Welcome to our community

Be a part of something great, join today!

  • 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 6k R3D RCX Full Rez... What Station you'r doing this with?

Getting 99% GPU usage in playback Does adding a second titan increase playback speed?

If you're talking about Resolve, or other apps that can use multiple GPUs, then yes. If you're talking about REDCINE-X, then no… Not yet, anyway.

I've come to like 8.1 if one applies the Start8 app. Haven't seen Metro UI in 8 months.

http://www.stardock.com/products/start8/

Yep, there are solutions. Unfortunately, most of those are difficult to propose to corporate clients or those who must answer to IT department software nazis when it comes to installing and configuring things the way they would like.
 
GPU-z is showing activity across all 3 CUDA GPU's I have here in both RedCine-X and Red Player.
 
can some of u guys elaborate further on the CPU vs. GPU processing that is done in RCX / Resolve in a system that does not contain a RR/RR-X card ?

I know that the CPU handles decrypting & decompressing and GPU from the debayering stage on, but what are other processes that the CPU must handle ?

I see posts from reduser that state their CPU can't keep up with full res playback of 6K footage and I wonder if we can we (roughly) estimate what the min CPU setup needs to be to handle 6K footage in full res... 2 x 8core ? 2 x 10-core ?
 
I built my current system for $10k so for me that is the sweet spot.

I don't need a portable system as it lives in the edit suite.

Connectivity: I need one slot for the Areca card to connect to my Maxxdigital SAS raid.

Windows 7 or 8? Either one, although with 8 I would have to run the classic shell mod to get the Windows 7 look as I hate the 8 design.

IMG_0251.JPG

Nice set up Matt!!
 
GPU-z is showing activity across all 3 CUDA GPU's I have here in both RedCine-X and Red Player.

I see the same thing, but GPU-Z doesn't match up with the GPU monitoring built into Resolve. It also seems to generate ghosts and assumed usage across multiple GPUs, probably due to it trying to monitor driver activity. Not sure how it would really dig in there to know for sure. Turn off and/ or pull your extra GPUs from the system and see what happens to your RC-X performance. I'm tellin' ya, RC-X doesn't use more than one GPU right now.

can some of u guys elaborate further on the CPU vs. GPU processing that is done in RCX / Resolve in a system that does not contain a RR/RR-X card ?

I know that the CPU handles decrypting & decompressing and GPU from the debayering stage on, but what are other processes that the CPU must handle ?

I see posts from reduser that state their CPU can't keep up with full res playback of 6K footage and I wonder if we can we (roughly) estimate what the min CPU setup needs to be to handle 6K footage in full res... 2 x 8core ? 2 x 10-core ?

Decompression or decoding of the R3D wavelet file is different than the demosaic (debayer) process. However, all of those tasks are handled by the R3D SDK module or API calls. RED has implemented acceleration support for all those processes for the Rocket, Rocket-X and GPUs via OpenCL and CUDA. It's up to the software developers to utilize a system's resources in the best way possible to balance the load across the different options. The GPU-enabled SDK has only just shipped out, so we are not seeing that acceleration incorporated into Resolve, Premiere or others just yet. For apps like Premiere and Resolve that already use GPU acceleration to speed up other tasks, they will have to balance R3D acceleration into the mix. It may even come down to requiring multiple GPUs to do it best as separate GPUs can be allocated to a different part of the process. In Resolve, two GPUs could be allocated to node processing, while a third can be allocated to R3D acceleration.

Prior to GPU acceleration in the SDK, things were much simpler. In Premiere/ Resolve, the CPUs did all the decoding and debayer of the R3D, or this process was accelerated by the Rocket, if available. GPUs were used to accelerate everything else afterward.
 
Decompression or decoding of the R3D wavelet file is different than the demosaic (debayer) process. However, all of those tasks are handled by the R3D SDK module or API calls. RED has implemented acceleration support for all those processes for the Rocket, Rocket-X and GPUs via OpenCL and CUDA. It's up to the software developers to utilize a system's resources in the best way possible to balance the load across the different options. The GPU-enabled SDK has only just shipped out, so we are not seeing that acceleration incorporated into Resolve, Premiere or others just yet. For apps like Premiere and Resolve that already use GPU acceleration to speed up other tasks, they will have to balance R3D acceleration into the mix. It may even come down to requiring multiple GPUs to do it best as separate GPUs can be allocated to a different part of the process. In Resolve, two GPUs could be allocated to node processing, while a third can be allocated to R3D acceleration.

Prior to GPU acceleration in the SDK, things were much simpler. In Premiere/ Resolve, the CPUs did all the decoding and debayer of the R3D, or this process was accelerated by the Rocket, if available. GPUs were used to accelerate everything else afterward.

thanks, Jeff !

I guess we gotta wait and see how the new SDK works out in real world application, but that sounds great on paper since GPUs are less expensive and we can just throw multiple in a box...

btw, any thoughts or release date info of GTX-780 6GB ?
 
What system would you be installing all that in, Blair?
 
The right tools

The right tools

What system would you be installing all that in, Blair?

That's a good question. I have a GlobalStor ExtremeStorDI rig with a good PSU, decent cooling and plenty of room. It's a few years old, so it would need a new mobo, X79 perhaps…

I was thinking E5-V2 CPU, probably the 2667. Also looking at the i7 4930K.

The strategy of stuffing slots with gamer cards is likely the best bang for the buck. I'm just looking at options that feature purpose built graphics boards like the RR-X, Tesla K20 and BlueFish444 Epoch Supernova 4K as potentially better suited to our world.

RR-X: I do a high percentage of my work with R3Ds so this is an easy call for me. I understand why others would rather go the GPU route since it speeds up other things. This card gets a lot more attractive when the video outs are activated - does anyone know if it will output 12 bit?

Tesla K20: Perhaps specialty cards aimed at compute intensive tasks are old school. Would love some perspective on bang for the buck vs gamer cards.

Bluefish444 Epoch Supernova 4K+: I really like the idea of 12 bit monitoring/output at 4K/60P. This seems like a card that could be the heart of a DI suite capable of doing client supervised sessions.

One part of this whole equation that I would love a proper explanation of, is internal signal path. Source media is read off a drive (RAID), decrypted, de-bayered, scaled, perhaps transcoded, etc. What's the method for determining which resources are utilized for various tasks? These types of issues may become even more important once we know more about encoding .RED files (assuming the format has a future).

Cheers - #19
 
We're working directly with Neil at Luma forge, Ted at Devil/Demon etc, and there are some very powerful and well-supported turn-key systems now available. This said, if you guys are interested in a group buy on AMD W9100 or Sapphire Radeon R9 GPU's, I am happy to help set that up.

Ketch, feel free to reach out to me privately at adamglick@sapphiretech.com

Cheers,

Adam
 
We're working directly with Neil at Luma forge, Ted at Devil/Demon etc, and there are some very powerful and well-supported turn-key systems now available. This said, if you guys are interested in a group buy on AMD W9100 or Sapphire Radeon R9 GPU's, I am happy to help set that up.

Ketch, feel free to reach out to me privately at adamglick@sapphiretech.com

Cheers,

Adam

Hey Adam, first welcome to our wonderful community that is "Reduser"... ;)

Still testing out few other options, especially with some cards not as pricy of those that can't afford to spend as much... But I'll definitely be more then happy to set up a AMD W9100 Group buy if that becomes the end card of choice...


Btw, please say hi to Neil for me!!
 
Many have had issues with the tesla cards and GPU acceleration applications such as Adobe. I would not suggest that for any editing applications that use GPU acceleration. Those are made for scientific and math calculation applications more than anything else. I suggest you get the Titan or 780Ti card instead. The 2667 V2 chips would be fine thought at that price I would just get the 10 Core's instead.

The Data is cached from the storage to ram based on the applications caching models. That means the amount of frames buffered when loading is specific to the applications and the system. The more ram you have available the caching will increase based on the models used by the applications.
 
That's a good question. I have a GlobalStor ExtremeStorDI rig with a good PSU, decent cooling and plenty of room. It's a few years old, so it would need a new mobo, X79 perhaps…

And new PSU. You'll want to match up a new PSU anytime you install a new motherboard. You won't think anything of it if you do, but will probably end up cursing yourself if you don't. At this point, you may want to be sure you're in love with the system case. You're essentially building a new system, may as well go all the way and continue to use the old one or sell it.

I was thinking E5-V2 CPU, probably the 2667. Also looking at the i7 4930K.

The 2690v2 is a better value for the same price. It's 10% slower at 3GHz, but gives you two more cores. It's my favorite CPU of the current Xeon lineup. :) It's only 100MHz slower when rocking along with 8 cores and the same max turbo speed for 1 and 2 core situations vs. the 2667v2.

If you want to go with a faster 8-core, then I would look to the X99 motherboards coming soon and the new i7 8-core CPUs. No official chip numbers just yet, but there's reportedly an 8-core i7 desktop class CPU clocked at 3.4GHz with 4GHz turbo. Liquid-cool that thing and I bet we can boost that to 3.8GHz w/ a 4.4GHz turbo. I look forward to giving that a try as soon as I get my hands on one. ;)

The strategy of stuffing slots with gamer cards is likely the best bang for the buck. I'm just looking at options that feature purpose built graphics boards like the RR-X, Tesla K20 and BlueFish444 Epoch Supernova 4K as potentially better suited to our world.

The K20 is a Quadro K6000 with no video out ports. The K20X is the same but clocked a bit higher. The GTX Titan is essentially the same thing, but less memory, clocked the same as the K20X plus you can also buy the SuperClocked version. It only has 6GB RAM and no ECC memory support, less robust memory controller and such, but when it comes to general CUDA acceleration in apps like Premiere and Resolve, it hangs with the K20X just fine. And you can add two of them for the same price as one K20X.

And Eric is right about the Tesla cards with Premiere. Or at least the latest Kxx series. Something about them and Premiere is a bit off. They work great in Resolve, though. Not worth the extra expense, IMO, unless you're specifically running something you know will use their extra features.

Not sure on the BlueFish444 cards. They have always been excellent cards, but this really comes down to software support these days. It's all about which card will work with the software you want/ need to use. Personally, I like AJA and some of the older BlueFish products I've used in the past, but I'm mostly on BMD cards lately because of DaVinci Resolve.

RR-X: I do a high percentage of my work with R3Ds so this is an easy call for me. I understand why others would rather go the GPU route since it speeds up other things. This card gets a lot more attractive when the video outs are activated - does anyone know if it will output 12 bit?

Lots of unknowns about the RR-X right now. I guess we patiently wait for them to ship out to more of us who are waiting and see what happens as all the features are enabled. I'm guessing RED has been slow on the ramp-up of production because they're making sure everything is working fine before sending out a ton of cards.
 
Rendering out the first project on the new Beast. Firstly, I LOVE that I could just copy over the Premier Project from my Mac and open it up on the PC with ZERO issues. Thank You Adobe!!!

Now, I'm rendering from resolve 5k Epic footage to HD DPX sequence at 35 fps, No Rocket.

Love my new machine!!!
 
"... Hey Adam, first welcome to our wonderful community that is "Reduser"... ;)..."

Actually, it's been a few years but I was on this forum before... Something happened to my old login/profile.

I developed the "RedBoxx" - the first system capable of playing 1/2 res , full quality R3d clips at 24fps. At least, until they released the original RR card...

ADAM
 
I think its difficult to come to a consensus when we all have different needs/budgets/and somewhat apps to a certain degree...


I would LOVE anything mobile that can encode fast and isn't a pain to bring to set. I am considering rocket x with a rmbp for this since mobility will rock. Apps I intend to use will be the new resolve lite, cs5, avid mc7 and potentially installing fcpx but not sure on the latter. Because of my location its usually much easier to take a train to sets then drive in, if possible. Also I am usually just encoding prores, but potentially can get hit with higher demands such as DPX or basically anything the client wants. Right now I own nothing but have a dragon on the way to me so the time has came to finally step up. I am split 50 50 on if I should get a trash can instead... seems like it will just balloon because of the monitor transportation needs if I go that route but it makes much more sense for non red jobs processing power wise so I am not sure what to do. I have to figure out some decent way to pack it all as small but as powerful as possible essentially...
 
Sorry quick Q:

I have no experience with non quadro nvidia cards... If i was going to try out a titan black (to replace a quadro 5000) which repackaging manufacturer is best. (Asus, Evga etc)? price is less of a concern than stabity/speed. It will go in a z800 (with 2 x6 core Xeons)...

I can note that they offer slightly different clock speeds and guarantees and I can pick just off this difference but since I have no experience with gamer style cards I wanted advice?

Thank you!

Michael
 
I was directed to this link to figure out a Mac Pro solution to replace my Red Rocket without going to a Red Rocket X. And there are 98 pages! 98! Without having to plow thru all of that - can someone tell me a viable solution or two? Would like to see Reduser have some kind of aggregate function to simplify some of this.

Have FCP 7. Can I edit in 4k ProRes files?

What drive speeds do I need?

Best card to process RED Dragon files. Best to output 4k?

Sorry to come in - just a bit daunted by the 98 pages.

Thanks,

Dale
 
FCP7 is a 32bit program. I don't think it can do 4k prores editing, but even if it could you would not want to use a 32 bit program since it can't utilize much Ram at all anyway. The consensus is some programs do better with more cores, some do better with higher processor clock speeds, so its highly dependent on what apps you will use is the answer as far as what trash can is the "best". are you talking about the new mac pro, if so not much card choice there i'd get the d700s no matter what amount of cores you buy...
 
Back
Top