My Big Idea: the Bottle Rocket

[David Battistella]

Thanks Wayne.

Doesnt REDCODE really need CPU horsepower to decompress the wavelet based Jpeg compression in the the file.

There is the debayer but also the decompression. If they can playback 1080p from the 9K$ camera CPU (scarlet) couldn't there be a much less expensive 4k card.

Maybe BMD or Aja could make it.

I just want to see this "hard to post RED" BS over with.

Not for me, but for the format in general.


David

[Wayne Morellini]

In the industry things are often missed because of too narrow a focus. It might seem to them that using general processing GPU's are not suitable, but GPU's general purpose processing ability is advancing, and it is a matter of how you process that effects it's suitability for GPU's, even if they have to change red code and wavelets. I am associated with a community that often used to redefine processing to speed it dramatically up, and also did my own waveformation technique. Wavelets however, are not my thing.

[Wayne Morellini]

Well, I am buying a modest quad core i7 (not 12 core+) and GPGPU graphics card hopefully, and hopefully that will be sufficient for the next 5 years. The next Macmini, might even be upto it. But I hope that this red post issue is sorted by then. Red Rocket was only a short term fix until gpu processing matured. I might have been the one that suggested the Red Rocket's processing chip to Jim originally (I came over from a previouse DIY digital cinema camera group that Jim followed around 2004).

I also have been thinking of posting a thread somewhere, on what is the minimum needed to make a good film, and it will be shocking how little camera wise, is needed.

[Joost Kelderman]

just $ 0.02

well.. having it look obsolete in 2 yrs might be a bit crude.. sure maybe in two years we'll have full decode on GPU; but you forget that at that point you'll still be putting a heavy lock onto system resources. having the thunderboilt debayer/wavelet device will keep your GPU free to do more suited tasks like encoding into proxies; first colors and if used in editing all the other stuff the GPU is already handling atm. been some years into editing but i've never heard an editor/compositor or anyone complaining about having to much raw power in a system [except for those moments that you're having a coffee break during renders] (-;

I think a lot of people/bosses would like something that saves them an hour of rendering time per day.. 300 hours a year would earn that back even if you only ask $10/hour. other thunderbolt devices might add just a couple of hundred CUDA-cores, just a couple of DSP-farms for audio. that is slowly starting to happen.
___
quote:
Going back to my point above, I'm not sure if it's really what people ultimately want. No one is going to give a crap about getting footage converted to ProRes and DNxHD or whatever once they can seamlessly play and edit R3D files just as easily. I prefer RED continue focusing their efforts on the software tools and SDK. If they want to invest in more resources and people, then put them on those tasks. The latest crop of GPUs coming up in the near future have vastly improved compute engines and should be much better suited for wavelet decoding -- the computationally-expensive portion of R3D decode. I would rather see efforts concentrated there. IMO, if software can evolve to match current and upcoming (known) hardware, then such a Rocket product would (should) be obsolete within a year or two.

[JanneJansson]

File formats are just relative when dealing with raw. If you combine the idea of having a RED-MAG reader with a Rocket and thunderbolt bus then you get some interesting possibilities...

When mounting a RED mag to your RED reader/rocket combo you could make the download and convert process the same operation. Just make the hardware in the reader/rocket display the raw files in "virtual folder" on the red mag and then you just "download" the files you like. When the transfer operation begins the hardware transcode raw to whatever need. Prores, Dnxhd, h264, etc.

[David Battistella]

That is a cool idea.

It sort of gets rid of a step.

I guess the idea is to have the same horsepower in a small enclosure that can travel from machine to machine, platform to platform and fit the wide ranging size of productions that currently use EPIC or Scarlett.

A high octane box, with a compact card inside it that just rips through REDCODE.

The R1 is to the RedRocket (form factor) what the EPIC is to the (would be) BOTTLE ROCKET = compact, ready for battle form factor.


David

File formats are just relative when dealing with raw. If you combine the idea of having a RED-MAG reader with a Rocket and thunderbolt bus then you get some interesting possibilities...

When mounting a RED mag to your RED reader/rocket combo you could make the download and convert process the same operation. Just make the hardware in the reader/rocket display the raw files in "virtual folder" on the red mag and then you just "download" the files you like. When the transfer operation begins the hardware transcode raw to whatever need. Prores, Dnxhd, h264, etc.

[David Battistella]

the camera turned the motion capture industry on it's arse.

This should shut everyone up and turn post production on it's arse.

David

[JanneJansson]

The camera is a reader/rocket combo already for some situations it could be useful to just connect the camera to a computer and download. So a thunderbold module to the camera could make for many-many new cool features :)


..but I think "Pocket Rocket" is a cooler name :P

[David Battistella]

I agree.

Let's make it Pocket Rocket.

But let's hope they make it!

David

[Wayne Morellini]

One day, that might be the size of it.

If you want to find out how efficiently and quick bayer wavelet compression can be done, ask David Newman at cineform. I trust his opinion in this as a premium expert on it, he made the first I've come across to do this for the SI camera.