[GameBreaker64]

I haven't gotten the chance to play around with a real SGI machine before, so I thought I'd try one out via remote access before investing in one of my own. How do I set up Xming in Windows 10 to connect to IRIX computers that are located in different states, countries, etc.?

[Raion]

I don't have specific instructions for you but I'm assuming that you have permission from someone to do that? If they aren't on your local network the latency is going to be terrible I really would suggest trying vnc instead. Nekoware has VNC.

Trying to use remote X over a open internet even with SSH tunneling is rather obtuse and inefficient. especially considering the crappy upload speeds that people in the United States or Canada usually have as a result of asymmetric broadband speeds.

[GameBreaker64]

Yes, I have permission from some users in the Silicon Graphics User Group Discord server. I was recommended VcXServ and PuTTY. I configured PuTTY to enable X11 Forwarding and set X display location to localhost:0.0. Whenever I get the chance to SSH into a machine, they said they'd help me out further.

I don't really care about speed. As long as all the programs work properly, I'm satisfied. MAME still has a long way to go from what I can decipher.

[Raion]

Well, no experience using GL over X (any of your graphics software is gonna use that) but I presume it can work since SGI pioneered that. But remote X is painful even on a LAN with 100mbit dedicated links between the client and server. You're really gonna have a bad time. Lemme repeat my suggestion of VNC, you'll have a better time with that.

MAME works with IRIX on the Indy, but the speed is glacial. You're barely, on a brand new AMD or Intel CPU, gonna reach a 50MHz R4000PC or so.

[weblacky]

OpenGL is client-side in Xserver (locally rendered). This is part of the standard and I proved this in front of a college lecture hall, putting a professor in his place on Linux using OpenGL gears. I created an SSH X redirect tunnel from the prof’s laptop (server) to my Dell precision laptop running Debian linux with Nvidia drivers (circa 2004). His laptop, running the program locally made about 37 fps, mine running his gears application remotely on my mobile Quadro did like 3760 FPS at the same window size (like 800x600 window size) over Ethernet! The larger the window was resized the slower my fps went.

That’s also why the older OpenGL standard has a client concept when issuing commands. That means that back in the day, using like a CAD or 3D Modeling application, you still needed a hot-sh*t local graphics setup, even if you ran the app from the central server for data security and stability (central server has UPS, local Indigo2 doesn’t - kind of office setup).

Still runs like champ because OpenGL is always Client -side rendered. It also means this all doesn’t normally work in most Windows XServer implementations because they don’t support the OpenGL commands via X. So apps crash or come up with blank windows. On MacOS and Linux I believe the Xserver projects people load do support this (I don’t know about recent MacOS since the removal of MacOS-provided OpenGL support in favor of METAL a few years ago.). I thought I was told MacOS OpenGL native support was pulled a year or two ago but Vulcan support can be added?

Linux should just work (even with software-only OpenGL acceleration).

Most Of Windows X implementations don’t even mention a yes/no support of OpenGL, so you’ll have to just try it. But since most don’t support it, that’s the reason VNC grew huge in UNIX community, it just copies the locally rendered server screen so all the client-side extensions are required for X.

Realistically it was a great idea for a hybrid terminal server feature back in the day. You could run a headless server with graphical clients. But still be running server apps, on the server, with a RAID storage and all that. The local stations only needed as much graphics acceleration as was required for each worker’s job. And since the server didn’t need to run a real graphical X server, it was very efficient to run many remote users at once with only raw CPU speed required.

But of course as Raion mentions, only over a local (trusted, low-latency) LAN as an integral solution for very expensive computers of the day while putting money toward CPU and storage in server and minimal ram and storage but huge graphics on clients.

Today, this really would only need to be done if you’re trying to do something like Windows RDP with remote FX but for UNIX, which means multimedia but you have to run it server-side for some reason?

It’s a scenario that as prices fell, just wasn’t needed anymore given the complexity and program-by-program finickiness of being X remoted.

[jpstewart]

Linux should just work (even with software-only OpenGL acceleration). 

It depends on the application.  An easy test is gr_osview.  It won't run on a remote Linux box.  You'll get errors about X_GLXCreateContext failing.  When gr_osview won't run, many other graphical applications won't either.

[weblacky]

Linux should just work (even with software-only OpenGL acceleration). 

It depends on the application.  An easy test is gr_osview.  It won't run on a remote Linux box.  You'll get errors about X_GLXCreateContext failing.  When gr_osview won't run, many other graphical applications won't either.

Apparently this is due to implementation details/differences.  So you may need to install mesa on Irix to maintain compatibility across platforms?  

https://praveen-palanisamy.github.io/blo...s-Remotely

[Raion]

There's no Mesa on IRIX.

[weblacky]

There's no Mesa on IRIX.

Well than that solves it, same implementation to same implementation required.

[Raion]

Mesa is a full on OpenGL implementation. Such a thing would require from scratch writing.