My dream Red product: The Redcode Card

[I Bloom]

How much would you pay:

It's likely that they already thought about this, but I just want to put it out there.

If the case is today that most computers aren't able to decompress Redcode quiet fast enough for realtime work. (Is this the case? Post your benchmarks http://www.redhax.net/wiki/Quicktime_Proxies#Playback_Benchmarks.)

Then maybe an interim solution until processors catch up is a card with some FPGA's that can be added to a PC or MacPro. The computer sends Redcode to the card, the card spits out Uncompressed RAW or Debayered RGB, or scaled RGB, for use in a NLE or Grading software. It might even be possible that the only programs on your system that need to interface with the Redcode card are the quicktime codec and RedCine.

Instant realtime workflow. Am I dreaming?

(Put it in the Expresscard slot on my MacbookPro, now that is the dream!)

IBloom

[Stephen Webb]

I was thinking a Red 4K PCIe card that both accelerates realtime performance and also runs the new Red 4K projector :sorcerer:

[Tony Lorentzen]

I think we'll see something like this from AJA - maybe even at NAB next year.

[Jeff Kilgroe]

I'm kinda thinking that too. It would be an interface board like one of the AJA cards that can drive a 4K display and will provide real-time benefits for REDCODE.

Yeah, I bet we'll see something like this whenever we see those 4K displays and projectors.

[MikeCurtis]

At NAB they discussed AJA as a tech partner - Kona Red or somesuch?

[I Bloom]

I'm kinda thinking that too. It would be an interface board like one of the AJA cards that can drive a 4K display and will provide real-time benefits for REDCODE.

Yeah, I bet we'll see something like this whenever we see those 4K displays and projectors.

Right, it seems though that a Red 4K projector or 4K monitor should have Redcode decompression and metadata interpretation built in, as well as RGB and other types of input. That would cut down on the bandwidth needed to communicate with the display.

But a standalone Redcode decompression tool, a card or a box, could allow Recode to easily render in your Canvas window on FCP and/or speed up render times all around would also be useful as well.

IBloom

[Lexicon]

This is definitely a possibility. AMD is incorporating a design into their architecture that allows for adding FPGA processors and other things (of any kind) via the PCI-E bus that will directly link into the main CPU and share resources. The current PhysX accelerators use a similar approach but it's a bit crude compared to what is on the drawing boards right now.

[Warren Kommers]

I actually emailed AJA about this and they replied that they could not start such work until FCP supported REDCODE. So perhaps sometime after that they can begin work. I wish Apple would give them something ahead of time so we didn't have to wait so long. I also hope the card could just display 1080p and 2k scaled versions of the 4k original. I mean even when the FCP support of REDCODE happens we will still only be able to use a 2k proxy right?

And is REDCODE really that hard on the processors at 27mb/s? That's nothing really. I guess it's the format.

[Kevin Halverson]

How much would you pay:

It's likely that they already thought about this, but I just want to put it out there.

If the case is today that most computers aren't able to decompress Redcode quiet fast enough for realtime work. (Is this the case? Post your benchmarks http://www.redhax.net/wiki/Quicktime_Proxies#Playback_Benchmarks.)

Then maybe an interim solution until processors catch up is a card with some FPGA's that can be added to a PC or MacPro. The computer sends Redcode to the card, the card spits out Uncompressed RAW or Debayered RGB, or scaled RGB, for use in a NLE or Grading software. It might even be possible that the only programs on your system that need to interface with the Redcode card are the quicktime codec and RedCine.

Instant realtime workflow. Am I dreaming?

(Put it in the Expresscard slot on my MacbookPro, now that is the dream!)

IBloom

There are several bottle necks and a hardware solution to handle the decompression address but one of them.

In your example the output of this piece of hardware would be "uncompressed RAW or DeBayered RGB" sent to the NLE. This still burdens the NLE with having to handle high rate data.

A better path would be for the NLE to process the most compressed data possible and then let the new hardware handle the conversion to a usable format. This would off load more of the processing burden from the host computer. The difficult part of this concept is that some if not most of the video effects would have to be handled by this piece of hardware as the NLE's native tools would no longer work. This would require a high level of integration between the NLE and the added hardware.

This is not a trivial problem at all and involves host, application and workflow issues.

[I Bloom]

I'm kinda thinking that too. It would be an interface board like one of the AJA cards that can drive a 4K display and will provide real-time benefits for REDCODE.

Yeah, I bet we'll see something like this whenever we see those 4K displays and projectors.


There are several bottle necks and a hardware solution to handle the decompression address but one of them.

In your example the output of this piece of hardware would be "uncompressed RAW or DeBayered RGB" sent to the NLE. This still burdens the NLE with having to handle high rate data.

A better path would be for the NLE to process the most compressed data possible and then let the new hardware handle the conversion to a usable format. This would off load more of the processing burden from the host computer. The difficult part of this concept is that some if not most of the video effects would have to be handled by this piece of hardware as the NLE's native tools would no longer work. This would require a high level of integration between the NLE and the added hardware.

This is not a trivial problem at all and involves host, application and workflow issues.

Actually, the card would only be used when the NLE is trying to display to a screen, or rendering to a useable format. Allowing us to cut larger proxies basically, but still keeping the data compressed as we cut of course. It's just a performance boost any time you need to decompress. So when I say "sent to the NLE" I mean via memory, while playback is going on, i.e. doing what the quicktime component does now in software instead with hardware. i.e the quicktime component or RedCine detects the card, it uses it. Plug and play.

[Sanjin Jukic]

Let it be

REDCODE

PCI Express Card.

[Gavin Greenwalt]

For those who haven't heard it has already been released:

http://www.bit-tech.net/hardware/2007/05/02/nvidia_geforce_8800_ultra/1

Edit: Hey post 1000 Neat!

[Jeff Kilgroe]

For those who haven't heard it has already been released:

...Well, if we're going to go that route, I'd prefer an 8900GTX please. Those GPUs are more than powerful enough. Oh and can I have one with drivers for a Mac Pro too? I'm on a Mac half the time ya know. Driving a 4K display via 2 x DVI|DL connectors isn't quite the elegant solution I had in mind, but it would work for color space at 10bpc and less. Or let's run two of them in SLI mode and then I've got all the output bandwidth needed for 4K @ 16bpc and then some.

[Gavin Greenwalt]

Why the GTX? The Ultra has a slight clock speed advantage.

Edit: Ohhhhhh 8900 my bad. Is that out? While we're at it lets' just go for the Geforce 8950:

Dual 550mhz GPUs with a total of 192 shaders/cores (Does a shader count as a core? Whatever 192 Parallel Operation Processors).
1GB of 2Ghz GDDR4 memory.

That should handle 4k REDCode ok. :)

[Rocco Schult]

apart from workflow issues and approaching it via pixel shaders, I would say:
1000-1500 USD

[Noah Kadner]

Give it a Firewire 800 option like the AJA Io HD, that way you can use it in the field with the camera. Looking at the proxies is definitely possible right off the CF cards with a FW800 reader but of course nothing would be a native Redcode Hardware decoder.

-Noah