Activity › Forums › Creative Community Conversations › CC-CS6-X render comparisons
-
CC-CS6-X render comparisons
Jason Van patten replied 12 years, 10 months ago 17 Members · 38 Replies
-
Bill Davis
July 15, 2013 at 6:27 pmI’d kinda bet that Grand Central Dispatch is also in play here.
Makes me wonder if they’re using that to divy up tasks to send to the multiple GPUs in some intelligent fashion, rather than just sequestering one GPU for background rendering and leaving the other for timeline calculations.
But that’s just a guess.
Know someone who teaches video editing in elementary school, high school or college? Tell them to check out http://www.StartEditingNow.com – video editing curriculum complete with licensed practice content.
-
Marcus Moore
July 15, 2013 at 6:28 pmApple has in the past held back features from hardware where it doesn’t feel the the user experience would be a net benefit- even though it’s “technically” possible to run a feature on hardware X, Apple says the experience doesn’t meet their threshold for “good”.
For example, see how certain features at staged from one iPhone model to the next. Even if you can run iOS7 on your iPhone4 some features aren’t deemed compatible based on the speed of the device.
I think unrestricted background rendering would definitely qualify here.
-
Erik Lindahl
July 15, 2013 at 7:58 pmWhen doing OpenCL work on the GPU the system can run into extreme responsiveness issues. What a next gen MacPro could do an iMac could not is dedicate X cores + 1 GPU for background rendering and leave the rest for the users front-end. As soon as user interaction stops it will use all 12 cores and 2 GPU’s for rendering.
The above wouldn’t be possible on todays systems with out severely sacrificing user-interaction performance. The new MacPro virtually has about 2-3 iMac’s of CPU power and probably 4-6 of it’s GPU’s. This opens up a lot of possibilities.
So, background rendering with out affecting “normal” working conditions would be a huge selling point for many I would imagine. If Apple goes that route. They have stated dual GPU as standard and even lower end FirePro GPU’s will provide quite a lot of horse-power.
There’s also been talk of what i/o the MacPro will have. We know there will be six Thunderbolt 2 ports and one HDMI-port but we don’t know exactly what all these ports will support. Will the HDMI-port be able to work as a real video-out port at up to 4K and is that HDMI-port going directly off the GPU, i.e. no more monitoring lag? This could for example lessen the need for a dedicated i/o box over Thunderbolt – again a selling point for quite a few.
Summary: The above would be unique MacPro features unless Apple opens up GPU-support for Thunderbolt chassis. That said, the above would probably work relatively good in a current-gen MacPro given you have high-end GPU’s. The next-gen will boot a lot of VRAM which helps for GPU-processing work (Adobe requires a minimum of 1GB for example).
-
Oliver Peters
July 15, 2013 at 8:25 pm[Mathieu Ghekiere] “have you sent feedback about that to Apple?”
Yes.
– Oliver
Oliver Peters Post Production Services, LLC
Orlando, FL
http://www.oliverpeters.com -
Chris Kenny
July 15, 2013 at 8:44 pm[Gary Huff] “Exactly how? What will FCPX be optimized for on the Mac Pro that it currently doesn’t support on a high-end iMac? Outside of dual GPU, which I assume would be included as part of the OpenCL framework anyway.”
OpenCL certainly lets you use two GPUs, but I don’t think this is entirely automatic — you’d have to set up a second dispatch queue, at least. And OpenCL supports reserving portions of GPUs for its own use, which you’d probably want to do differently on a dual GPU system. So there’s that.
Apple’s wording also almost makes wonder if maybe the HDMI port on the machine can actually be used directly for broadcast monitoring, and FCP is adding support for that. (Though I’d consider this a bit unlikely.)
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
-
Ronny Courtens
July 15, 2013 at 9:11 pmMy own tests with the software have Premiere (CS5.5 and CC) handling my tasks in literally 1/2 the length of the clip. For FCPX? Twice the length of the clip. The main difference between my tests and yours: I’m not starting with, nor transcoding to/from ProRes. I’m starting with AVCHD and ending with h.264 MP4.
Nope. We have had this discussion on another forum. The most important difference is that you did your tests using an upclocked nVidia GTX570 GPU that is great for CUDA but lousy for OpenCL. So it’s quite normal that PPro performs better than FCPX on your system. As another poster pointed out you are also running FCPX on a fast but older MacPro:
“Aside the OpenCL GPU FCP X is much faster on Ivy or Sandy Bridge processors since the code makes good use of the AVX on those processors. Older Mac Pros just don’t cut it (no matter how many cores) like the new iMacs and MBPs. I’ve noticed this especially with H.264 material.”
So with all due respect you are comparing apples with lemons. Oliver has done his tests using different cards, both for CUDA and OpenCL. That’s why I think his comparison tests are reliable. As I have explained we have worked with native XDCAM HD 422 footage at the Olympics using FCPX on several new MBPs and we got faster than realtime exports. So this has little to do with ProRes or native footage.
-
Marcus Moore
July 15, 2013 at 9:44 pmIf not for Monitoring, what could the rationale behind an HDMI 1.4 port be?
Granted, even if it does give users a 4K display out, it likely won’t be compatible with some 3rd party products like Resolve, which still require a BlackMagic I/O for monitoring from DaVinci.
-
Jason Van patten
July 15, 2013 at 9:53 pm[Ronny Courtens] “Nope. We have had this discussion on another forum. The most important difference is that you did your tests using an upclocked nVidia GTX570 GPU that is great for CUDA but lousy for OpenCL.”
Read the whole thread Ronny. Specifically the OP’s machine specs. He’s using a single 5870 card in a MUCH SLOWER SYSTEM than mine. Do you think that ancient AMD card is enough to make that much of a difference? (Answer: it isn’t).
His results are so good because he didn’t transcode. Period.
[Ronny Courtens] “As another poster pointed out you are also running FCPX on a fast but older MacPro:”
So is the OP. Your point?
-
Chris Kenny
July 15, 2013 at 9:54 pm[Marcus Moore] “If not for Monitoring, what could the rationale behind an HDMI 1.4 port be?”
It could just be feeding a typical ‘computer’ video signal. That’s what the HDMI ports on Apple’s other systems do. Of course if Apple adds 10 bit monitor support in Mavericks, makes sure the system’s color management doesn’t get in the way, and has sufficiently high-quality software de-interlacing in FCP X, a ‘computer’ video output could actually be a viable alternative to a ‘real’ broadcast video interface for most purposes. Whether this will happen is anyone’s guess. The fact that in the keynote they went from discussing triple 4K support to showing a photo of a guy with three monitors — one of them being a broadcast monitor — sure is suggestive, though.
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
Reply to this Discussion! Login or Sign Up