Welcome to our community

Be a part of something great, join today!

Register 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 :)

IBC - GPU Build REDCINE-X Beta 21

Kalulu Ngilo said:
What is the typical size per EXR file? I'm getting around 108MB for my exports, that is the correct file size I imagine?
Click to expand...

5k FF is 79.1mb EXR with no compression. It writes the full file, its just a corrupt image.

If I apply any scaling on EXR export I get an OpenCL error.

I think the problem is probably EXR is Float and all other exporters are 16bit or less.

CPU/Rocket EXRs work, GPU is the only one that is corrupt.

Ticket #4321 filed.


GPU.jpg
 
You are right. I tried opening the EXR files with PS CS6, this is the result I got.

*****Later EDIT**** ONLY HAPPENS ON BETA RCX 21.2.

corruptedEXR.jpg



RCX Build 20.1 opens Phil's 6K files, exports EXR files and Now PS CS6 is able to see the files. I should also note that I am using Beta nvidia drivers(331.40) too!

workingEXRPSCS6.jpg
 
Last edited:
Thanks for the EXR reports, we'll take a look at that. It was supposed to have been fixed already, so something must've gone wrong with this build. We'll get it sorted out.
 
Mike P. said:
Hey Rob, is there any possibility that using Intel's Iris IGP (the one with OpenCL 1.2 support) will actually help debayering in RCXp21b? I think it only has 128m eDRAM and then shares with system, but it'd be cool if even the lower end 13" Retinas (and presumably 15" if the rumours of them ditching discrete gpus is true) had a little extra help in the debarring/playback department.
Click to expand...

That's not enough memory by far, we require 768 MB at least. And I would definitely not recommend that to be shared either.

I don't believe that IGP would have enough processing power, but we'll take a look at that after Cuda implementation.
 
Hi Rob, do you know why I can't get the beta builds to open up my R3D without crashing? As soon as I open up a clip it crashes.
I'm on a MBP Retina 2.7Ghz 16gb Ram on Mac 10.8.5
 
Rory: yes, you don't want it badly enough ;)

Could you send us a verbose (can be enabled in RCX preferences) RCX log so we can see why it might be crashing? Please also send the OS X crash log with that. You can email both to redcinex @ red

Thanks,

Rob
 
Daniel Reed said:
I tried the older build of RCX... 21.03... which incorrectly(i suspect) reports the video card as only having 512MB RAM - MAC sys info and the original mac box both report the Radeon HD 6750M as having 1024 MB VRAM... so hmmm... where do I send bug reports to? I have no other displays connected atm, just using the plain built in LCD atm.

On a side note it definitely seems accelerated on the 21.03 build... renders & playbacks previews way faster of 4K content w/ openCL turned on!

...but OpenCL won't enable on the 21.1 build and renders and playbacks previews of 4K content painfully slow... a single frame takes approx 10 secs to render.

Thanks!
Click to expand...

It seems that with some cards OSX only reports 512MB to OpenCL.

Unfortunately we only have access to the memory available on the card through OpenCL and require 768MB minimum.

A google search for "OpenCL 1GB only 512" will show more information about this issue with OSX and OpenCL on certain cards.
One particular link with the issue relating to some ATI cards is: http://parallelis.com/some-thought-on-opencl-on-mac/
 
Rob Lohman said:
Thanks for the EXR reports, we'll take a look at that. It was supposed to have been fixed already, so something must've gone wrong with this build. We'll get it sorted out.
Click to expand...

I know it's late but I wanted to let you know that 21.2 Beta outputs the EXR files just fine on the hackintosh. I used preview 6.0.1 to look at the EXR files, they appeared fine.... :rolleyes: the files are about 113.3MB in size vs 108MB in Win7.
hackintosh i7, gtx670, 16GB ram.

Hope that helps narrow the problem down.
 
Last edited:
Thanks for the continued reports. I've fixed the EXR issue for good now .. it was hit or miss. Sometimes it would work, other times not. We'll put out an update soon.

Thanks,
David
 
Is the Titan going to be a significant speed increase over the 7990? Would just like to get somewhere close to real-time 4k, so saving a few hundred bucks by getting a 7990 would be ideal.

Anyone have any experience with both cards and this release yet?
 
A 7990 is actually 2x7970's on a single card, and shows up to the OS as 2 cards. We don't yet support multiple GPUs. Additionally, we've run into some issues with AMD's HD 7900 series that causes them to run much slower than they would otherwise. We're working on resolving that issue with AMD now, but there's no ETA on when they'll be delivering a fix for end users.

That said, I have a 7970 GHz Edition at home and love it :)
 
Grant Limberg said:
A 7990 is actually 2x7970's on a single card, and shows up to the OS as 2 cards. We don't yet support multiple GPUs. Additionally, we've run into some issues with AMD's HD 7900 series that causes them to run much slower than they would otherwise. We're working on resolving that issue with AMD now, but there's no ETA on when they'll be delivering a fix for end users.

That said, I have a 7970 GHz Edition at home and love it :)
Click to expand...

Let me ask this question, Can I possibly have an AMD and Nvidia in the same computer? I have a gtx670 right now but a 7970 as a solo render card could be a nice addition to my computer. What do you think?
 
It's definitely possible to run both in the same machine as far as I know. With RCX, you can only use one or the other for the time being, though.
 
In preliminary tests, how much of an advantage is CUDA going to give users? 10% over vanilla OpenCL? 20%? 50? 99%? OH MY GOD... 400%?

...but seriously though, will it make enough of a difference for REDusers to want to go with a CUDA card over any other?
 
Brian M Lynch said:
Is the Titan going to be a significant speed increase over the 7990? Would just like to get somewhere close to real-time 4k, so saving a few hundred bucks by getting a 7990 would be ideal.

Anyone have any experience with both cards and this release yet?
Click to expand...
I'm getting a little faster than realtime (about 26fps) from 4KHD to FullHD ProRes with full debayer on a GTX580. So anything that's about as powerful (or more powerful) should be able to get you realtime.
 
Grant, thanks for the info -- that's good to know. I think for the time being I might get the inexpensive 7970 GHZ edition as you stated, then grab another one later when multi-gpu support is added. Anything is an upgrade over what I have now (Gtx 280).

Jaakko -- thanks for your input too, that's very good to know!

Edit: Just been doing a bit more research -- the R9 280x seems to be the same card as the 7970 ghz edition, for a bit cheaper. Would this be a better option?
 
Last edited:
Decklink HD-SDI, got it to work, I did not know that we had to press the monitor button. I guess it's a good idea to read the manual.

However: I only get half size output. Probably 720p, the Resolution drop-down on the Video-out tab in Perferences are blank...

Any ideas?
 
I tried installing the beta on Window and it would not launch.
This is on a HPz800 with nVidia Quadro 5800.

Seems strange I can't get neither the Mac or PC betas to work
 
Jarred and Red Team,
The GPU performance is quite good in RedCine-X. How will this performance translate to Premiere when Adobe implements the new features from the SDK?

For example, RedCine-X debayers frames to memory ahead of time so that playback is smooth even with a full debayer (at least until the cache is empty).

Will Premiere be able to use this feature or is this RedCine-X functionality only?

Also, will number of simultaneous frames or allowable memory size set in RedCine impact anything in Premiere?

Thanks for listening
 
You must log in or register to reply here.
Back
Top