SGI: hinv

Why not?

165MHz output ( single link DVI ) should be capable of 1920x1200 ?

You're right, it should. I be wrong. But the docs I read ages ago said 1920 x 1080. Maybe they thought no one would be interested in them outdated old resolutions ? Or just a typing error on the internet ?

So okay, here's another possibility - 16 O350's with a single V10 in each driving four Compositors to drive one T221 or a setup like Recondas has.

Could you fit all that in one rack or would you need two ?

Haven't purchased a free-standing DCD in quite some time, so I'm probably not the most up-to-date source for that info.
As hamei has already mentioned, if your usuage will be focused on non-3D stuff, the the rough road that's been described for 3D apps with Onyx4/ATI graphics will probably be fairly smooth. Aside from the you-already-have-everything-needed/nothing-to-purchase/free issues hamei mentioned, you could also have all four DVI ports (2x FireGL AGP boards) into a single O350 G2N and leave the master O350 headless.

No router means a max of two linked O350 modules, but that shouldn't present an issue. Two linked O350 modules will work nicely with the T221 and Onyx350IP/Odyssey/V12 or Onyx4/Voyager/ATI graphics. With Odyssey graphics each module will house a V12 graphics pipe (preferably with a DCD each, though you could get by without the DCDs and run two 1920x2400 stripes like hamei is currently doing); with Voyager graphics all four heads would be in one module (you'll still need a second module for system I/O- no IO9 in the AGP-only G2N).

To my knowledge, there aren't any publicly available photos of the interior of a factory Onyx350 IP (and especially both sides of the graphic boards and the cooling fan arrangement). Next time you're over there *please* consider taking your camera.

As hamei mentioned, there aren't any, but if you installed the vidtogfx example programs in the optional DM10/ML dev stuff you may have enough to conduct some basic tests.

Might you have been thinking of IR graphics? Two of the 24"/1920x1200 SGI CRTs (attached to the same DG) exceed IR pixel clock bandwidth, but 2@1920x1080 will fit.

Kind of like ircombine with IR graphics, the Compositor uses sgcombine (rather than setmon/xsetmon) to load configure and load graphics 'combinations'. The default install of sgcombine includes a couple of predefined 1920x1200 combination files.

I'm a little surprised, it's just like an origin 350 with a Vpro, just like the rackmount tezro. Next time I open either, I'll snap some picts. Do I own a rare beast?

I've reviewed the "Silicon Graphics UltimateVision Graphics Porting Guide" which has relevant T221 information:
http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi/hdwr/bks/SGI_EndUser/books/Onyx4_Port/sgi_html/ch05.html#LE11137-PARENT

I don't have the pretension to understand all the Modeline options, but the config file shows:


12.7Hz isn't the kind of resolution I am hoping for, do you think that's all the UltimateVision can drive?

No, that's what you can get over one cable. If you go to two cables you can run at 33 hz refresh. At four cables, 48hz refresh. The same max resolution for all refresh rates. The limit is the bandwidth of a single-link dvi connection.

I'm not aware of the specs on the FireGL cards ... if they will do dual-link then you can use your converter box and one single-link plus one dual-link connection to get the 48 hz refresh. In our case tho, four single-link cables is probably easier. Matrox sells an lfh60 - to - 2 dvi cable reasonably cheap. In China, where they are made, they are way-cheap. You guys get so fleeced by the finance "industry". What a con.

I believe you can go to 60hz refresh externally but internally, the panel (DG-5) refreshes at 48 hz no matter what, so no point. DG-3 refreshes at 41 hz.

For all the people going "oh yuck", you can't tell. No matter what your spec-minded brain says, for human beings there is no difference between 33 hz refresh and 60 hz. I've used the same monitor at both, the biggest difference is that at the lower refresh and higher res you have to slow down or your eyes lose the cursor. I've got to biggify that thing ...

You can also use lower resolutions, such as 1600 x 1200. I don't believe you get all the screen real estate that way though, in multi-linked mode. With a single connection you do. That monitor will interpolate to almost any normal resolution.

When the monitor arrives, you can check it easily by feeding it a 1920 x 1200 @ 60 signal over connector #1. (They are labelled). Windows or Irix will both work. Probaly Linux also but I didn't try that. It will also sync to 640 x480, 1024 x 768, 1280 x 1024, and 1600 x 1200 all at 60hz over a single connector.

In Irix, there were problems with the vfc compiler creating a 3840 x 2400 @ 13hz vfo for a single connection. 2@ 25, 31, and 33 work okay. So for a quick-n-dirty check when you first get it, do a 1@ resolution over a single cable. Then maybe try a 2@ with one of the vfo files which the wreck so kindly created ?

Then go to the G2N and a different x server and you're on your own, but we'll be listening for updates

Okay, I get it now.

So first things first, Big thanks to Recondas and Hamei for providing a tremendous help here.



Right, 4 links is probably easier for me. There's not much difference between 3 and 4 links, either will take 2 cards.



This is what I have gathered from the forums.



I took the liberty to cut you off there, yes I do perfectly understand that.



Most certainly, I will have to explore all this. Hopefully with your help.



Does this mean you have not been able to get it to work with the DCD?



Of course, didn't Squeen manage to get this working with the T221?


Cheers!

I have read that they only devised that peculiar setup because there were several popular graphics cards made with one dual link and one single link connection at that time. Nowadays that feature is not very useful.


If you guys would just light up the torches and storm the castle then I could shut up, you know. Revolution has to come eventually, there's no point in everyone suffering for a hundred years before y'all finally grow the balls to do something about it. FDR saved the banksters' lives the first time around but there's no FDR now ... those idots are looking for a date with Mme Guillotine. Unmitigated greed is an amazing thing, isn't it ?


No, sorry, I meant for testing. I've been running with the dcd for several months now. But I would expect you'd want to make sure the monitor works right away so you could send it back if it's broken. I pulled the dcd and connected to the single socket on the VPro. Everything should work fine. If you plug in the number one cable and set the computer to 1920x1200@60, you should get a fullscreen desktop. You can try other lower resolutions also, they should all work.

Maybe easier to do that on a Linux or Windows computer but I don't have one

Then add the dcd and if you are running Xsgi, plug cable # two into the second socket. You should get two desktops side by side on the one monitor. Mr Recondas' excellent 2@1920x2400_33hz vfc files can then be selected and you should get a double-wide single desktop.

Better have the magnifying glass handy, the fonts will now be real small

If it doesn't go, don't freak out. Just try a couple times. Some times it syncs up all by itself on boot, sometimes on the first "load" from xsetmon, sometimes when the "load" doesn't take and I click "cancel" it syncs*, very rarely it takes a second full go-through. But then it is happy until I shut down so I'm not complaining.

* "sync" is probably not the correct word, the monitor displays the desktops fine but they are flashing left-right, left-right the same as when you try to view a single monitor on the center socket when using a dcd.


Squeen had a 221 and didn't tell us ? that rat ! He should be a trove of information then. Let's storm his castle on the way to the banksters' place !

The devil is always in the details - and there aren't many details available for the Onyx350 InfinitePerformance. Since the V12 used in the Onyx350 InfinitePerformance has a unique part number (030-1909-00x) and fits into a much more restrictive space (Fuel/Tezro boards overlap the O350's processor cooling shroud), it would be interesting to see:

• if it shares the same form factor as the Fuel and Tezro VPros;
• the O350 VPro heat sink (Fuel/Tezro VPro heat sinks foul the O350 processor cooling shroud);
• the fan and air flow ducting used to cool the O350 VPro;
• the mounting hardware.

Thanks! It'd be much appreciated As far as I'm concerned you do - it'd rank right up there with its doppelganger, the rackmounted Tezro.

I haven't noticed any internal difference between my rackmounted Onyx350 and rackmounted Tezro, except the faceplate. There might be some differences I have failed to notice; pictures will tell as soon as I get to it.

I was reading some old posts (2011, 2012) whereas Recondas and Hamei had managed to drive (or did I misread) the V12+DCD to two channels of 1920x2400 (hacking vfo files). And I still wonder which solution would work best for me (this doesn't exclude trying a few options); those options are:

* V12+DCD (Fuel or Tezro deskside) and two channels of 1920x2400 for a resulting picture of 3840x2400.
* g2-brick on an Origin/Onyx/Tezro rackmount.
* Onyx 350IR + Tezro rackmount "numalinked" (each having a DCD).

Thank you, I'm weighting my options, and curious to understand the differences.

I like to delude myself into thinking that working out the (completely undocumented) method of creating 2@ DCD compatible formats was at least a slightly more elegant than a hack job. It does work, but with a single V12/DCD you risk the occasional fail-to-sync-without-rerunning-xsetmon issues hamei described (earlier in this thread). If you have two Onyx350/V12 systems they might provide a more stable solution, even if neither has a DCD (though two V12s, two DCDs would be optimal).
Possible with an Origin350, but if the Onyx350 and Tezro have internal graphics, Odyssey/V12/Xsgi and Voyager/ATI/xFree86 aren't compatible in the same system. Combining a RMWS Tezro with an Onyx4 G2N brick will also cause serial number incompatibilities (see below for more info).
SGI implemented a L1 serial number security routine in IP35 generation hardware (excluding the Fuel). Origin350/Onyx350/Onyx4 systems have serial number string prefaced with an "M", Tezro serial numbers are prefaced with a "P" (and for the record, O3k systems with an "L"). Because of the differing serial number prefixes, the L1 serial number security routine in would prevent one or both from starting if NUMAlinked.

One possible solution would be to convert the RMWS Tezro into an Onyx350 InfinitePerformance system with the L1 command "make rmws 0". The process is described in the following thread, but please note it carries the potential for risk. Even if we set aside issues regarding the rarity of a RMWS Tezro, if you don't have an L2 controller as a recovery device I'd strongly recommend against trying the make rmws procedure. viewtopic.php?f=3&t=16721235

And recondas did all the work. All I did was load the file and connect the cables ...



If you've already got this ready in hardware, it's trivial to connect the two cables and load the vfo file.

I would suggest doing this first because it is so easy and it is possible that you won't like the display. For graphic work I think it's great but for text mode, damn, you best have some good eyes !


Ja, but as a first step it's trivially simple. From a self-serving standpoint, it would be interesting to see if you get the same problem or if that is some glitch in my hardware. And there is no guarantee there won't be the same problem with two dcd's, either. As wreck pointed out, the video format compiler bitches when you feed it a y number larger than 2400. Four stripes of 960 x 2400 is still bigger than 2400 .. oh yeah, you could do four tiles at 1920 x 1200. That would tell us whether the video compiler has the glitch or the dcd's ... go for it


This would be interesting from the standpoint of

You already have one, they are rare, it would be more compact, you should be able to run the four channels with a single 2U box which would be physically nice. Your description of your work is not-mainly-OpenGL-ish so the alleged weakness of the Infinte Disaster graphics might not be a drawback.

On the other hand, it's a step into the unknown.



Oh. An O350 plus a Tezro. I thought it was two O350's. Serial number problems

How come my O350 lmhostid starts with a "d", btw ?

I would like to do the four channels because in some applications (mostly Pro/E for me) you can see the lack of color depth. If you didn't have a comparison you'd never notice, and running terminals and text editors I don't imagine you would ever see a difference, but a side-by-side comparison of low-memory and high-memory screens shows that high memory is better. What a surprise

You would also get a slightly higher refresh. (33 to 48 hz step up.) But that's not something you can see with the T221 because it always refreshes the screen at 48hz. And you can't type or move the mouse faster than 33 hz anyhow. The whole "refresh" thing on a T221 is complex to talk about because the screen update is not the same as the "refresh rate." And the mouse is so tiny that you can't move it fast or your eyes will lose it so the benfit of a "fast refresh" is just not there.

Note to those who want to open-source Irix : hardware hacking the V12 to 512 megs of memory would bring the same (or a better) result. Maybe start with a simple project ?


So you don't really need four channels and you don't really need a faster refresh for typing and the panel refreshes at 48 hz anyhow and maybe if you just plug in the two cables and load the vfo you'll be plenty happy and stop right there. And a Fuel is smaller, quieter, and just about as fast as the O350 ... You might consider a 900 mhz Fool with the low-noise fan hacks. No 75 foot cables, repeaters, noise and distance hassles, just stick it under your desk. I bet it would suit your needs admirably.

Well, I guess I'll have to try either solution. But Hamei's comments are sound, the simplicity of a Fuel/Tezro is hard to ignore, but I'm not too worried about running 100ft dvi cables either, this has the advantage of being able to keep the hardware in a colder and well ventilated machine room.

As far as changing the Tezro's identification, it shouldn't really be a problem, and I also have a L2 (somewhere) with (thanks to Recondas) the right power adapter.

It's a little sad though that a compositor wouldn't drive high resolutions such as the 3840x2400 this monitor offers.

Stay tuned for more.

Here are my concerns:

* One thing I fear about is that the signal "quality" will be lower on a V12+DCD (please note: single V12) because when pushed at an extreme resolution (1920x2400) it might be a little much per single link dvi cable (please educate me here).

* As such, I think the Quad FireGL configuration, hooked up to an Origin (no need for the extra V12 overhead here), might provide me with a crisper image, because each channel (and here, there's 4 of them, instead of 2 above) pushes a lower throughput through each pipe? Also, the extra cooling offered by the g2-brick (compared to V12) might make this system more durable (perhaps).

* I'd rather not run 1xTezro+1xOnyx350IR (although I will try that and report my results) because of the power drawn, which, in this case will be close to 600W, which is quite a bit for a "workstation".

* Another alternative would be to run 4xPsitech rad4 cards, but they're analog, so the signal would have to be converted somehow. I think I won't bother testing this; there's probably a high loss in signal quality here.

* In a few days, I'll probably ask which 100ft DVI cables to buy, should I decide to stay with the G2-brick+Origin configuration (most likely will), this isn't relevant for testing as I'll simply move the display closer to the rack for now. This also makes me realize that I'll also need 2x100ft usb cables for keyboard and mouse (or only one and a usb hub perhaps).

On a side note, I wish my deskside Onyx2 could drive the T221.

Since you're up for an adventure anyhow, I'd look into this. I know, in general the IR graphics can't go to four channels of 1920 x 1200. However ... I have seen hints here and there that with the latest, largest memory cards it might be theoretically possible.

If so, four analog-to-digital adapters will cost less than all the cabling and repeaters you are going to get into.

If you get that running there will be several very jealous people out here

The single-link DVI specification supports a maximum pixel clock of 165MHz. Recondas' 1920x2400 resolution used a low 33Hz refresh rate. Multiplying 1920x2400 pixels times 33Hz gives 152064000 pixles/second, or about at 152MHz of bandwidth required. So it's actually well within the spec for single-link DVI.

In other words, the unusually high resolution is nicely offset by the unusually low refresh rate.

I doubt that. The problem is not the amount of memory on the raster manager but the bandwidth of the pixel bus to the DG5.

OTOH, if you lower the refresh rate you bring down the bandwidth consumption and it might work.