SGI: Hardware

Hello all,

I just picked up a 1600sw display with a multilink adapter for my Octane2 system with v12 graphics. I'm using a proper 13w3 to VGA adapter and the vga to dvi cable that comes with the Multilink adapter. I'm running on 1280x1024_60 resolution and the display is perfectly fine. I can see the 1600X1024_60 resolution option in the display control panel, but it doesn't look at all proper. It uses the same space as the 1280 resolution, but all the pixels are squished. If I try the alternate timing mode on the multilink adapter, the display takes up the whole screen, but the timing is very wrong and everything has vertical lines through it. I tried adjusting the display timing modes and such through the multilink adapter, but nothing I do seems to solve the problem. I know that there's a way to create a custom display resolution using the VFC software, but I don't have the slightest idea how to do it. I'm sure if I spent the time reading I might be able to figure it out, but I was hoping somebody has already created one that works with this configuration. I've already done many searches on this forum and on Google, but all I see are people in my situation with no resolutions, or links to a proper file/settings for this configuration.

Any ideas??

3dchris wrote: I'm sure if I spent the time reading I might be able to figure it out, but I was hoping somebody has already created one that works with this configuration. I've already done many searches on this forum and on Google, but all I see are people in my situation with no resolutions, or links to a proper file/settings for this configuration.

I don't recall seeing one either, so if no one responds with a pre-made vfo, you might have try your hand at creating your own. There have been several reports of success using the VFC via the BlockSync method. I took a look at the BlockSync examples in the VFC Manual on TechPubs and tried the BlockSync template to create a VPro-specific 1600x1024_60 format. It produced a vfo file on a V12 equipped Tezro, which I unfortunately can't test because I don't have a monitor that will display 1600x1024. if you'd like to try it on your Octane, here's the command I used (run it as root and enter the entire thing as a single line):
If it also compiles on your Octane it should produce a file named "1600x1024_60-1600SW.vfo"; copy the resulting vfo into "/usr/gfx/ucode/ODSY/vof" (or run the command from within /usr/gfx/ucode/ODSY/vof).

Looks like 1600x1024_60-1600SW.vfo won't be dynamic. That means you'll have to restart the Xserver to load it, so if you aren't already familiar with setmon and single-user mode some basic knowledge of either might be useful if the new format doesn't produce a display.

If that doesn't work, then you might want to try using a modeline specific to the 1600SW to create a format that's targeted a little more directly at the 1600SW, though I'm not certain what effect, if any, the analog to digital conversion performed by the MLA might have on the process. If it might prove helpful, I ran across a set of 1600SW modelines that inludes one modeline for use with an 1600SW/MLA combination (which at least suggests the inclusion of an MLA might require a specifically tweaked vfo file).

If you'd like to try the modeline method, the info in this post by rooprob looks like it might be very helpful: viewtopic.php?f=10&t=15616&p=7329782&#p7329782

Thanks, Recondas for the display vfo info. I can compile it on my Octane, so that's not a problem. I also tried it out, but it's giving me a very similar effect compared to the display option in the setmon by default. One difference is that my multilink can display the video widescreen when set to widescreen. The default vfo file displays the same way whether or not the multilink is set to widescreen or standard mode. It only displays widescreen when the multilink is set to alternate timing mode.

It looks like I will have to somehow create a custom file based on the multilink, but I have no experience at all with this. Any help would be appreciated

3dchris wrote: I also tried it out, but it's giving me a very similar effect compared to the display option in the setmon by default. One difference is that my multilink can display the video widescreen when set to widescreen. The default vfo file displays the same way whether or not the multilink is set to widescreen or standard mode. It only displays widescreen when the multilink is set to alternate timing mode.

It's possible the presence of the MLA will necessitate a custom vfo. The 1600SW modelines I linked above include one for the MLA that shows opposite values for HsSync and Vsync:

3dchris wrote: It looks like I will have to somehow create a custom file based on the multilink, but I have no experience at all with this. Any help would be appreciated

Ditto on the no experience, but if you get the chance, take a look at rooprob's write up - it looks like it could be very helpful.

I used to use a Fuel V12/V10 with the MLA, and that worked well enough with the stock .vfo. I think there's an option somewhere in the menu that needs to be set / adjusted to get a proper 16:10 picture. I'm mostly using a Pixelink box now, which is working flawlessly for digital (DVI) signals.

I did something that my computer doesn't like and now I don't get a display when it boots into the login screen. I can see the boot sequence, but now I get an unsupported resolution message on the monitor . I tried to find the command to get it back to the 1280x1024_60 file that I was using before, and I thought I was doing it properly in the command monitor at bootup, but obviously I'm doing something wrong. What's the correct command to get it back to the default resolution?

Also, Canavan, the Fuel is using a DVI connection, but my Octane is using a 13w3->VGA adapter and a VGA->DVI cable to connect to the Multilink. I'm sure this is what's throwing it off because it's an analogue signal, not a digital one.

Thanks!

There's a template (.vfs file) posted on line that was written to take advantage of the analog connection on the 1600SW Multi-Link Adapter. The catch is it was intended for use with an O2 rather than a VPro-equipped Octane (though it doesn't contain anything that looks glaringly O2-specific). For additional details and background be sure to see the 'read me' that's included as a comment inside the .vfs file: http://dart.ncsa.uiuc.edu/slevy/o2-ster ... 1600sw.vfs ).

I took a shot at compiling it using the VPro chip and board definition files, and it compiled with only the single non-fatal error (at line 652) that all VPro template files seem to generate: No 1600SW or MLA here, so the resulting format is completely untested (not to mention having been originally intended for O2 graphics). As a check I did run the resulting format with the "-a" option in VFC, which performs an an analysis of the video format and outputs the results as ascii text or a postscript file: Here's a copy of the analysis:

If you're willing to try the otherwise untested format I've attached a copy of the one I compiled.

I'd suggest using xsetmon to examine the compiled format before you load it - xsetmon can be run from a command line or from Toolchest > System > Display Properties. As long as you don't press the "Load" button, you can open multiple xsetmon windows. Open one and click once on the new format (in the left panel), then do the same (in a second instance of xsetmon) with the SGI supplied 1600x1024_60 so you can compare the two if you like - before you attempt to use the 1600x1024_60-1600SW-MLA.vfo.

If the format leaves you with no display, the following should get you back to 1280x1024_60:

Press and release the power button, your Octane will cleanly shut itself down;

Once it's completely power down, restart it.

As it boots watch for the "press escape" prompt and enter the PROM

Select option "5" from the PROM menu to access the PROM command line;

type the word "single" (w/o the quotes) and press enter.

Enter the root account/password info when prompted, and run execute "/usr/gfx/setmon -x 1280x1024_60"

reboot.

Thanks for all the detailed information, Recondas! I compiled everything on my system, but the strange this is that the MLA is showing the display as a 1280x1024 image. The image looks similar to the standard 1280x1024_60 vfo file. That is how it appears on the login screen. When I logged in though, it seemed like IRIX was thinking that it's indeed a wider screen because the toolchest is off the screen as though it was in the top left corner of a wider monitor. Also, my desktop icons are off the screen as well. It's as though the system is displaying 1600x1024, but the graphics is only showing 1280x1024. Any ideas what might cause this?

3dchris wrote: I compiled everything on my system, but the strange this is that the MLA is showing the display as a 1280x1024 image. The image looks similar to the standard 1280x1024_60 vfo file. That is how it appears on the login screen. When I logged in though, it seemed like IRIX was thinking that it's indeed a wider screen because the toolchest is off the screen as though it was in the top left corner of a wider monitor. Also, my desktop icons are off the screen as well. It's as though the system is displaying 1600x1024, but the graphics is only showing 1280x1024. Any ideas what might cause this?

Considering that it was originally intended for an O2, it is encouraging to see that it will compile and produce a viewable display.

With the newly compiled vfo running, what resolution does the Multi-Link Adapter's On-Screeen Display indicate, and (not to belabor the obvious) have you had the opportunity to try adjusting the display with the adjustments in the MLA's Advanced Display Menu (or the "Force Detect" option in the MLA's general menu)?

This usenet post uses provides another template/.vfs file with slightly different timing values. The post also makes specific suggestions to tweak the MLA's Advanced Display settings to match the resulting format (phase to 13, clock divider to 125). So you may be able to adjust the phase and clock divider with the format you've already created or even compile another format with the values provided in the usenet post.

EDIT: I went ahead and compiled a format using the template provided by Avi Bercovich in the 2002 usenet post referenced in the previous paragraph - there's a copy attached to this post. It complied with only the usual line 652 error.

If you'd like to contrast the differences between Ari's format and the first one (from Stuart Levy's site), I attached a screen shot of both analysis files in gdiff (I named Ari's format 1600x1024_60-Analog_MLA to avoid overwriting the first):
It might also be worth taking a look at the cable and/or adapter that connects the 13W3 port on your V12 to the VGA port on the MLA to make sure that's up to spec. The MLA may need to receive DDC info , which probably won't happen if you don't have a properly wired SGI 13W3 cable or adapter. See Amigo's post in this topic for confirmation of a working V10-to-Analog-MLA configuration - particularly his mention that the "VPro series actually has H/Vsync out on the 13W3 AND DDC2 pin output, which allows MLA to configure itself to compatible hardware".

I tried the new vfo file you attached and the display to me looks almost exactly the same as the previously compiled vfo file. I'm still getting missing pixel information on the left and right sides of the screen as though it's only showing 1280x1024. The MLA is still showing the display as being 1280x1024 and not 1600x1024.

With the newly compiled vfo running, what resolution does the Multi-Link Adapter's On-Screeen Display indicate, and (not to belabor the obvious) have you had the opportunity to try adjusting the display with the adjustments in the MLA's Advanced Display Menu (or the "Force Detect" option in the MLA's general menu)?

I have already tried that on each iteration of the vfo files. I've tried the force detect, wide/zoom, overlapping modes, reseting the monitor, adjusting phase, clock etc...

What I'm seeing with the last two files appear to have a much darker, distorted horizontal band of graphics at the top portion of the screen. Maybe 50 pixels tall or so.

On your previous post, you linked to the vfs page that was written. I took another look at the readme at the beginning and the poster says "Anyway, enough of my troubles. The 1600x1024_60p.vfo only
works if you have the flat panel adapter AFAIK, and the timings are
set appropriately for digital out (no front porch!). This 1600x1024
works with the analog output and sets the timings appropriate for
analog out." What is this Front Porch? I can see some entries on both of those files for horizontal and vertical front porch, maybe that's what they are referring to? Maybe that's something that needs to be changed?

Another interesting thing I noticed when fiddling with the MLA adjustments is the overlapping mode. In the default mode, I get a "normal" looking picture, but the sides of the screen are black. In the Alternate mode I can actually see most of the image, but the colors are all messed up and the display is showing the left halve of the screen on the right and vice versa. You posted much earlier a link that stated " These lines are required for use with the SGI Multilink Adapter and the SiliconGraphics Flat Panel display.
Modeline "1600x1024g" 108.0 1600 1616 1656 1704 1024 1027 1030 1056 -Hsync -Vsync
Option "OverridePolarity" "1""
I wonder if the syncs need to be reversed and this OverridePolarity needs to be flagged.

I think we are getting close to solving this mystery. I'm sure I'm not the only one out there that has this sort of a setup. Once we get this sorted, we should definitely include the vfo file into the Wiki about the MLA and the 1600sw.

One last thing, the 13w3 adapter I'm using was purchased from a site that specifically makes these for different models of SGI's and SUN computers. They have a computer selector/configuration for the different types of 13w3 adapters and mine was specifically for SGI's and Octanes. I don't think it's a bad adapter because I've used a few other monitors with the computer (analogue and digital) and never had any problem before. If it turns out that the adapter is to blame, then do you know where to get a proper one?

Thanks again for all your help with my crazy setup!

3dchris wrote: IWhat is this Front Porch?

http://techpubs.sgi.com/library/tpl/cgi ... /ch03.html

3dchris wrote: Another interesting thing I noticed when fiddling with the MLA adjustments is the overlapping mode. In the default mode, I get a "normal" looking picture, but the sides of the screen are black. In the Alternate mode I can actually see most of the image, but the colors are all messed up and the display is showing the left halve of the screen on the right and vice versa. You posted much earlier a link that stated " These lines are required for use with the SGI Multilink Adapter and the SiliconGraphics Flat Panel display. Modeline "1600x1024g" 108.0 1600 1616 1656 1704 1024 1027 1030 1056 -Hsync -Vsync Option "OverridePolarity" "1"" I wonder if the syncs need to be reversed and this OverridePolarity needs to be flagged.

I made another following the xorg modeline for the analog MLA as closely as possible (I have yet to find s method to explicitly specify -Hsync -Vsync). Here's the VFC analysis... .... the resulting vfo file is attached.

3dchris wrote: I'm sure I'm not the only one out there that has this sort of a setup.

Actually, you may be. I remember quite a few posts in the newsgroups from people trying to get the 1600sw to work on an Octane. If you succeed at this, you may be the first.

It's kind of funny because there is quite a bit of marketing literature which shows an Octane with a 1600 SW. The implication is that you can do that, but in Real Life no one was ever able to make it work.

I gave this another shot. It's better than the previous one in the sence that there is much less of a distortion at the top of the screen. Almost none. I am almost able to get the quality of the picture to be as close to the 1280x1024_60 by fiddling with the MLA controls. What I'm still seeing is that the picture is displaying as 1280x1024 still. The MLA is detecting that as the resolution. Were you able to get the Override Polarity option set? Maybe that and the -syncs will solve the problem?

We're close. I can feel it. So very very close.....

3dchris wrote: What I'm still seeing is that the picture is displaying as 1280x1024 still. The MLA is detecting that as the resolution.

You've probably been there, done that, but I'll ask just in case.... The manual mentions the MLA's firmware has an overlapping modes feature that can cause a 1600x1024 input to be displayed as 1280x1024:

Silicon Graphics MultiLink Owner's Guide wrote: Overlapped Modes—Chooses the correct resolution for display in special situations. Certain combinations of timing modes are indecipherable to the MultiLink. In such cases, you may need to override the MultiLink's settings through the Overlapped Modes menu item. Consult Table A-8 for a listing of the affected timing modes.

3dchris wrote: Were you able to get the Override Polarity option set? Maybe that and the -syncs will solve the problem?

You'll have to remember that list of 1600SW modelines was acquired from and intended for Linux/Xorg, and so was the mentioned 'Override Polarity option'. If there's an equivalent setting for VFC I haven't discovered it yet.

3dchris wrote: We're close. I can feel it. So very very close.....

The MLA manual also includes a mention (in table A-7) that when correctly set in 'SGI Operating mode' with a resolution of 1600x1024_60, that mode will have a total horizontal x vertical setting of 1704 x 1056 (where 1704 represents the TotalPixelsPerLine and 1056 TotalLinesPerFrame values used by VFC). The intended-for-the-analog-MLA-port vfs examples obtained through web searches that we've tried so far have come close, but none was exactly 1704 x 1056.

I can try changing the vfs formula to the 1704 TotalPixelsPerLine and 1056 TotalLinesPerFrame later this evening, or, since it's very close, you're welcome modify the last attached vfs and try it yourself (changing either of those values will require recalculating the values inserted in to the vfs template for horizontal and vertical back porches). It'd be nice to be able to test while working on specific formats, but unfortunately I don't have a 1600SW or an MLA.

EDIT: I'm back and have attached a .vfo with the aforementioned total horizontal x vertical setting of 1704 x 1056. If you haven't done so already, I'd suggest checking the 'Overlapping Modes' menu setting in the MLA OSD first.

Thanks for the response. I'm going to try this vfo again tonight.

Regarding the overlapping modes in the MLA, I've already tried adjusting them from default to alternate for every vfo that we've tried so far. I can see more pixels on the screen when it's set to alternate mode, but the image is REALLY messed up looking with the right half of the image on the left half of the screen and vice versa. Also the colours are all messed up in this mode, like they are missing one of either R,G or B. Probably they sync is messed up in this mode.

I'll let you know when I give it another try tonight.