SGI: hinv

Onyx350 IP Dual-Head: 8x700MHz | 16GB | DM8+RAD | Compositor - Page 4

jpstewart wrote:
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.


Do frontporch/backporch pixels and overscan lines need to be considered with DVI? That would round up the bandwidth a fair bit...

_________________
:Onyx: (Maradona) :Octane: (DavidVilla) A1186 (Xavi) d800 (Pique) d820 (Neymar)
A1370 (Messi) dp43tf (Puyol) A1387 (Abidal) A1408 (Guardiola)
jpstewart wrote:
mia wrote:
* 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).

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.
It's actually a little fatter than 152MHz - here's the pertinent section of the format analysis run when it was compiled:
Code:
2@1920x2400_33_T221.vfo:
Total lines per frame:   4866
Total pixels per line:   2044
Active lines per frame:  2400
Active pixels per line:  1920
Pixel clock:             164.111 MHz, period = 6.09345 nsec
164.111MHz is bumping right up against the 165MHz ceiling for a single link connection - which is why I think hamei occasionally has issues locking the signal to to his T221 during a cold start.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
jan-jaap wrote:
OTOH, if you lower the refresh rate you bring down the bandwidth consumption and it might work.

The DG-3 has a max refresh of 41 hz. The DG-5 48 hz. You can feed them a 60 hz signal but internally the max they refresh the panel is 41 or 48 hz.

You can watch movies fine at 33 hz.

The SW1600, by the way, will sync at 50 hz (2 hz faster than 48 for arithmetic-challenged Amurricans) and works wonderful at that refresh. This whole thing with refresh rate is a hangover from crt physics.


I'm not up on IR but iirc, one possible hurdle was the way the second head in a deskside is more limited than the primary head ? It would be an awesome display for a deskside if one could get it to work though. Analog to digital converters are readily available that will do 1920x1200.

_________________
waiting for flight 1203 ...
hamei wrote:
I'm not up on IR, but iirc one possible hurdle was the way the second head in a deskside is more limited than the primary head ?
Regardless of the number of channels/heads available/in use, the maximum output bandwidth for a DG5 (used in IR2/3/4 graphics sets) is 300 Mpix/sec .

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
recondas wrote:
164.111MHz is bumping right up against the 165MHz ceiling for a single link connection - which is why I think hamei occasionally has issues locking the signal to to his T221 during a cold start.

Foodling with this some more (5 minutes or so on a daily basis, kinda) ... it doesn't seem to be a sync problem. This thing would sync to a tadpole, it is one smart monitor.

What I suspect is happening is that for some reason Xsgi is not locking in the correct startup setting. When it starts, it acts *exactly* like any of the dcd-equippped graphics cards do when you have setmon misconfigured to a 1@ res. I have a feeling that the dcd is feeding two single-monitor low-res default setups and I am seeing the backy-forthy that happens when you do that. Then the monitor is interpolating whatever resolution they send to fill the screen. It does that on its own. I could easily be seeing the default res of 1280 x 1024 or even 640 x 480 blown up to half-screen and shifting back and forth. Then I choose the 2@ res and it settles down.

Why it doesn't want to keep the 2@ setting is the mystery. Maybe it's time to experiment with setmon rather than xsetmon ...

recondas wrote:
Regardless of the number of channels/heads available/in use, the maximum output bandwidth for a DG5 (used in IR2/3/4 graphics sets) is 300 Mpix/sec .

What do you get for bandwidth for 4@1920x1200 @ 31hz ? 30 hz seems to be the limit factor for human eyesight. The T221 will sync to 25 hz if necessary but 31 would seem to be good. I can run the 31 hz vfo also, no visual difference to 33 hz for general use but haven't tested against movies.

If you want to knock out some more vfo's I don't mind to check them out to see what could be comfortable refresh rates.

T221 would be majestic on Infinite Reality :D

_________________
waiting for flight 1203 ...
recondas wrote:
It's actually a little fatter than 152MHz - here's the pertinent section of the format analysis run when it was compiled:

Thanks for sharing the details, recondas. It also nicely answered guardian452's question about porches. (Which I had completely forgotten to account for, obviously!)

But at least it's still within spec for single-link DVI.

_________________
:Indigo2IMP: :Octane: :Indigo: :O3x0:
Sun SPARCstation 20, Blade 2500
HP C8000
hamei wrote:
If you want to knock out some more vfo's I don't mind to check them out to see what could be comfortable refresh rates. I can run the 31 hz vfo also, no visual difference to 33 hz for general use but haven't tested against movies. What do you get for bandwidth for 4@1920x1200 @ 31hz ? 30 hz seems to be the limit factor for human eyesight. The T221 will sync to 25 hz if necessary but 31 would seem to be good.
The pixel clock in that 2@1920x2400_31_T221 vfo is 154.165MHz, so it might be worth seeing how it performs under more extensive testing.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
recondas wrote:
The pixel clock in that 2@1920x2400_31_T221 vfo is 154.165MHz, so it might be worth seeing how it performs under more extensive testing.

Aye aye, skipper ... coming to 31 hz now, sir.

_________________
waiting for flight 1203 ...
hamei wrote:
Aye aye, skipper ... coming to 31 hz now, sir.
If 31Hz works out, now that you've made some improvements on the font issue it might be worth taking another look at the 2@ 24Hz config :
A while back, hamei wrote:
With a single V12 and dcd and custom vfo of 2 @ 1920 x 2400 24 hz refresh, the monitor runs at full res. The refresh rate is not a problem. The cursor is so small you can't zoom around with the mouse anyhow. Move too fast and the damned thing disappears :) Movies are also fine with one exception (and this is a clue to how the dcd works) - if you display the movie across the center line of the screen, you can see where the two outputs from the dcd alternate. This is not apparent at all in any other operation and I'd bet that if you spread a movie over two monitors you wouldn't be aware of it because the bezels would mask that action. It's just that with the two sections directly adjacent, with motion across the full screen you can see the alternating refreshes.
If it doesn't look too non-functional/ugly, the pixel clock (for each 1920x2400 stripe) is only 119.4MHz. Analog-to-digital conversion issues aside, two channels at 1920x2400@24Hz might fit within the IR display generator bandwidth.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
interesting, on my list of things to test (after the ATI though).

_________________
:Onyx2:
recondas wrote:
If it doesn't look too non-functional/ugly, the pixel clock (for each 1920x2400 stripe) is only 119.4MHz. Analog-to-digital conversion issues aside, two channels at 1920x2400@24Hz might fit within the IR display generator bandwidth.

Don't quote me on this but I think there are some other intermediate frequencies the monitor will accept. 27 and 29 hz come to mind ... one of those should still be under 150 Mhz limit but be a little faster. Seems to me the magic number for human eyes is right about 30. Video refreshes at 29.x, doesn't it ? And film at 30 ?

What happens at 150 Mhz ? Does the compiler absolutely refuse to work or is it just "out of spec" ? I can't imagine that a pixel clock of 151Mhz would explode the cables ...

About edid, there's something non-standard ab0ut that, too. People make and sell edid-interposers that you put in the cables to feed low-bidder computers an edid they can deal with.

No problem with analog-to-digital converters, there are several available that will function at 1920 x 1200. Seems too bad to go digital to analog then back to digital again tho :(

Somone also figured out how to hack the hardware to make the monitor refresh faster than 48hz ... but given the sophistication of the equipment, that sounds to me like some pimple-faced teenager improving Tony Stewart's engines.

_________________
waiting for flight 1203 ...
Film at 24 but the traditional cinema projector has a 3-sprocket shutter of sorts that flashes the same frame 3 times. So 72 Hz, which is a bit of a misconception on lower refresh == "film-like" looks.

_________________
:Onyx: (Maradona) :Octane: (DavidVilla) A1186 (Xavi) d800 (Pique) d820 (Neymar)
A1370 (Messi) dp43tf (Puyol) A1387 (Abidal) A1408 (Guardiola)
NTSC video is indeed 29.97fps.

_________________
smit happens.

:Fuel: bigred , 700MHz R16K, 4GB RAM, V12, 6.5.30
:Indy: indy , 150MHz R4400SC, 256MB RAM, XL24, 6.5.10
probably posted from Image bruce , 2x2x2.5GHz PowerPC 970MP, 8GB RAM, Mac OS X 10.4.11
plus IBM POWER6 p520 * Apple Network Server 500 * HP C8000 * BeBox * Solbourne S3000 * Commodore 128 * many more...
hamei wrote:
No problem with analog-to-digital converters, there are several available that will function at 1920 x 1200. Seems too bad to go digital to analog then back to digital again tho :(
Somone also figured out how to hack the hardware to make the monitor refresh faster than 48hz ... but given the sophistication of the equipment, that sounds to me like some pimple-faced teenager improving Tony Stewart's engines.


* I haven't tried, but it seems to me that going to analog would obviously introduce some noise, and degrade the signal itself.
* I've researched the frequencies higher than 48hz and, unless I was told wrong information, the upgrade is a simple eeprom flash, but this also comes with additional heat and implies a modification of the power supply. The gain is minimal, except for some edge applications (gaming, perhaps?).

_________________
:Onyx2:
recondas wrote:
If it doesn't look too non-functional/ugly, the pixel clock (for each 1920x2400 stripe) is only 119.4MHz. Analog-to-digital conversion issues aside, two channels at 1920x2400@24Hz might fit within the IR display generator bandwidth.

31 is okay. It might even be better - the cursor is more controllable. I'll have to use the spaceball a little at this refresh - at 33 it was mostly good but occasionally would blow up. Video is okay but might be a little fuzzier ? Or that might just be the old video ... Peter Albin looks funny. His guitar is almost as big as he is :D
Attachment:
not_beyonce.jpg
not_beyonce.jpg [ 82.7 KiB | Viewed 174 times ]


Four inputs might be better than two. We might be pushing the envelope at two :(


mia wrote:
* I haven't tried, but it seems to me that going to analog would obviously introduce some noise, and degrade the signal itself.

Yeah, but IR is analog, no choice there. Unless one could hack into the circuitry upstream of the d-to-a section ... :D

Quote:
* I've researched the frequencies higher than 48hz and, unless I was told wrong information, the upgrade is a simple eeprom flash ...

Except Cirthix is an idiot and the IBM engineers were not. This thing is not a toy and the people who designed it knew what they were doing. I could go into Tony Stewart's heads and make them ports real big, too. And then he'd kill me when the car couldn't get out of its own way.

Between IBM and Cirthix, I'll take Door Number One, please ...

btw, I could be mistaken but take a look ...

http://www.deskography.org/people/oPi18 ... hotos/203/

recondas : here's two modelines I found for XFree :

# 1920x2400 @ 25Hz for IBM T221, VS VP2290 and compatible, hsync: 60.8kHz
Modeline "1920x2400" 124.62 1920 1928 1980 2048 2400 2401 2403 2434 +hsync +vsync
# 1920x2400 @ 30Hz for IBM T221, VS VP2290 and compatible, hsync: 73.0kHz
Modeline "1920x2400" 149.25 1920 1928 1982 2044 2400 2402 2404 2434 +hsync +vsync

looks like the second one would be 1.5 mpix under the max capability for an IR (this glass is half-full) ... and you have an IR ... ?? Will your panels accept analog input ? Probably won't sink at 30 hz tho, hunh :(

If you want to make another vfo I can give it a test.

_________________
waiting for flight 1203 ...
hamei wrote:
recondas : here's two modelines I found for XFree :

# 1920x2400 @ 25Hz for IBM T221, VS VP2290 and compatible, hsync: 60.8kHz
Modeline "1920x2400" 124.62 1920 1928 1980 2048 2400 2401 2403 2434 +hsync +vsync
# 1920x2400 @ 30Hz for IBM T221, VS VP2290 and compatible, hsync: 73.0kHz
Modeline "1920x2400" 149.25 1920 1928 1982 2044 2400 2402 2404 2434 +hsync +vsync
As is, those are a no-go with VFC and the VPro definition files - both modelines generate crash-n-burn errors.
hamei wrote:
you have an IR ... ?? Will your panels accept analog input ? Probably won't sink at 30 hz tho, hunh :(
Analog input wouldn't be a problem, but regardless of the refresh rate I don't have anything that'll display 1920x2400.
hamei wrote:
If you want to make another vfo I can give it a test.
Sent you another @1920x2400_30, built using another of the SGI-provided modelines in the Onyx4 Porting Guide. That one comes in at 149.192MHz:
Code:
2@1920x2400_30.vfo:
Active lines per frame:  2400
Active pixels per line:  1920
Pixel clock:             149.192 MHz, period = 6.70279 nsec
so if building against the IR definition files doesn't escalate the pixel clock, it ought to be possible to squeeze two of 'em in (just) under the DG bandwidth ceiling.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
recondas wrote:
Analog input wouldn't be a problem, but regardless of the refresh rate I don't have anything that'll display 1920x2400.

Two or four monitors at 1900x1200 ? how many outputs them IR3's got ? I remember with eight you can only go to 640x480 but were those specs thinking about low refresh rates ?

Quote:
Sent you another @1920x2400_30, built using another of the SGI-provided modelines in the Onyx4 Porting Guide. That one comes in at 149.192MHz:

So far so good ... ran it all day. Maybe I'll leave it here, save some wear and tear on the oscillators.

Quote:
so if building against the IR definition files doesn't escalate the pixel clock, it ought to be possible to squeeze two of 'em in (just) under the DG bandwidth ceiling.

If you can get your resolution up by getting the refrsh down and the analog-to-digital doesn't care about the frequency, might can do. Would be cool. Would definitely make a good-looking pair with a deskside Onyx2 :D

Oh look ! J-J - that's where I got the idea a Compositor could only run 1920 x 1080. Run < sgcombine > and check out the default setups they give you ... File -> Open -> a lot of 1920 x 1154 (I know, 60 pixels. But I was specifically looking for 1920 x 1200 ...)

Here's a new one for the hinv collectors :)
Code:
urchin 67% ./gfxtopology
pipe  0 g @ 001c01/IX:12

input   k @ 001c01/IX:15/pci/0/2b/usb/1 m @ 001c01/IX:15/pci/0/2a/usb/1

_________________
waiting for flight 1203 ...
hamei wrote:
If you can get your resolution up by getting the refrsh down and the analog-to-digital doesn't care about the frequency, might can do. Would be cool. Would definitely make a good-looking pair with a deskside Onyx2 :D
Ok - if somebody is willing to be the IR/T221 analog-to-digital crash test dummy, I've built a 1920x2400 format using the IR board and chip definitions. This one has a 24Hz refresh rate/119.402 pixel clock to stay well within the DG bandwidth (if it works we can try bumping the refresh right up to the bleeding edge):
Code:
IR-1920x2400_24.vfo:
Total lines per frame:   2434
Total pixels per line:   2044
Active lines per frame:  2400
Active pixels per line:  1920
Frames per second:       24
Fields per frame:        1
Swaps per frame:         1
Pixel clock:             119.402 MHz, period = 8.37505 nsec
Hardware pixel rounding:  every 2 pixels
Line analysis:
Length:                 2044 Pixels, 1 Lines, 17.1186 usec; (line 0)
Frequency:              58.416 KHz, period = 17.1186 usec
Horizontal Sync:         52 Pixels, 435.502 nsec; (line 32)
Horizontal Back Porch:   64 Pixels, 536.003 nsec; (line 32)
Horizontal Active:       1920 Pixels, 16.0801 usec; (line 32)
Horizontal Front Porch:  7.99997 Pixels, 67.0001 nsec; (line 32)
Field Information:
Field Duration:           4.9751e+06 Pixels, 2434 Lines, 41.6667 msec; (line 0)
Vertical Sync:            8176 Pixels, 4 Lines, 68.4744 usec; (line 0)
Vertical Sync Pulse:      8228 Pixels, 4.02544 Lines, 68.9099 usec; (line 0)
Vertical Back Porch:      57232 Pixels, 28 Lines, 479.321 usec; (line 4)
Vertical Active:          4.9056e+06 Pixels, 2400 Lines, 41.0846 msec; (line 32)
Vertical Front Porch:     4088 Pixels, 2 Lines, 34.2372 usec; (line 2432)
If anyone wants to try it everything you need to build your own format is in the build analysis text - or drop me a PM and I'll send you a vfo.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************
Hm, could you point me to a reliable analog to digital converter?
I'd be surprised if that worked though, what's your confidence that this might?

_________________
:Onyx2:
mia wrote:
Hm, could you point me to a reliable analog to digital converter?
:D While I fully intend to watch, I ain't got a dog (or T221) to put in the fight.

_________________
***********************************************************************
Welcome to ARMLand - 0/0x0d00
running...(sherwood-root 0607201829)
* InfiniteReality/Reality Software, IRIX 6.5 Release *
***********************************************************************