Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

DRAGON debayer options - Red Rocket-X hardware versus RCX-P software only

Robert Ruffo New said:
When I'm hired as a DP I don't write the story - and if I do that's a separate contract and negotiation, that may as well be another person a steh two jobs are very separate. As a DP - regardless of who wrote the script - I offer that the finest details will be taken care of in the most attentive and sensitive way possible within the confines of available time and money. As a DP I am paid to deliver the best visual quality possible. If you think thats' not important, why are you a DP? Think of what you are saying, it's the equivalent of saying you aim low-ish in your work because anyway who cares, the writer's job is all that matters. I doubt you really want to say that, not here, and likely not even to yourself.

It doesn't matter what people managed to do 40 years ago - at the time they were using the best technology they coudl afford in the best way they could, and I try to do the same now as best I can.

RR debayer is not the best, a better alternative is always affordable to any production (any production can afford a few hours of render time at the very end of post), and so choosing that better alternative is an obvious choice. Sometimes complicated light set-ups and so on can eat too much set time and negatively affect story or actor comfort, and there compromise is indeed needed of visuals vs story, there it's important to remember the importance of story and performance - but here - software debayer does not affect story at all. The actors have gone home when debayer is done - story will be the same, just the image quality will be better.
Click to expand...

Debayer quality aside, I look at being a DP more than delivering the best quality on whatever media. There are so many elements to telling a story beyond the script. Naturally we have to add the visual as the cinematographer. The aesthetics of the camerawork is more important to me. The decision for a DP to go handheld vs. sticks is major. A dolly shot vs. static is major, etc., etc. It's all tone and thats what gets me excited. I'm on Reduser and I enjoy the technical side of my craft as many others on this forum. But when I look at the legacy of films that have shaped my knowledge of the craft, my emotional response is mainly derived from "the dance". I love blocking out scenes with the director. For me, if the blocking isn't sitting well, it sets off a chain of bad compromises and the scene falls flat. After blocking, the decisions for light come into play. A super key ingredient to the story. Then lenses and framing decisions which incorporate how animated the camera is going to get. All decisions before the debayer stage. Perhaps we are saying the same thing? In any case, this thread is indeed about debayering and perhaps you misunderstood me, I do prefer software encoding over hardware.
 
Bob Gundu said:
Debayer quality aside, I look at being a DP more than delivering the best quality on whatever media. There are so many elements to telling a story beyond the script. Naturally we have to add the visual as the cinematographer. The aesthetics of the camerawork is more important to me. The decision for a DP to go handheld vs. sticks is major. A dolly shot vs. static is major, etc., etc. It's all tone and thats what gets me excited. I'm on Reduser and I enjoy the technical side of my craft as many others on this forum. But when I look at the legacy of films that have shaped my knowledge of the craft, my emotional response is mainly derived from "the dance". I love blocking out scenes with the director. For me, if the blocking isn't sitting well, it sets off a chain of bad compromises and the scene falls flat. After blocking, the decisions for light come into play. A super key ingredient to the story. Then lenses and framing decisions which incorporate how animated the camera is going to get. All decisions before the debayer stage. Perhaps we are saying the same thing? In any case, this thread is indeed about debayering and perhaps you misunderstood me, I do prefer software encoding over hardware.
Click to expand...

I did misunderstand you - and agree with everything you said. I feel the same way, but to me it all should be as good as it can be, including lighting. I tend to moved by lighting as much as anything else. I got into this business because I learned photography when I was a teenager, and I realized that making a beautiful image of women, perfectly lit, subtly photoshopped, with perfect makeup, helped them feel wonderful about themselves. It was an awesome gift I could give them. I fell in love with that perfected fantasy world in those pictures, which I think is among the reasons many people like movies, need movies even.

You can see why I am not a fan of many Canadian productions - most (even commercials) value "the gritty truth" which is the exact opposite of what I love and value, and think also is one of the reasons why so many Canadian films are so unpopular with Canadians and the rest of the world.. They are often devoid of fantasy, magic and imagination. They are dreary, sad and depressing, drab, like a view out the window in November in Mississauga, and I think part of that is abysmally bad scripts but the other part is ugly, half-hazard lighting stuck in 1970s quasi-solialist ideologies about naturalism. Indeed government filmmaking is about as sexy as a government office - which makes sense in a way.

But just to be clear I am adding to what you said, not arguing with you - an I am not implying that I think you are typical of Canada in any way.
 
Here is an example of decode quality example for you, it shows pretty clearly how CPU is different/better, just unfortunately extremely slow because the red software is single threaded :banghead:.

CPU vs GPU vs Rocket

Maybe now with the red dress in the closet and smpte over, Neil could post a similar example for us of rocket-x vs cpu ?


Dave Summers said:
OK, so I'm really confused at the thought of getting better quality with a RRX instead of software debayering???
A full debayer should produce the same results, with or without a rocket.

As far as I know, the only difference should be the obvious increase in rendering time when using software. This makes me wonder if RCP-X has been engineered to cause a slight downgrade in quality in the absence of a RR card?:confused1:
Click to expand...
 
Christopher*Seguine said:
Here is an example of decode quality example for you, it shows pretty clearly how CPU is different/better, just unfortunately extremely slow because the red software is single threaded :banghead:.

CPU vs GPU vs Rocket

Maybe now with the red dress in the closet and smpte over, Neil could post a similar example for us of rocket-x vs cpu ?
Click to expand...

This reflects my old findings...
This is a comparison of RED ROCKET vs. RED Software decoding.
Please read the text inside the images carefully to understand what you are looking at.

Actor: Gen Seto
Movie: Sushi in Suhl, 2012
With kind regards of StarCrest Media



Important: Neither the new GPU stuff nor the new RED ROCKET-X is being used here.

The source material is a RED R3D shot on RED ONE MX as far as I remember.


This image shows different kind of artefacts when pushing gain strongly:


https://dl.dropboxusercontent.com/u/22067399/RED%20Software%20SDK%20vs.%20RED%20ROCKET%20-%20GAIN.png

RED%20Software%20SDK%20vs.%20RED%20ROCKET%20-%20GAIN.png



This image shows sharpness and aliasing issues:

https://dl.dropboxusercontent.com/u/22067399/RED%20Software%20SDK%20vs.%20RED%20ROCKET%20-%20SHARPNESS.png

RED%20Software%20SDK%20vs.%20RED%20ROCKET%20-%20SHARPNESS.png



We made this for our customers to understand why we use software only decoding for VFX and final RED output work.

As to our experience, the difference is always there.

Regards,
Axel
 
Hi Everybody,

since our first Dragon is gonna land back in Germany soon, the decision of if or how many RED Rocket X's to buy is on the doorstep.

Did anyone ever since make a comparison with RED Rocket X vs Software / CPU Debayer?

thanks a lot!
 
The RRX is better than the old card. Differences between that and software are now more subtle.

RRX is still lagging behind though, and I very much dislike the artificial sharpening it adds, whether we want this or not. Not nice to women's skin, for sure.

I should also warn that if you grade things with big HSL qualified pushes further shortcomings will become visible in any RR or RRX (and thus technically non-optimal) decode, as is the case if you do digital zooms or try to track small objects. Here no strong pushes are done, and so the underlying shortcomings are less visible.
 
Why not? Don't we all benefit from playing around with high quality footage? Art inspires art?

With Phil's permission, would it hurt anyone?


Neil W. Smith said:
Check these out and let me know what differences you see .. please don't apply any "artistic improvements" ... just look at the images as they on the best screen/monitor you have and see what you can see
Click to expand...
 
Robert Ruffo New said:
The RRX is better than the old card. Differences between that and software are now more subtle.

RRX is still lagging behind though, and I very much dislike the artificial sharpening it adds, whether we want this or not. Not nice to women's skin, for sure.

I should also warn that if you grade things with big HSL qualified pushes further shortcomings will become visible in any RR or RRX (and thus technically non-optimal) decode, as is the case if you do digital zooms or try to track small objects. Here no strong pushes are done, and so the underlying shortcomings are less visible.
Click to expand...

Robert, or anyone else who has got a Rocket X in their Hands, it would great if it would be possible to post some 100% Size Comparison Images, maybe even blue channel only, similar like in this thread: http://reduser.net/forum/showthread.php?107462-CPU-vs-GPU-vs-Rocket-vs-RocketX-Quality or like Axel Mertes Comparison with the Gain, to see if there are still strong unwanted Artifacts in the RR X Decode.

Thanks a lot!
 
What about dailies transcoding?

What about dailies transcoding?

I am seeing a lot of talk about high quality transcoding speeds using the Rocket. Considering that most high end post houses use software debayer in the final render due to quality concerns, I'm more interested in lower quality transcoding for dailies, where speed is far more important than quality. I don't own a Rocket, but would love to see some tests comparing 1/4 debayer OpenCL to a Rocket and to a Rocket-X. For dailies, I usually make 720p ProRes with 1/4 debayer, and I can get about 25fps on 5K material. I think adding an original Rocket to my system wouldn't help much, if at all, with this.

If anyone is willing to do some tests like this, I think that would be very interesting. Also, how does a Rocket-X affect speed with lower res 4K material? I would be happy to run and document these tests if I can borrow a Rocket from somebody.
 
Right, the Rocket always does full debayer. That's why its value goes down in a situation when lower debayer is adequate. I would like to know if it's worth buying one, or if my CPU+GPU is just as fast.
 
Elliott Balsley said:
Right, the Rocket always does full debayer. That's why its value goes down in a situation when lower debayer is adequate. I would like to know if it's worth buying one, or if my CPU+GPU is just as fast.
Click to expand...

Its hard to tell until its ready to ship. Lots of optimization is going gone right now.
 
does anybody had the chance to test this ?
Is rocket-x quality equal to software debayer ? or , like the first red rocket , software debayer is still superior ?

thanks

g
 
One small note on doing Rocket debayering vs. 1/4 or similar for dailies: because the Rocket does a Full debayer, running transcodes on it can be used as part of the QC/footage verification process. Because REDCode is a wavelet compression scheme, debayering at 1/4 (or 1/8, 1/2 etc.) can work even when there may be some kind of corruption in the data in the higher resolution range. If you transcode on the Rocket, and there is anything wrong with the source file at all, the transcode is likely to fail or at least come out with visible artifacts that may not be present with a lower resolution debayer. Of course, you can get the same verification without a Rocket by doing a Full debayer on CPU/GPU, but obviously that is time consuming.Just another thing to think about when considering whether a Rocket is the right choice for your purposes.
 
I have a 1 frame R3D that I could send someone who has the Rocket-X for verification.
I know the results from the software deBayer and the old Rocket for this frame, so we could easily tell.

If anyone has such a card at hand, please contact me via mail to axel at magnamana dot com.
 
Eric Cameron said:
One small note on doing Rocket debayering vs. 1/4 or similar for dailies: because the Rocket does a Full debayer, running transcodes on it can be used as part of the QC/footage verification process. Because REDCode is a wavelet compression scheme, debayering at 1/4 (or 1/8, 1/2 etc.) can work even when there may be some kind of corruption in the data in the higher resolution range. If you transcode on the Rocket, and there is anything wrong with the source file at all, the transcode is likely to fail or at least come out with visible artifacts that may not be present with a lower resolution debayer. Of course, you can get the same verification without a Rocket by doing a Full debayer on CPU/GPU, but obviously that is time consuming.Just another thing to think about when considering whether a Rocket is the right choice for your purposes.
Click to expand...

Fine, but this issue has never occurred in my real life, maybe 20 000 clips into my adventures shooting with Red. One time I got a single skipped frame, and it was visible at all resolutions.
 
You must log in or register to reply here.
Back
Top