Hardware For Sale/Trade

EBAY US - What's this? Infinite Performance

-----------------------------------------------------------------------
Hey Ho! Pip & Dandy!
:Octane2: :O2: :Indigo: :Indy:
-----------------------------------------------------------------------
From what I found online for it:

The Scalable Graphics Compositor (SGC) can recieve 2 or 4 DVI digital video inputs and combine them into a single video output with increased performance. Input is provided by 2 or 4 InfinitePerformance graphics pipelines (in V-Brick) that render only a part of the full screen.
Systems in use:
:Indigo2IMP: - Nitrogen : R10000 195MHz CPU, 384MB RAM, SolidIMPACT Graphics, 36GB 15k HDD & 300GB 10k HDD, 100Mb/s NIC, New/quiet fans, IRIX 6.5.22
:Fuel: - Lithium : R14000 600MHz CPU, 4GB RAM, V10 Graphics, 72GB 15k HDD & 300GB 10k HDD, 1Gb/s NIC, New/quiet fans, IRIX 6.5.30
Other system in storage: :O2: R5000 200MHz, 224MB RAM, 72GB 15k HDD, PSU fan mod, IRIX 6.5.30
So yes this conbines 2-4 VPro inputs and makes a 2-4kimage out of it.
:O3x02L: R16000 700MHz 8GB RAM kanna
:Octane: R12000 300MHz SI 896MB RAM yuuka
:Octane2: R12000A 400MHz V6 2.5GB RAM
:Indy: (Acclaim) R4600 133MHz XL Graphics 32MB RAM
:Indy: (Challenge S) R4600 133MHz (MIPS III Build Server)
Thinkpad W530 i7 3940XM 3GHz, 32GB, K1000M Windows 8.1 Embedded rin
Thinkpad R40 Pentium M 1.5GHz 2GB RAM kasha
Raion-Fox wrote: So yes this conbines 2-4 VPro inputs and makes a 2-4kimage out of it.

Nope, the output is a single link DVI so still limited to 1920x1200

IIRC it interleaves the pipes so you can get 4x the frame rate. Or maintain framerate with a more complex scene.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. ( IRIS 4D/70 Superworkstation Technical Report )
jan-jaap wrote:
Raion-Fox wrote: So yes this conbines 2-4 VPro inputs and makes a 2-4kimage out of it.

Nope, the output is a single link DVI so still limited to 1920x1200

IIRC it interleaves the pipes so you can get 4x the frame rate. Or maintain framerate with a more complex scene.

Can it AA the final output, rather than splitting the workload, or increasing the complexity?
Thanks for the info chaps, I was curious as to what it actually did, I get the idea now.
-----------------------------------------------------------------------
Hey Ho! Pip & Dandy!
:Octane2: :O2: :Indigo: :Indy:
-----------------------------------------------------------------------
spiroyster wrote: Can it AA the final output, rather than splitting the workload, or increasing the complexity?

It appears that each input is a tile of the output. From the SGI™ InfinitePerformance™: Scalable Graphics Compositor User’s Guide :
The scalable graphics compositor is a hardware graphics compositor that is capable of
receiving two or four DVI digital video inputs, and then combining them into a single
video output to increase graphics performance. Each input is from one SGI
InfinitePerformance graphics pipe (VPro V12 board) residing in a V-brick. The video
output can display on digital and analog monitors at the same time.

I was interested in combining two DVI output (a V12 w. DCD) into a single high resolution output, but the Compositor is not the answer to this problem because of it's single link DVI output.
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. ( IRIS 4D/70 Superworkstation Technical Report )
There's an IBM monitor that can do it, forget the name but Hamek uses one.

I misunderstood the purpose of the compositor, my bad.
:O3x02L: R16000 700MHz 8GB RAM kanna
:Octane: R12000 300MHz SI 896MB RAM yuuka
:Octane2: R12000A 400MHz V6 2.5GB RAM
:Indy: (Acclaim) R4600 133MHz XL Graphics 32MB RAM
:Indy: (Challenge S) R4600 133MHz (MIPS III Build Server)
Thinkpad W530 i7 3940XM 3GHz, 32GB, K1000M Windows 8.1 Embedded rin
Thinkpad R40 Pentium M 1.5GHz 2GB RAM kasha
Raion-Fox wrote: There's an IBM monitor that can do it, forget the name but Hamek uses one.

I misunderstood the purpose of the compositor, my bad.

The bottleneck is the single-link DVI (max is 1920x1200)?


Cool! It can!
Pixel Average
: This selection combines the inputs from the InfinitePerformance
pipes connected to the compositor and displays an average of those inputs.
(This selection works only if you have inputs from four InfinitePerformance
pipes.) This selection can be used for anti-aliasing (smoothing out of jagged
edges on items on a display)

AA can be achieved using the accumulation buffer, but would have to be implemented on a per application basis. This contraption would AA all inputs (including non GL contexts).

What an awesome bit of kit! shame the cogs and pistons on the other end of the 4 DVI inputs are so difficult and expensive to obtain and run.... unless you already have it of course ..0.o

Nice logo illumination too 8-)
lets pretend someone bought this thing - what can one currently run/do with this thing? Any software that already exists or would someone have to code their own stuff to use this? I'm still confused...
That's the problem with SGI's foray into broadcast media, they kicked out all this high end stuff over the period of about two years, and it could do just about anything (for a big big big price), but they had so little success finding customers for it (other than Autodesk, which used it with Smoke/Flame, and NBC used it for the Barcelona Olympics), and none of it was ever properly documented at techpubs, and now the end result is that it's all high-end abandonware... :x
Project:
Temporarily lost at sea...
Plan:
World domination! Or something...

:Tezro: :Octane2:
gijoe77 wrote: lets pretend someone bought this thing - what can one currently run/do with this thing? Any software that already exists or would someone have to code their own stuff to use this? I'm still confused...

Both, manual says there is a utility that allows you to customise the output of all the graphics pipes (user configurable tiling) as the user desires, or this could be done on an application basis since there appear to be some specific extensions for this hardware, so the developer could choose how they wish to output (which gfx device), multi-contexts all with complex geometry in a system operating in a non-distributed way (in terms of CPU and mem, I'm unsure if the vram can be shared between the different graphics pipes, or it must be replicated for each pipe?). Certainly, it is a novelty item, and you need something in the big-iron class (iirc the ATi's in the Onyx4 support this thing as well) pushing the pixels, but in terms of 'graphics related hardware', its pretty cool imo, while specialised. I don't I know of another breakout box which can perform this as a seperate part of the pipeline other than provided by the gfx, other systems rely on multi GPUs on the same board (or same discrete graphics subsystem) and then perform the mixing as part of its pipeline before outputting through the primary display. e.g Quantum3D AAlchemy.

The AA capability is neat, and the 'tiling' interesting. I don't think the AA would require a special code path for the GL program, since it can interleave all the inputs (which would duplicate all the geometry/texture memory, and jitter each output) for multiple pipes and mix at a hardware-level. Normally this would have to be done by rendering 4 frames (using the same gfx pipe) and then compositing via the accumulation buffer, or (if the pipes' texture buffer can handle it) rendering an image 4x the size and relying on the bilinear filtering when drawing to a downsized quad (not so easy on sgi's tbh).

@vishnu
This item is certainly applicable outside of broadcasting too since it can be used for higher-bandwidth visualisations and perhaps simulations which would make use of multi-context rendering. I think (I have seen) there are dedicated boxes with stuff like BNC ports and wotnot for the dedicated broadcast equivalent.... DMedia Pro breakout box which has a pretty similar livery. ??? :?
spiroyster wrote:
gijoe77 wrote: lets pretend someone bought this thing - what can one currently run/do with this thing? Any software that already exists or would someone have to code their own stuff to use this? I'm still confused...

Both, manual says there is a utility that allows you to customise the output of all the graphics pipes (user configurable tiling) as the user desires, or this could be done on an application basis since there appear to be some specific extensions for this hardware, so the developer could choose how they wish to output (which gfx device), multi-contexts all with complex geometry in a system operating in a non-distributed way (in terms of CPU and mem, I'm unsure if the vram can be shared between the different graphics pipes, or it must be replicated for each pipe?). Certainly, it is a novelty item, and you need something in the big-iron class (iirc the ATi's in the Onyx4 support this thing as well) pushing the pixels, but in terms of 'graphics related hardware', its pretty cool imo, while specialised. I don't I know of another breakout box which can perform this as a seperate part of the pipeline other than provided by the gfx, other systems rely on multi GPUs on the same board (or same discrete graphics subsystem) and then perform the mixing as part of its pipeline before outputting through the primary display. e.g Quantum3D AAlchemy.

The AA capability is neat, and the 'tiling' interesting. I don't think the AA would require a special code path for the GL program, since it can interleave all the inputs (which would duplicate all the geometry/texture memory, and jitter each output) for multiple pipes and mix at a hardware-level. Normally this would have to be done by rendering 4 frames (using the same gfx pipe) and then compositing via the accumulation buffer, or (if the pipes' texture buffer can handle it) rendering an image 4x the size and relying on the bilinear filtering when drawing to a downsized quad (not so easy on sgi's tbh).

@vishnu
This item is certainly applicable outside of broadcasting too since it can be used for higher-bandwidth visualisations and perhaps simulations which would make use of multi-context rendering. I think (I have seen) there are dedicated boxes with stuff like BNC ports and wotnot for the dedicated broadcast equivalent.... DMedia Pro breakout box which has a pretty similar livery. ??? :?


ok, pump the breaks. One needs a Onyx4 with ATI's graphics to use this? Not VPro or earlier Onyxs (I don't see any DVI output on the IR4 pics listed in the other thread)? I'm not aware of any Autodesk software working on the Onyx4/ATI platform... I was under the impression nothing really works on the Onyx4...

I looked into the DMediaPro for a while and it looks like the typical Discreet setup (with DM5 daughtercard) basically takes a part of the VPRO display and pushes it out either the SD-SDI or HD-SDI BNC port. I once found the command line commands to do that with anything on the display (you would have the specify what region of the screen you want to send to the SDI out port - I didn't really mess with it though) but I guess thats about as good as it gets. I suppose the TMDS version for IR3/IR4 graphics serves the same purpose.
gijoe77 wrote: I suppose the TMDS version for IR3/IR4 graphics serves the same purpose.

Correct. I have a DG5-TVO in my Onyx2, which connects to a VBOB w./ DM5 with a pair of DVI cables. This a HD-GVO (graphics to video) option. The system has a DM2 as well, which provides the regular (HD/SD) video I/O.

What it lacks is a Discreet Inferno installation, so it's not very useful :(
:PI: :Indigo: :Indigo: :Indy: :Indy: :Indy: :Indigo2: :Indigo2: :Indigo2IMP: :Octane: :Octane2: :O2: :O2+: Image :Fuel: :Tezro: :4D70G: :Skywriter: :PWRSeries: :Crimson: :ChallengeL: :Onyx: :O200: :Onyx2: :O3x02L:
To accentuate the special identity of the IRIS 4D/70, Silicon Graphics' designers selected a new color palette. The machine's coating blends dark grey, raspberry and beige colors into a pleasing harmony. ( IRIS 4D/70 Superworkstation Technical Report )
gijoe77 wrote: One needs a Onyx4 with ATI's graphics to use this? Not VPro or earlier Onyxs (I don't see any DVI output on the IR4 pics listed in the other thread)? I'm not aware of any Autodesk software working on the Onyx4/ATI platform... I was under the impression nothing really works on the Onyx4...

Apologies.. I meant I think it works with Onyx4 as well as V-brick, but not G-brick. Most stuff which doesn't use SGI specific extensions should work just fine on the Onyx4 (including Autodesk...Inventor/AutoCAD, however I don't have one so can't confirm), at the end of the day its nothing more than a GL implementation provided by ATi so in theory should work fine. We are still talking GL1.5 though, no shaders, the GLSL spec had been released for GL1.5 (GLSL was devised/promoted by ATi at Khronos), however wasn't fully introduced until GL2.0, by which time neither the FireGL nor V-Pro couldn't really handle anything non-standard outside the fixed function pipeline (there were specific SGI extensions for dealing with non-standard stuff which eventually found its way into GL standard, but a lot got superceeded by fragment functionality). I can't think of many CAD programs these days which require shaders tbh. They certainly use compute (GPGPU functionality) but not fragment/vertex shaders. CAD doesn't tend to push the 'looks nice' category, rather 'look how many vertices I can have on screen representing a bad ass mesh and get high FPS as possible' category.

Yes some features of this contraption, the applications had to explicitly implement however... as most things concived by SGI, it eventually made its way into the standard for those bespoke times when the application has more control over this aspect of the display (hyperpipes). Very bespoke though.

https://www.khronos.org/registry/OpenGL ... erpipe.txt

I think this was the first though :) . Not many other systems allowed this level of configurability for their graphics.