Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 4 Mar 2012 05:15:39 +0400
From: Solar Designer <>
Subject: Re: CUDA & OpenCL status

On Sun, Mar 04, 2012 at 01:53:31AM +0100, Lukas Odzioba wrote:
> 2012/3/4 Solar Designer <>:
> > Why do you?  Not enough PCIe x16 slots?
> I've got 2*8 PCIe or 1*16. I had problems installing amd and nv
> drivers (X's didn't want to start as far as I remember),

In my case the NVidia card is currently primary (by slot number) - so
BIOS and Ubuntu bootup messages go to it.  Then which card is used by X
is determined by current xorg.conf contents - I can make it either way
without swapping the cards physically.  I suppose I could also use both
and connect multiple monitors if I wanted to, but I mostly just use this
new machine over SSH from my main desktop machine now.

FWIW, at the time of the tests that I posted, I had X running (but not
in active use since I was logged in via SSH instead) on the AMD card
only.  The NVidia card was displaying some text in a framebuffer (no X
running).  With lsmod, I can see only the AMD driver in use:

user@...l:~$ lsmod | egrep 'fglrx|nvidia'
nvidia              12294232  0
fglrx                3129133  106

When I run X with xorg.conf set to use NVidia, it's 0 use count for
fglrx and some non-zero number for nvidia.

...I've just started "LD_LIBRARY_PATH=/usr/local/cuda/lib64 ../run/john
-te=10 -fo=phpass-cuda &" and quickly re-ran the lsmod command above:

nvidia              12294232  12
fglrx                3129133  106

So the driver is indeed made use of by CUDA/OpenCL programs and this
does not require having X use the driver as well.


Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ