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 :)

RED Rocket update...

My RR is working on the sub MacPro that I have for background rendering while the main machines are used for clients.

With:

Hardware Overview:

Model Name: Mac Pro
Model Identifier: MacPro3,1
Processor Name: Quad-Core Intel Xeon
Processor Speed: 2.8 GHz
Number Of Processors: 2
Total Number Of Cores: 8
L2 Cache (per processor): 12 MB
Memory: 4 GB
Bus Speed: 1.6 GHz
Boot ROM Version: MP31.006C.B05
SMC Version (system): 1.25f4
Serial Number (system): YM844017XYL

And ATTO R380 with RAID5, Apple Fiber Card with metaSAN working.

Great.
 
Red Rocket update...

Red Rocket update...

"When this baby hits 88 miles an hour, you're gonna see some serious shit." :dita:
 
Питер Энрайт;457779 said:
"When this baby hits 88 miles an hour, you're gonna see some serious shit." :dita:
Click to expand...

Hey Peter,

Why not use English characters to write your name on an English forum?
 
To change the subject a little, here is stop motion of us unwrapping the RED Rocket™™. It's kind of dorky to do something like this, but I saw all the other people taking pictures and we thought we'd try something little different.

http://shorts.creativecow.net/player/player.swf?vid=1390

-Russ
 
falcor said:
I don't know that much about this RED system but I ran across the RED Rocket™ and was kind of surprised. If this really is just a decoder/raw converter than what keeps you from running this on ~$100 commodity nvidia graphics hardware? The same is done with many formats, from mp3 decoding to BluRay. If there's any interest and it doesn't require a lot of reverse engineering on my part I can implement this in CUDA/OpenCL as a quicktime plugin to run on your graphics chip. Has the r3d format diverged much from jpeg2000? Any interest in this?
Click to expand...

hi, you should contact red directly. you know all the users in red are from the company red.

just drop them an email.

if they like your ideas there will be no need for reverse engineering.

guess jarred land would hunt you if you start reverse engineering the redcode.

i am not sure how much the jpeg2000 stuff is different from reds redcode...but they should be pretty similar i guess. (my knowledge is too small to be able to judge that)

btw. this is a real name forum.

dont feel attacked, just go the nice road. red will listen.
 
Hi Carson,

As of right now, RED hasn't opened up their file format. But the discussion of using GPU acceleration does come up here every now and then. I think thus far, as powerful as GPU solutions are, they just aren't quite up to the task unless we get into specialized configurations with multiple GPU cards like the nVidia QuadroPlex. There is also the issue of dealing with multiple hardware and software platforms. nVidia + CUDA is only one solution... I think OpenCL will address that issue as it develops more.

RED's REDCINE application does make use of a system's GPU for color matrix operations and that sort of thing and it gives decently smooth playback on most current systems at half res (2K) and a medium quality setting. Some more powerful systems, latest i7 Nehalem Xeons and overclocked i7's can do better. I think the RED Rocket is a promising product that will alleviate some of the burden on the debayer process for the next couple years until workstation CPUs / GPUs catch up to the task.

The RED SDK is freely available -- email developer@red.com to get signed up, there is a simple NDA. But like I said, they haven't opened up the file format itself or given developers free reign to do their own debayer operations at this time.

And check out the Real Names thread in the Recon forum, you can post there to get your username changed. In the meantime, you should probably put your name in your signature so your account doesn't get locked out by one of the other mods. We're just trying to get everyone to play by the same rules here.
 
Yeah thanks for the warning but it looks like my threads got nuked already.

OpenCL will run across vendors (and even multicore cpu if needed) but is a lot slower on certain tasks at the moment though it is maturing and will be released to the general public in the next few months so it should hopefully see rapid developement. This is assuming that the operation would even require scatter operations (if the location of the raw data doesn't directly map to a pixel) so it couldn't just be done through the traditional graphics pipeline which is usually faster anyway if you can fit the problem into a rasterization operation.

So Redcine already does this then or is that just for certain filters? I wouldn't be surprised if someone has already tried it and I'm just missing something though.

I'm kind of surprised they would have that much trouble with the gpu. The main bottleneck would be bandwidth (compute is essentially free for vector operations on these cards) as you have to push each frame to the card. A cheap 1gig consumer card however can still store a lot of frames though. for a 4k image 4k*4k*4 bytes assuming floating point is a size of 64 megs. Since the data is compressed that results in about 32 megs per frame lets say as a conservative estimate. The card can then store a full second of 30fps data. I can push a gig of 3D volumetric data, run sub-texel operations on it and display it to the screen in about 1fps on my card I so don't see this as being too bad. PCI-E 3.0 supports 1 GB/s x 16 = 16 GB/s transfer rates. Do they publish the specs of the red rocket anywhere? I can't find much info on it. There could be something specific to the debayer that needs a specifically designed hardware acceleration but image processing tasks are usually straightforward on gpus.

Quadros are great for scientific applications given their large storage and double precision support, in fact we're building a super computer out of them. They can be slower for some operations (ie don't use them for games) however and comsumer level cards can be very impressive in their own right especially for image processing applications like this one. Quadros are very much subject to the law of diminishing returns given the price.

Given the huge volume of GPUs driven by the gaming market we pretty much use them whenever we can as proprietary hardware ends up costing a lot more. Pretty much why Ageia physX didn't have much of a change unless they were bought out, it costs too much to make small volume hardware and its operations could be run on gpus.

That's too bad about the format being closed, way too much work to work around that.
 
Lachlan Ward said:
I was wondering what you guys think of the ATI FirePro and ATI Stream Encoder Plug-in for Adobe Premiere Pro CS4.

http://www.techworld.com.au/article/312946/amd_launches_powerful_firepro_graphics_card

Is anyone testing out the ATI Stream Encoder?

http://forums.amd.com/amdlive/messageview.cfm?catid=366&threadid=114667&enterthread=y

Here is a not so great article talking about how they got an 8x boost to there encoding times.

http://it.tmcnet.com/news/2009/06/16/4227124.htm
Click to expand...

interesting for sure at $1799. Marking the thread for follow up. anyone test this card yet?
 
JonathanF said:
Redcode IS jpg2000 as has been pointed out here:

http://reduser.net/forum/showpost.php?p=275296&postcount=7
Click to expand...

Unfortunately it seems to be encrypted though specifically so you can't read it in as a jpg2000 stream. That seems to be the only hurdle I can see from writing a cuda/opencl program and running the same operations on a gpu though and then you could use the gpu for other things. you can get a nice quadro 5800 for 3.5k at the high end and put that 4 gigs of video memory and over 100 GB/s memory bandwidth to good use. Does anyone know of the technical specs of the red rocket?

Someone was saying earlier that redcine did implement the same operations on a gpu though and can't do full resolution however is that correct?

The part I don't understand with this though (and this fully displays my ignorance on the subject) is that monitors only display up to 2.5k res anyway aside from powerwalls. Do people really work in real time with such high resolution video in some sort of quad monitor setup or something crazy like that? Or is this more useful for faster rendering? I'm always amazed at how long it will take to render out a presentation video in motion.
 
You must log in or register to reply here.
Back
Top