[kuban]

Octane MXE is supported for flame v9, but I haven’t seen it anywhere running on octane1.

[kuban]

Xavier wrote:

I would guess Flinux uses FibreChannel disk arrays, probably identical to the ones found on Flame and Lustre.

But I don’t know for sure. Can anyone confirm?
— Xavier

I looked to installation guide for Flinux. It uses etiher QLA2342 or QLA2344. These are both 2Gb fiber SCSI controllers. 2342 is 2 port, and 2344 is 4 port scsi controller. They both use the PCI-X 133Mhz bus. And this bus should be capable of doing more then 1Gbyte transfers per second. So the controller and bus type seems to be ok for HDTV reads from SCSI. If you have a disk array, which is fast enough. The flinux should read enough data. But one question remains, is the memory to gfx bus fast enough for HD playback?

If you ask me, even my homePC should be capable of 8bit HDTV playback from RAM to screen. Anybody tried this on a flinux?

[kuban]

If that VTR is compatible with RS422 Sony protocol, 1 to 1, 9pin cable won’t work with the octane. You have to connect different pins on both ends. But don’t remember the pin layout for octane now.

[kuban]

Jonas wrote:

I don’t know why our test system don’t run 1920×1080 – it actually does a short amount and then drop frames – Maybe it’s possible when the disk is total defragtmented or with another aray – You might be able to do it – but I don’t know if discreet will promise it to run HD.

Pilsjar wrote:

Jonas wrote:

I think there is a playback from ram – that would defiently play a short sequence of HD.

Hmm, strange, the flinux I have seen could play back HD in real time. It even played realtime from disk when zoomed out.
The differences with Flame/Inferno is 3 software options, namely Modular Keyer, Motion and 3D tracking. Hardware maxes out at 8bit. All the rest is in there. At one third of the flame/tezro pricetag, it becomes very tempting. If you can live/work with the limitations, and throw the rest of your money to burn, it can actually get pretty wild imho.

I would check my SCSI controller bus type. For playing HD, you will need a fast system bus. Maybe a PCI-X SCSI controller will be fast enough.

And anyone knows if current version of flint on linux supports FX3400G? It is the PCI-X version of FX3000G.

[kuban]

patdawg wrote:

I’m not sure that I believe that the Linux smoke box can beat the Tezro…I’d have to see it right in front of me to believe it, but either way you can’t play 10-bit 2K clips in realtime on the Linux box.

Yes, you are right. That is a software limit, which will be be probably demlimited in newer versions.

When it comes to believing, V12 gfx is equal to Quadro2 gfx by Nvidia. And Nvidia FX3000G should be the fastest Quadro4. And it has 256MB of memory for gfx. But V12 has only 128MB of memory. The only problem is system bus on Linux bus. But if nvidia FX3400G is also supported, you can also forget about bandwidth problems.

[kuban]

Flint on linux! The linux boxes might be even faster than tezro in near future. It might be even faster now, for texture fill operations. Current flint uses nvidia FX3000G(AGP card), and it is a fast board. I hope it will support FX3400G too(PCI-X card), and it will have a faster memorygfx bus in both directions.

Yes guys, with the introduction of PCI-X bus on PCs, the bandwidth is now great. And discreet started to support linux. Lets see what happens in near future.

[kuban]

-k wrote:

…but does Inferno actually make use of more than 4 CPUs? I heard different opinions on this one.
I think the offical discreet statement is no. Not sure though.
-k

The old days, onyx RE2 had to tile textures to 512 x 512 pixels. That was a hardware limitation. And discreet chose to process that tiles with different processors. For keying process, color correction, etc… And if you work on PAL res, 720 x 576 pixels, there was no meaning of having more than 4 processors. But if you work on higher resolutions, more processors start to help. So, if you work higher than PAL resolution, even only in action, you might see lots of acceleration with more than 4 CPUs.

But if you ask me, the main power of onyx is, the IR4 gfx subsystem. That gfx has 1024MB of texture memory. And the V12 has only 104MB of texture memory. And it has to swap between texture and system memory. And this might decelearate Tezro a little bit. If you work on SD resolution, you have to put lots of layers for going above 104MB texture, but if you work on 2K, you don’t have many layers.

So you have to go for the batch way. I don’t know how discreet optimizes texture memory transfers in batch, but if you have lots of small action setups connected to eachother in batch. Tezro won’t be a lot slower than onyx.

But lets come to a tougher point! To the texture fill rate. With 4 raster managers, how many pixels will IR4 fill per second? Probably 10 times more than the V12. If I don’t remember wrong, the V12 should do around 500Mpixels/sec. And this not a great texture fill. For one sample processes, the CPU processing power might be very important, and this low texture fill speed won’t be too important. But what if you prcess 16 antialiasing and 20 samples of mo-blur? Then the OpenGL has to process 320 samples per frame!

Let’s think that our batch setup needs 1 second to process the CPU tasks per frame. But it needs to process 10 secons of texture fill on the tezro. The total rendering time will be 11seconds/frame. But if we have the same CPUs on the onyx, and have a 10 times faster texture fill, the onyx will process the same setup 2seconds/frame. Which will be 5.5times faster in total. And I have used 16 samples many times, and moblur is widely used anyway. So that texture fill speed is important.

In concluesion, if you ask me, the highest value of an inferno system is: the name itself. 😯 But onyx has a faster gfx system, and it’s a bit more expensive.

[kuban]

Cmistudios started selling the DVD’s on 16th April.

[kuban]

I might say a few things about v6 inferno:

– import photoshop layers into action
– paint node within batch

and some more good stuff, but don’t remember them all 🙂

[kuban]

>>> Foetz
SGI seems to give the bandwidth numbers as a total. Especially on the onyx series. Onyx series are multibus machines. Multibus seems to affect compositing speed a little, but mainly the communication speed between system memory, and texture memory is very important. So about the onyx2, we might talk about XIO bus speed. Yes, XIO is faster then PCI-X. But how expensive is a PCI-X bus?

Also on onyx series, SGI likes to tell, all the added raster managers, gfx systems, etc… And add them, and tell a max. texture fill rate. Which is a gigantic texture fill. But none of the onyx owners configure such an onyx.

[kuban]

Xavier, I see we share the same ideas about discreet/sgi/PC

I totally agree, that the PCI-X is a very fast bus architecture. But I also want to add that AGP bus has some problems fro OpenGL rendering. The 8x bus can write 2.1GB/sec, but can not read with this speed. So you can preview your opengl gfx on your monitor very fast. But when it comes to reading that gfx back to system memory, AGP isn’t optimized for that. But SGI hardware is bidirectional. There are lots of cables going from video to gfx, from gfx to video. Even the OnyxRE2 could have live video textures. And remember, sirius was a seperate board. Even at very old times, 12 yrs ago. SGI hardware could do that.

According to INTEL:
AGP 8x technology is intended to be the last parallel interface step that meets the industry’s requirements before transitioning to a PCI Express-based serial graphics solution in 2004. The PCI Express architecture is a high-speed, general-purpose, serial I/O interconnect that provides a unifying standard, consolidating a number of I/O interconnects within a platform.

So the PCI-X seems to be the future, also for gfx.

[kuban]

Officially every permanent license from discreet comes with a dongle. But are the systems working? I mean does flame execute, when you run it?

If it is running, you probably have a temporary license. Sometime, some machines have serial port problems with dongle, and those work with 1 year license from discreet. By chance that flame can be one of those.

I guess you have to talk to discreet, to understand what licenses you have.

[kuban]

Also if your foreground image is premultiplied with blck, you can use “add” mode to composite the foreground image. This way, no black edge problem will happen, but transparency won’t work in layers.

[kuban]

I also never thought that MR would be used with MIPS processors. Since MR is very good for network rendering, burn* is a very good product to run MR. So we will just tell action to render using MR. That would be a very nice option.

And also shaders, I also want to see how to edit shader within action. That would be very interesting. And something like IPR would also be nice, to fine tune the shading parameters.

[kuban]

A much more painful, but more accurate way would be like this:

Scale the animation 2 times. And deinterlace your source clips. Now render the animation in frames. Then interlace the progressive rendered frames. This way, you might keep the fileds in video images.

hope it works,

[Author]

Posts