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 :)

REDCINE-X Beta CUDA Build 23.0.0 and Release Build 22.2.0

paul schefz said:
not sure it is that simple....first tests are coming out...like these from barefeats...not saying you are right or wrong...

http://www.barefeats.com/tube05.html
http://www.barefeats.com/tube08.html
http://www.barefeats.com/tube10.html
Click to expand...

I'm basing myself on Resolve and PPro tests done by guys over at Lift Gamma Gain. Many other tests use the stock GPU cards that came with the old Mac pros (which is ridiculous - you could get a better GPU card for under $300).
 
For me, performance on both the release version (OpenCL) and the beta version (CUDA) is below the last beta (Open CL).

On the last version, regardless of the debayer setting, all of my CPU cores went full-tilt. Now, with 5K footage, they run at 100% for 1/2 and full, around 50% at 1/4 res, and around 25% for 1/8 res. As a result, I can't get real time playback at any resolution with the current releases, whereas the previous OpenCL beta allowed me to playback 4K at 1/2 res and 5K at 1/4 res with room to spare.

Is this resource scaling expected behavior, or should I start sending logs?

Specs for reference:
i7 3820 @4.3GHz
GTX 770 4GB
16GB RAM (4x4GB)
Win7

RC-X OpenCL/CUDA set to 8 frames and 12GB.

EDIT: I also tried this (for kicks) on my y500 laptop (i7 3630QM and GT 750M 2GB, Win 8.1). Same CPU performance scaling.
 
Last edited:
I have 2 GTX Titans installed (just added the 2nd) and get an error when trying to use both. The error in the status pane is "RedImageExporterToExternal::cancel failed to receive stop message from client". This happens immediately upon clicking the Export button. Using RCXP 22.2 in OS X 10.9.1. I tried it both with no display and with a display connected to the 2nd Titan. Thoughts?

EDIT: Nevermind. I had aborted an export earlier and even though I had the Overwrite option selected for the export preset, it wouldn't overwrite the AAF creating folder. I deleted the folder and all is good.
 
Quick question(s) -- in RCXp 22.3 (latest OpenCL release build) you can choose to "check" both integrated intel graphics (4000 or Iris) as well as dedicated GPU (650m or 750m) on Retina Macbook Pros. Does selecting both mean it uses both? If it only uses one, which does it default to?

Another thing; the Integrated GPU has 12 OpenCL compute units compared to the dedicated GPU's 2 compute units... Is there potential for more performance with more compute units?
 
Last edited:
Mike, the Iris Pro should have more OpenCL power compared to the 750m...but I don't know the performance difference on these...if you do some benchmarks let us know!!!
 
Tommaso Alvisi said:
Mike, the Iris Pro should have more OpenCL power compared to the 750m...but I don't know the performance difference on these...if you do some benchmarks let us know!!!
Click to expand...

Anecdotal tests show the obvious -- Currently Iris seems to be about half the speed of the 750m... I'm more interested in taking advantage of BOTH simultaneously to eek out even more OpenCL performance.

EDIT: Forgot to mention, it seems that when they're both selected, it's slower than when just the 650/750m is on, but still substantially better than just Iris (I didn't compare to just CPU). Also my mind playing tricks.
 
Last edited:
I doubt you'd get very good performance on any integrated graphics. The integrated graphics using CPU power to function, so using your integrated graphics the GPU acceleration/debayering is going to significantly hurt the performance of other CPU functions. You'll always get better CPU performance and in general when using a discrete graphics chip. This is why you're seeing significantly better performance from the 750m by itself vs combined with integrated graphics.
 
Mike P. said:
Anecdotal tests show the obvious -- Currently Iris seems to be about half the speed of the 750m... I'm more interested in taking advantage of BOTH simultaneously to eek out even more OpenCL performance.

EDIT: Forgot to mention, it seems that when they're both selected, it's slower than when just the 650/750m is on, but still substantially better than just Iris (I didn't compare to just CPU). Also my mind playing tricks.
Click to expand...

In REDCINE X 22.3 Prefs what is recommended setting for 'Memory Limit' and 'Number of Simultaneous Frames to Process?' The default setting is 2GB (memory) and 8 simultaneous Frames.

If running a maxed out MBP (newest) 2.6 Ghz, 16GB, 1TB Flash, 750m coupled with a Rocket in a Thunderbolt Sonnet Express what settings are recommended for optimal performance?
 
Mike are you sure about that? I was under the impression integrated graphics are essentially separate circuitry that piggy-backs on the CPU die and does NOT in any way reduce CPU performance (it may generate more heat, which ends up throttling things down, but IGPs on the CPU die or northbridge do not significantly eat at up CPU cycles). There are a multitude of reasons why the dedicated card would perform better; it has dedicated ram, instead of shared with system memory (arguably the biggest), faster clock speeds, better architecture/more efficient performance, etc. My question is more concerned with doubling down with both the IGP and the GPU; using both to get an additional 5% or 10% (or 20%?) playback/rendering performance boost.

Dan, I'm not sure what's ideal/recommended but I'd go ~75% of your RAM (12GB) and 8 frames simultaneously (4core CPU with HT = 8... I've read that stability decreases when you run more simultaneously frames than you have thread processors.)
 
Mike P. said:
Mike are you sure about that? I was under the impression integrated graphics are essentially separate circuitry that piggy-backs on the CPU die and does NOT in any way reduce CPU performance (it may generate more heat, which ends up throttling things down, but IGPs on the CPU die or northbridge do not significantly eat at up CPU cycles). There are a multitude of reasons why the dedicated card would perform better; it has dedicated ram, instead of shared with system memory (arguably the biggest), faster clock speeds, better architecture/more efficient performance, etc. My question is more concerned with doubling down with both the IGP and the GPU; using both to get an additional 5% or 10% (or 20%?) playback/rendering performance boost.

Dan, I'm not sure what's ideal/recommended but I'd go ~75% of your RAM (12GB) and 8 frames simultaneously (4core CPU with HT = 8... I've read that stability decreases when you run more simultaneously frames than you have thread processors.)
Click to expand...

Thank you!
 
Jarred,

do you think it's possible to give us the same behaviour as Camera Raw and Adobe Apps sliders in general where if you double click it it returns to the default value?

It would be really nice and intuitive I think.

Guys?
 
Gabriel Scindian said:
Whoa!!! Beta 23 on my Win 8.1 Rampage Extreme 4 3930K 64gb Ram with raid0 512ssd and RevoDrive x3 with Gtx Titan and GTX 690 5K playback at 24fps is very quick. Thanks
Click to expand...


Seriously I gotta wonder, do you need so much gfx powerz? A titan and a 690.... :/
 
I'm looking for information on the timeframe for the release of the GPU accelerated SDK. Is it rather weeks, months, years?

Maybe it is somewhere in the thread but I missed it.

We are building a grading suite based on Davinci Resolve and if the SDK is close we would rather invest in more GPU power that CPU.

Any info is much appreciated
 
I'm building a new machine, handling RED footage and rendering is the main-task. So, the big questions in regard of choosing a GPU are: NVIDIA vs AMD? GTX770 (4GB) vs R290X (4GB)? Two GTX770 vs one TITAN Black?

Any thoughts? ;)
 
Maik Müller said:
I'm building a new machine, handling RED footage and rendering is the main-task. So, the big questions in regard of choosing a GPU are: NVIDIA vs AMD? GTX770 (4GB) vs R290X (4GB)? Two GTX770 vs one TITAN Black?

Any thoughts? ;)
Click to expand...

I'm on a Titan and I don't think REDCine is using a fraction of its power. Bottleneck is still CPU on a 3930k.
 
Maik Müller said:
I'm building a new machine, handling RED footage and rendering is the main-task. So, the big questions in regard of choosing a GPU are: NVIDIA vs AMD? GTX770 (4GB) vs R290X (4GB)? Two GTX770 vs one TITAN Black?

Any thoughts? ;)
Click to expand...

If you are building a machine I assume you are on Windows and you have alot of options. Its probably best if you need it right now to go LGA2011 sockets and get 2 of them if you can. 2 CPUS and 2 Graphic cards is a whole new experience...
 
Gavin Greenwalt said:
I'm on a Titan and I don't think REDCine is using a fraction of its power. Bottleneck is still CPU on a 3930k.
Click to expand...
Thanks Gavin, good to know.

Jarred Land said:
If you are building a machine I assume you are on Windows and you have alot of options. Its probably best if you need it right now to go LGA2011 sockets and get 2 of them if you can. 2 CPUS and 2 Graphic cards is a whole new experience...
Click to expand...
Thank you very much for your input Jarred. Yes, you are right, i'm on Windows and i'm really considering going for a dual-cpu system this time. Not exactly sure about 8core 3.4ghz vs 10core 3.0ghz yet. GPU-wise, i'll probably go for a TITAN Black, and add another one later. (...and keep space free for a ROCKET-X, hehe)
 
Last edited:
You must log in or register to reply here.
Back
Top