SGI: Hardware

So I bought this saturday a nice 1600SW+Multi Link Adapter from Nekochan member bplaa.yai (thanks a lot for the delivery at home !). I then fiddled the whole WE to make the combo work with V12/DCD equipped Octane. Not very straightforward but it kinda worked in the end, thanks to the invaluable infos shared in here. Curiously, the various 1600x1024 vfos found in my install didn't work at all. I finally used the 1600x1024_60-1704TPPL-1056TLPF.vfo file (to be found here : http://forums.nekochan.net/viewtopic.php?f=3&t=16725775 ).

Still, there remains a minor annoyance that I couldn't solve. When the Octane boots up with the 1600SW+MLA plugged in one of the DVI slot, the above mentionned 1600x1024 vfo disappears from the list in xsetmon and some default 1280x1024 is automatically selected. At this point, there is no way to get back to 1600x1024. When I boot up the machine without any monitor connected until the graphic login screen, and then hotplug the DVI, the vfo is back in the list and I can select it properly and make it the default choice. That means : every time I reboot or start up, I need to unplug first the DVI (so that IRIX can't detect the 1600SW+MLA contraption) and plug it later to get the native 1600x1024 or else I'm back at 1280x1024.

It is workable as it is but kinda annoying. Is there's a way to avoid that ?

This is really cool!! Strange though, that Octane2+DCD+1600 SW isn't working right out of the box, since this was once upon a time a possible configuration from SGI(?)

Geoman wrote: This is really cool!! Strange though, that Octane2+DCD+1600 SW isn't working right out of the box, since this was once upon a time a possible configuration from SGI(?)

octane2_2.jpg

I'm running 6.5.22 at the moment. Maybe I'm missing some needed patch to make the whole thing smoother ?

Perhaps - you could check supportfolio for any DCD/1600 SW related patches (if any they would be $...)

Or someone could find a way to prevent IRIX from autodetecting the MLA - time to dig deep inside IRIX' configuration files

Maybe the dcd isn't supported, although I can't recall if the analog input on the MLA is even SOG.

I use the digital input of the MLA so Sync-On-Green is not the problem...

Geoman wrote: This is really cool!! Strange though, that Octane2+DCD+1600 SW isn't working right out of the box, since this was once upon a time a possible configuration from SGI(?)

It's funny you would choose that photo, since there was quite a bit of laughter on the newsgroups about it at the time. When they put that ad out, it was not possible. I'm a little surprised JL could get it working at all, almost no one else has been able to. Maybe it was the dcd that did the trick ? Or disconnecting the monitor during bootup ....

hamei wrote:

Geoman wrote: This is really cool!! Strange though, that Octane2+DCD+1600 SW isn't working right out of the box, since this was once upon a time a possible configuration from SGI(?)

It's funny you would choose that photo, since there was quite a bit of laughter on the newsgroups about it at the time. When they put that ad out, it was not possible. I'm a little surprised JL could get it working at all, almost no one else has been able to. Maybe it was the dcd that did the trick ? Or disconnecting the monitor during bootup ....

Therefore the "(?)" at the end of my sentence ^^ I found that TIFF recently while consolidating my stuff to a NAS and think it's remarkable, since I newer seen such a setup in real life -- now I know why

In the year 2000 or 2001 when they had this published, I have not been into SGI newsgroups that much; #IRIX and #SGI were my main source of information concerning SGI. (greetings to Duck, Joerg, Opi, Rhoenie, Schleusel et al )

EDIT: So my brain just came up with this: I think around 2005 Schleusel pointed out to me, that a 1600sw+Octane DCD was a beautiful advertising image, but really didn't work. So Kudos to the OP and Nekochaners, who finally came up with a working vfo!

I confirm it does work with the aforementioned vfo. I was still tinkering with the setup (looks exactly like the pic above) this morning before going to work. But I'd really like to make it work without unplugging the the DVI each and every time...

Jack Luminous wrote: Curiously, the various 1600x1024 vfos found in my install didn't work at all.

It's likely the timing in those vfos was intended for the 24" CRTs SGI sold with Octanes.

Jack Luminous wrote: I finally used the 1600x1024_60-1704TPPL-1056TLPF.vfo file (to be found here : http://forums.nekochan.net/viewtopic.php?f=3&t=16725775 ).

Been a while since I made that one, but I'm pretty sure it was created from a random xFree86 1600SW modeline captured from the internet.

Jack Luminous wrote: Still, there remains a minor annoyance that I couldn't solve. When the Octane boots up with the 1600SW+MLA plugged in one of the DVI slot, the above mentionned 1600x1024 vfo disappears from the list in xsetmon and some default 1280x1024 is automatically selected.

As mentioned in your other Octane/monitor thread IRIX looks at the DDC or EDID info from the monitor and presents only those graphics formats it deems appropriate. Be nice if that process was fully documented / could be updated.... (we'll leave that pipe dream for another time).

In the mean time you might try connecting the DVI port of the 1600SW/MLA to another system to see if you can capture the EDID info and try for a 1600SW/MLA tailored VFO using the EDID template .

Jack Luminous wrote: At this point, there is no way to get back to 1600x1024. When I boot up the machine without any monitor connected until the graphic login screen, and then hotplug the DVI, the vfo is back in the list and I can select it properly and make it the default choice. That means : every time I reboot or start up, I need to unplug first the DVI (so that IRIX can't detect the 1600SW+MLA contraption) and plug it later to get the native 1600x1024 or else I'm back at 1280x1024. It is workable as it is but kinda annoying. Is there's a way to avoid that ?

Probably not the elegant workaround you were hoping for, but instead of hot-plugging the DVI cable have you tried disconnecting the MLA power supply before booting (to see if that stops the exchange of DDC/EDID info)?

recondas wrote: In the mean time you might try connecting the DVI port of the 1600SW/MLA to another system to see if you can capture the EDID info and try for a 1600SW/MLA tailored VFO using the EDID template .

I don't have a DVI equipped computer at home apart from the Octane. Though I googled some EDID info about 1600SW+MLA and found this :

<00ffffff ffffff00 4ce90016 55a1aec6 04030103 00100a00 0e000000 00000000 00000021 0800a900 81400101 01010101 01010101 0101302a 40686000 20401030 1300a820 64000000 302a0068 50002040 c8701300 a8206400 00000000 00fc0053 4749204d 756c7469 6c696e6b 000000fd 003c463c 4619000a 20202020 202000b9>

Which translate to :



It looks like it's the same values you used for the previous vfo : Horizontal Image Size : 1704 mm Vertical Image Size : 1056 mm. But I'm not sure.

recondas wrote: Probably not the elegant workaround you were hoping for, but instead of hot-plugging the DVI cable have you tried disconnecting the MLA power supply before booting (to see if that stops the exchange of DDC/EDID info)?

I tried and apparently the EDID is read as long as the DVI cable is plugged, even when the MLA is not powered. I haven't tried to disconnect the 1600SW itself. I'll try tonight.

Jack Luminous wrote: I don't have a DVI equipped computer at home apart from the Octane. Though I googled some EDID info about 1600SW+MLA and found this

Nice find, though it looks like that info might have been obtained when the MLA was switched to analog mode:

Jack Luminous wrote: It looks like it's the same values you used for the previous vfo : Horizontal Image Size : 1704 mm Vertical Image Size : 1056 mm.

That'd fit with the EDID mention of "video input definition: analog". Been a while, but ISTR the xFree modeline I used (to build the vfo you're currently) using was for an MLA in analog mode. I'll have sort through some archived data and see if I can find the vfoinfo file from that format. (EDIT: found some additional info in this post - looks like I manually modified/swagged an xFree modeline to use the 1704 TotalPixelsPerLine and 1056 TotalLinesPerFrame values mentioned in the MLA manual - hence the inclusion of "1704TPPL-1056TLPF" in the name assigned to that format).

Using EDID info from an MLA in digital mode would be the next step. Hopefully that would resolve the issue with IRIX rejecting the MLA's offer of 1600x1024 at boot, but until it's tested we won't know for certain. The DDC/EDID info presented by the MLA/1600SW still might fall outside of the range accepted by IRIX for an Octane with VPro graphics. Since the 1600SW/MLA does work/boot when attached to an O2, the inconsistency might be attributable to code that was inserted into the O2 PROM to support the MLA/1600SW EDID info ( the 1600SW got added for specifically for the *O2* beginning with the IRIX 6.5.2 update ) but didn't make it into the Octane PROM. If that's the case, then you may still see the occasional Octane/MLA glitch ( scroll down to the highlighted line ).

Jack Luminous wrote: I tried and apparently the EDID is read as long as the DVI cable is plugged, even when the MLA is not powered. I haven't tried to disconnect the 1600SW itself. I'll try tonight.

Let us know - might be helpful to someone else down the line.

Attached a format built using the EDID info you found for an MLA in analog mode. It's pretty close to the timing I swagged into the 1600x1024_60-1704TPPL-1056TLPF.vfo. The EDID generated version does provide slightly wider horizontal and vertical back porches (8 pixels and 4 lines respectively).

Here's the VFC -generated format analysis:


If the idea of hot-swapping *two* 1600SWs appeals to you I also found the 2@/DCD version of 1600x1024_60-1704TPPL-1056TLPF.vfo (mentioned in that older thread).

EDIT: While poking through my VFC archives, I also ran across a couple of VFO's I built for a 1600x1024 Apple Cinema Display (the old school ADC version). One has a 60Hz refresh rate, the other 50Hz. The versions I archived use the same 1704 TotalPixelsPerLine and 1056 TotalLinesPerFrame called for by the 1600SW, so if all else fails (and you're feeling lucky) let me know and I'll send you copies You could also try the ones ZoontF attached to the thread (the TPPL and TLPF in those were adjusted to something other than 1704TPPL/1056TLPF to suit ZoontF's Apple ADC monitor).

I'm glad it found a good SGI enthusiast home, and that it's living on the edge of unsupported SGI configuration

bplaa.yai wrote: I'm glad it found a good SGI enthusiast home, and that it's living on the edge of unsupported SGI configuration

LOL Thanks a lot for your saturday delivery !

So I had 10 minutes this morning before going to work for further testing. here are the results :

- the DVI must be physically disconnected to avoid the EDID reading. Even if both the MLA+1600SW are both shut off, the VPRO card still detects everything and reverts back to "authorized" vfos list.

- I tried with the VPro_1600x1024_60-Analog_MLA.vfo posted above and it gives the same result as the other one : it disappears from the list if the DVI is connected at startup. But it works fine with the hot-plugging scenario.

To be continued...