I basically use my Linux system (Slackware 13.1/64-current with 2.6.35.6 kernel) through SSH, mostly because it connected with a KVM switch to share monitor/mouse/keyboard with another computer and most of the time someone is on the other computer that sharing these devices. On occasion I need to have a visual display (basically because what I trying to do at the time I can't remember the command and need to do it through the desktop (currently KDE 4.5.1). In the past I have used VNC to do this but now most text doesn't show up (like all text in the KDE menu is missing. I want to start investigating and playing around with the Desktop Effects which need 3D acceleration. I tried installing VirtualGL and the Mesa libraries and so on required but that failed and all it did was mess up my 3D acceleration (at least that is what I think did it) I had a post here http://www.linuxquestions.org/questi...s-card-839195/ trying to get it to work, which it does now. Doing searches on this it looks like I could use XDMCP and a client on the Windows machine to connect like Xming-mesa or Hummingbird Exceed and Hummingbird 3D (if I want to go commercial). I have tried XDMCP with Xming-mesa and I get it to connect but I have NO 3D acceleration to me that means a few things:

• I am missing something in my xorg.conf
• X doesn't support 3D acceleration remotely
• Xming-mesa does not supprt 3D

Other then a few font directories not being found there no errors in my Xorg.0.log file. It doesn't matter how I do it as long as I can get remote 3D, but with only using it on occasion would prefer for the server (Linux) something open source or free, and for the client (Windows7) something along the same line. If there is another way to do this I am open for suggestions and willing to try anything.

You can rule out at least some of these.

X does support remote 3D acceleration (I have tested this on a linux->linux system, running only mesa software acceleration on the client (X server) end. It works, though it is fairly slow!

Xming-mesa must support 3D, because that is what mesa is (3D support including software emulation). Not sure how up to date the Xming version is, I don't have a Windows machine here to try it.

You could do a simpler test by using putty to do an ssh connection (with X11 forwarding enabled on the tunnel), and see what glxinfo reports, rather than using XDMCP.

If I do that I get the following (I manually typing it so it prone for errors)
I set LIBGL_DEBUG=verbose and ran glxinfo again and got exact same thing

If I switch over to the Linux system (sitting at console) and run the same command I get
I know my xorg.conf file is using the nvidia driver I don't know where the NV-GLX is coming from. This is for a nvidia 8500GT 1GB card. I just used my normal shell program to forward X11 (SecureCRT). You said that XDMCP was slow is there another solution that would be faster, it would just be used on the LAN and only allow the desktop I am using currently until my laptop comes back from being repaired then allow that as well which can be done with iptables.

I was using ssh, not XDMCP, and the reason for the slowness is that not only does the X and GLX stuff have to be marshalled over the LAN, but in my setup the client machine has no graphics acceleration, so it is all being emulated in mesa. That might not matter if the only acceleration you care about is the limited amount done in the desktop.

When testing using ssh access, the xorg.conf on the server is not relevant, because it is the X server on the client that is being used.

Just some questions:

Are you using the current version of Xming (the 7.5 version)?

If so, are you using the (experimental) '-wgl' flag to invoke it? (probably best to avoid it until after you get the mesa working)

You can force GLX redirection of all rendering (but this might be too slow to be useful), eg by the following (all on one line):

OK when I evoke XDMCP it seems slow after you type in username/password (at least until kde loads all the stuff it does) once it is loaded and on the desktop it seems fine. As far as the desktop 3d acceleration goes it has an onboard graphics with an intel 82945G chip, so I can't expect much here, but my laptop when I get it returned has an ATI Radeon HD4300 with DDR2 512MB so that should do fine.


I think I been doing it wrong then because I was not even running Xming I was using just Xlaunch

I was using the xming-mesa-6-9-0-31 version but will upgrade to the newer version. As far as flags I wasn't using any just going through the Xlaunch dialogs setting things (wasn't saving it for wanted to get it working before I saved).

The xming 7.5 are the fonts at least where I am looking

Using Xlaunch is fine, that starts the X server on your client machine. The ssh to the remote machine (with X11 forwarding enabled) will forward any X requests back to the local machine, and the X server carries them out.

It is a bit more complicated with the GLX extensions, because they are low level, and some of the processing may be handed off to the remote machine before being forwarded (but they will still not involve the X server on the remote machine). My understanding is that the LIBGL_ALWAYS_INDIRECT flag forces the GLX requests to come via X first, which will force all the processing onto the local machine (which will cost network bandwidth, but may solve this NV-GLX issue).

It looks like the current version of Xming includes the mesa library (where in the older version the Xming-mesa was a separate variant).

I see where you getting the 7.5 release I was looking at the "Public Domain Releases" the 7.5 is not in those it is in the "Website Releases" which require a username and password to download. I know it just a donation to get one but I wanted to see how well it worked before I did that. I think what I will do is do something a little faster like vnc which I know doesn't support 3D acceleration, but I only have to use it when someone else is using the other computer attached to the KVM switch, or I believe another option would be do something like VirtualGL.

Yep, that is fair enough.

I guess another option is to turn off the 3D support on the remote machine (eg, using the nv or nouveau driver instead of the nvidia one) so that your vnc connection has a better chance of working, but it is a less than satisfactory solution.

You should definitely try VirtualGL, as it's designed to solve precisely this problem. You have to have 3D acceleration on the server, but no 3D is required on the client.

(disclaimer: I developed VirtualGL, so I'm a little biased.)

1 members found this post helpful.

Hi, I have installed Nvidia card on Redhat 5.7 server and I am able to see the direct rendering enabled locally.

glxinfo | grep "direct rendering"
direct rendering: yes


However when I try to access it through VNC or Xmanager, the direct rendering doesn't seem to be enabled.

glxinfo | grep "direct rendering"
direct rendering: no

I am using the Nvidia Quadro K 600 1 GB DDR3 cards.


How do i use VirtualGl to resolve this?

Set up VirtualGL per the instructions in the VirtualGL User's Guide, then do

vglrun glxinfo

in your VNC session. If you don't use vglrun, then the OpenGL application will be trying to use the built-in OpenGL renderer in the VNC server (or whatever other X11 proxy you're using), which will be software rendering only. VirtualGL intercepts and re-routes the 3D rendering commands to the hardware-accelerated (DRI-enabled) display (usually :0).

If you need additional help, please post on the VirtualGL-Users mailing list (http://sourceforge.net/p/virtualgl/mailman/)

1 members found this post helpful.

Hey,

I tried running the command and getting this error.

[root@nvidia ~]# rpm -qa | grep Virtual
VirtualGL-2.3.90-20140717


[root@nvidia ~]# vglrun glxinfo
[VGL] ERROR: Could not open display :0.