Amazon EC2 PBuffer FBO rendering dual Tesla xconfig EE /dev/fb0

[SpanHard]

Hi,
I am trying to configure an Amazon Machine Image with Ubuntu 11.04 to run a cg1.4xlarge type machine with two Nvidia Tesla s2050 in order to use hardware accelerated OpenGL offscreen rendering using Frame buffer objects.
I managed to install the latest Nvidia drivers on it without problem.
However, in the attached nvidia-bug-report I can see that no /dev/fb0 could be found.
[ 88.482] (II) LoadModule: "fbdevhw"
[ 88.483] (II) Loading /usr/lib/xorg/modules/libfbdevhw.so
[ 88.486] (II) Module fbdevhw: vendor="X.Org Foundation"
[ 88.486] compiled for 1.10.1, module version = 0.0.2
[ 88.486] ABI class: X.Org Video Driver, version 10.0
[ 88.486] (EE) open /dev/fb0: No such file or directory
[ 88.486] (II) Loading sub module "vgahw"
[ 88.486] (II) LoadModule: "vgahw"

I installed X and when attempting to start I also get the message that /dev/fb0 could not be found.
I have read in this thread:
http://www.nvnews.net/vbulletin/showthread.php?t=159580
that jesmith has managed to run OpenGL on EC2s Teslas. I saw in his nvidia-bug-report that no libfbdevhw.so but a libfb.so is loaded instead and no /dev/fb0 is missing in his log:
(II) Loading sub module "fb"
(II) LoadModule: "fb"
(II) Loading /usr/lib64/xorg/modules/libfb.so
(II) Module fb: vendor="X.Org Foundation"
compiled for 7.1.1, module version = 1.0.0
ABI class: X.Org ANSI C Emulation, version 0.3
(II) Loading sub module "wfb"
(II) LoadModule: "wfb"
(II) Loading /usr/lib64/xorg/modules/libwfb.so
(II) Module wfb: vendor="NVIDIA Corporation"

How can I setup the Tesla cards to use libfb.so intend of lbfbdevhw.so?? Also does libfb.so support full hardware acceleration?
I have tried with
sudo nvidia-xconfig -a --use-display-device=None --virtual=1280x1024
and also
sudo nvidia-xconfig -a --connected-monitor=DFP --virtual=1280x1024
to create an /etc/X11/XF86Config that could use OpenGL and GLX with the Tesla cards on EC2s cg1.4xlarge instances.

Any help would be very appreciated!
Many thanks,
Bernhard

[AaronP]

Did nvidia-xconfig create an /etc/X11/XF86Config file? If so, please rename it to /etc/X11/xorg.conf. XF86Config hasn't been used in many years.

libfbdevhw is used by the Cirrus driver driving the primary display device. The Tesla devices are not being configured to use the NVIDIA driver at all because the X server is not finding your configuration file and is falling back to the default configuration.

[SpanHard]

Wow, I cannot believe that this was the problem. I started from a plain Ubuntu 11.04 and installed the NVIDIA driver on that. I don´t know why there was an X86FConfig file in /etc/X11 there and why nvidia-xconfig changes that file instead of the xorg.conf. Well, deleting the X86FConfig and rerunning
sudo nvidia-xconfig -a --use-display-device=None --virtual=1280x1024
solved the whole problem and I could run X with OpenGL support to perform offscreen rendering.
This is great. AaronP I owe you a pint!
Bernhard

[jdesha]

I know this post is a couple of years old now, but my company is also interested in doing off-screen OpenGL rendering with hardware acceleration on the Nvidia Tesla cards at Amazon Web Services. In searching through the AWS forum, the party line seems to be:

“Under Linux, OpenGL can be made to work on the Tesla M2050s in cg1.4xlarge, but it is not officially supported by NVIDIA.”

https://forums.aws.amazon.com/thread...275825&#275825

It sounds like you were successful in making OpenGL work on the Tesla s2050. Did you run into any more problems? How far did you take the project? Are you still doing rendering on Amazon Web Services? Do you know of others who are doing the same? I'd appreciate any guidance you can give me.

Thanks,
Jenny