« Last Edit: 24 Feb 2012, 11:48:49 pm by Jason G »

Offline riofl

  • Knight o' The Round Table
  • ***
  • Posts: 240
Re: SETI MB CUDA for Linux
« Reply #791 on: 25 Feb 2012, 07:49:10 am »
ugh. just reviewed the specs of the nvs 420 and 450.. neither will work for me :( they only have 512mb ram . i would clobber that in a heartbeat. typically my setup chews up a good 1gb+  vidram and thats without boinc running..

Offline Mike

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 2427
Re: SETI MB CUDA for Linux
« Reply #792 on: 25 Feb 2012, 08:12:51 am »
ugh. just reviewed the specs of the nvs 420 and 450.. neither will work for me :( they only have 512mb ram . i would clobber that in a heartbeat. typically my setup chews up a good 1gb+  vidram and thats without boinc running..

You can get a EVGA 450 SC with 1GB.

Mike

Offline sunu

  • Alpha Tester
  • Knight who says 'Ni!'
  • ***
  • Posts: 771
Re: SETI MB CUDA for Linux
« Reply #793 on: 25 Feb 2012, 09:56:43 am »
If it is only for display, what do you need 1GB for? Maybe use 2 NVSs?

Nvidia gives extra capabilities/configuration options to their professional series of cards. For example those cards support BaseMosaic. BaseMosaic seems the perfect solution in your case:
Quote
Option "BaseMosaic" "boolean"

    This option can be used to extend a single X screen transparently across all of the available display outputs on each GPU. This is like SLI Mosaic mode except that it does not require a video bridge connected to the graphics cards. Due to this Base Mosaic does not guarantee there will be no tearing between the display boundaries. Base Mosaic is supported on all the configurations supported by SLI Mosaic Mode. It is also supported on Quadro FX 380, Quadro FX 580 and all G80 or higher non-mobile NVS cards.

    Use this in conjunction with the MetaModes X configuration option to specify the combination of mode(s) used on each display. nvidia-xconfig can be used to configure Base Mosaic via a command like nvidia-xconfig --base-mosaic --metamodes=METAMODES where the METAMODES string specifies the desired grid configuration. For example, to configure four DFPs in a 2x2 configuration, each running at 1920x1024, with two DFPs connected to two cards, the command would be:

        nvidia-xconfig --base-mosaic --metamodes="GPU-0.DFP-0: 1920x1024+0+0, GPU-0.DFP-1: 1920x1024+1920+0, GPU-1.DFP-0: 1920x1024+0+1024, GPU-1.DFP-1: 1920x1024+1920+1024"

« Last Edit: 25 Feb 2012, 10:07:12 am by sunu »

Offline riofl

  • Knight o' The Round Table
  • ***
  • Posts: 240
Re: SETI MB CUDA for Linux
« Reply #794 on: 27 Feb 2012, 08:08:08 am »
ugh. just reviewed the specs of the nvs 420 and 450.. neither will work for me :( they only have 512mb ram . i would clobber that in a heartbeat. typically my setup chews up a good 1gb+  vidram and thats without boinc running..

You can get a EVGA 450 SC with 1GB.

Mike


yeah i was just looking at he power consumptions... 40w max. i if i get a 590 or equivalent 690 ?, i am afraid the power requirements would leave me lacking for my display card if i went with a mainline card.

 

Welcome, Guest.
Please login or register.
 
 
 
Forgot your password?
Members
Total Members: 97
Latest: ToeBee
New This Month: 0
New This Week: 0
New Today: 0
Stats
Total Posts: 59559
Total Topics: 1672
Most Online Today: 24
Most Online Ever: 983
(20 Jan 2020, 03:17:55 pm)
Users Online
Members: 0
Guests: 31
Total: 31
Powered by EzPortal