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 :)

IBC - GPU Build REDCINE-X Beta 21

You want as many CPU cores as you can get so they can keep the graphics card busy
 
We are attempting to use a AJA Kona-3G card for dual-link 1080 output over SDI. Fast dual Xeon system with 64GB RAM and Titan with 6GB RAM.

Beta 21 doesn't seem to like that combo. Has anyone tried something similar with success?

Cheers - #19
 
Blair: what does "doesn't seem to like that combo" mean? What sort of issue are you having? Any errors?
 
Keep getting errors on a macbook pro mid 2012 with the 512mb version of the nvidia 650m :( I guess 768mb is the minimum. Damn, should have gotten the higher model
 
Wow, new RCX is way faster. My laptop that is a 1.5 years old that use to playback real time in 1/8 resolution, does 1/2 resolution with no problem. This is awesome!!!!!

Thank you RED Team!
 
Rob Lohman said:
You want as many CPU cores as you can get so they can keep the graphics card busy
Click to expand...
Interesting. Maybe I'll need to look at getting a new processor as well as a new GPU, then. I currently have an i7 3820 (4 core/8 thread) and a GTX 570 2.5GB. Running 4KHD files at full res got me 12fps, with the CPU slammed at 100%, but the GPU at 30-40%, and ~900MB of VRAM used.
 
You can use a Rocket with GPU, in the sense that any image processing (like effects) the Rocket can't do the GPU will do.
 
Saumene Mehrdady said:
Keep getting errors on a macbook pro mid 2012 with the 512mb version of the nvidia 650m :( I guess 768mb is the minimum. Damn, should have gotten the higher model
Click to expand...

Does it work at half res or below? It would be good if you could email us a verbose log file: redcinex at red

That is not a lot of memory for full-res, that's for sure. I would say 1GB is sort of minimum if you want to have a nice experience.
 
Rob Lohman said:
Does it work at half res or below? It would be good if you could email us a verbose log file: redcinex at red

That is not a lot of memory for full-res, that's for sure. I would say 1GB is sort of minimum if you want to have a nice experience.
Click to expand...

Playback is fine, I can play 1/4 res smoothly and mostly realtime now instead of 1/8, which is a great improvement.

It just gives me a memory warning in the system settings menu and also in the status tab at the bottom (next to audio, markets, export, etc) and when I try to export a clip in the viewer to 422 it also gives me an error, and the file comes out corrupted. The effective render time estimate is way, way lower than versuon 17.5 though, which is awesome.

Im going to keep playing around with it for the next day or two
 
Does anyone know if a flashed MacVidCards Titan or 780 will work good in RedCine-X on a 8 or 12 core Mac Pro?
 
Rob Lohman said:
You want as many CPU cores as you can get so they can keep the graphics card busy
Click to expand...

Big improvement over the last version! , but seems likes your limited to 32 cores or 2 numa nodes max utilization.

With 4x e5-4650 procs, 64 HT'd cores, only 32 get used, and 2 numa nodes.

Not a disk i/o bottleneck.

Setting for simultaneous frames to process has no effect.

Also, 1/2 resolution decode utilizes the 32 cores it is accessing efficiently, but full resolution decode barely uses 15%.

It does nicely max out 256gigs of ram for playback now.
 
OS X doesn't support the Titan. Hackintosh is not gonna cut it guys.



Christopher: you wanna send us that system to debug? :) We don't have a 64-thread system, but I promise we're not artificially limiting it to 32.

Your limit is likely coming from somewhere else (assuming there's no bug). What graphics card do you have in your system? You'd probably have to go to a multi-GPU setup (once RCX supports that) so the GPU card is not the bottleneck for the decode (assuming you have a Titan or something in there). You can use GPU-Z to see how much GPU is in use.

You mention it's not disk I/O, but how much bandwidth do you have? Depending on your clip you could easily need 1GB/s of bandwidth to keep 64 threads busy
 
You must log in or register to reply here.
Back
Top