"Redcard" Redcode accelerator?

[lordtangent]

Now that Red is working on Red Ray, I'm wondering... couldn't they take some of that technology and put in on a PCIe card to help accelerate Redcode workflow?

I realize the driver issues to host it in a given software package / OS could be insane. But maybe if they implemented it correctly, say as a codec in Quicktime or Direct Draw so that every software saw it equally it wouldn't be quite as insane. It could work a lot like the QT proxy work flow in that respect. (only accelerated even at full res HQ)

Basically it would be the same J2K and debayer engine they are putting in the Red Ray, only on a PCIe card. The system would hand it Redcode and get back RGB buffers. "Moores Law" might make the development of such a device not worth while. But since I'm hearing people give times like 1.5 FPS on transcode from Redcode, maybe not? At the very least, it would free the system CPU from the work of decoding and de-bayering and let it work on other stuff. If the price were in the same ballpark as Red Ray and gave real-time or near real-time de-J2K and de-bayer at 4k, I'm sure a lot of people would be interested. I'm thinking about when I get a Scarlet and I know I would certainly be interested if it sped decoding up for me.

What do you think?

 

[Dylan Reeve]

It's a good idea, but invariably software and hardware improvements tend to render these sort of accelerators somewhat unnecessary fairly quickly.

Certainly not a bad idea at all though.

 

[Sven Seynaeve]

This would be a nice idea, anyway I think some application to transcode on multiple pc's would be better.

But a card being able to give us 4xhdsdi output or hdmi would be very nice if we could use this in combi with our edit or grade station.

 

[lordtangent]

This would be a nice idea, anyway I think some application to transcode on multiple pc's would be better.

You mean like a render farm manager? There are no license fees for Redline. It seems to me that on the Mac especially, this would not be too hard to set up (using Q-master, rsh, ssh or one of the myriad commercial render farm manger tools available ) Maybe "Crimson Workflow" could be extended to do it? Each clip could be sent to a network machine for rendering. It's not a very complex problem. And Red has already provided enough for some enterprising individual to do it themselves. I'm not sure if it would be worth it for Red to have to deal with network rendering.

I agree with Sycophant than Moores Law and software improvements tend to obsolete specialized hardware while it's still in the shrink wrap. The only reason I even suggest the idea of a "Redcard" is because I noticed the price of Red Ray being so low. Even two or three years from now getting enough network render computers to get software-only 4k decode & demosaic from 1.5 FPS to real-time would cost way more than $1000. It seems to me that makes something like "Redcard" worth while if the price is in the same ballpark as a Red Ray. $1000 is cheap enough that I think most Red One users would probably go for it once they did the math.

 

[Dylan Reeve]

Assuming that RED is using a moderately powerful FPGA in the RED One (and other products) with an instruction set tuned to the handling of the RED media, there is no real reason that such a card couldn't be made fairly simply I believe.

Certainly I can see a benefit to having a PCIe card that provides native or near-native R3D image handling (in much the same way we used to have MPEG-2 cards way back when).

 

[lordtangent]

Red might be using one of the J2K chips like the AD212 available off-the-shelf from Analog Devices. It's a $25 part.

http://www.analog.com/en/prod/0,2877,765%255F810%255FADV212,00.html

But yeah, a FPGA wouldn't be too expensive either.