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 :)

DaVinci Resolve 7.1 Update

Put a x16 GPU in an x8 slot and see what the performance difference is. I have done a little testing and it was 0 difference. I have read benchmarks and they found minimal to no speed increase. Most video cards do not use the full bandwidth. Maybe the latest and greatest cards do, but do some research and test for yourself.

I am very happy with my Cyclone 600-2707 PCIe expander. I choose it over all the others. Now I need to get Resolve and a couple more video cards.


Dusty
 
Two FX4000 will use full 16x2=32 lanes. That leaves only 8 lanes for GT120, Decklink or RR. Unless I'm missing something, I don't see how this is going to work on todays Mac...

Very few cards actually use the full 16x pci-e 2.0 bandwidth which is 8GB/s. Big B not little b. Try experimenting by exchanging cards and do a battery of gpu centric tests. While I don't have a Quadro card to test I've seen no difference in the radeon 4870 or gtx285 between 8 or 16 lanes. A 16x card will require a 16x electrical and form factor but most expansion boxes include 16x slots even if not providing more bandwidth. How many gpu cards will saturate a 16x channel I don't know but my feeling is that for the cuda multiprocessing you can use 4 cards easily.

-noel


Edit: Oops I left the browser unrefreshed and loads of info from others who answered this question better than me.
 
Great update, the most important update I was hoping for from a Red workflow perspective is lacking: AUDIO RENDER

If anybody from BM is listening:

WE NEED AUDIO RENDER TO AVID CODECS AND QUICKTIME CODECS FROM R3D FILES

I am using Resolve with Redrocket amongst finishing work to CREATE DIGITAL DAILIES. If my dailies have sound, I have to do crazy workaround work to extract and reconnect sound in other software. This makes it as a dailies tool way too inefficient.

The render files NEED TO PASS THROUGH THE AUDIO FROM r3d files otherwise the usefulness of Davinci in a Red workflow is greatly diminished. It seems obvious but I have mentioned this in various forums and tech support emails and all I've gotten is "it's a finishing tool". I have a vision for this tool as a film-less, tape-less "telecine" powerhouse, returning the intricacy and quality of my dailies back to that level...

Please some listen and implement this.

thanks!
 
Hi Florian, thanks for your feedback. We do have this feature request on our list but as you may see from the 7.1 release we had a packed schedule of features already in development since IBC. Please keep the ideas coming. 'Dailies workflows' is certainly something we have a great deal of requests for. We can't address them all as quickly as we want but will do our best.
Peter
 
Hi Peter,

this is a absolutely fantastic release. Can you confirm whether your adding support for the Euphonix Color Panel and any AJA hardware?

Thanks so much for your hard work!
 
Last edited:
Very few cards actually use the full 16x pci-e 2.0 bandwidth which is 8GB/s. Big B not little b. Try experimenting by exchanging cards and do a battery of gpu centric tests. While I don't have a Quadro card to test I've seen no difference in the radeon 4870 or gtx285 between 8 or 16 lanes. A 16x card will require a 16x electrical and form factor but most expansion boxes include 16x slots even if not providing more bandwidth. How many gpu cards will saturate a 16x channel I don't know but my feeling is that for the cuda multiprocessing you can use 4 cards easily.
I'm sorry, but I don't have time to experiment, I'm kinda busy, working:-)
As you said it yourself, you don't know, if two FX4000 cards will saturate 32 lanes of bandwidth. If you don't, then you can't claim, that it will work properly. If you still insist on trying, then I, personally, would rather go the Hackitosh route, where there are more than 40 lanes available, on the same hardware as Linux Resolve and you can run multiple GPUs properly, with no expansion chassis needed. BTW, someone had mentioned, that multi GPU on Linux is the same as on Mac. Unfortunately, it doesn't, for the reason I just had mentioned.
Now, with new Intel architecture in the upcoming Macs it's different story altogether...
 
does anybody know if you can mix and match gpu's? gtx 285 in the mac pro, expansion chassis with a quadro 4000 in there? will it harness the power of both cards, or does it have to be identical gpu's?
 
I'm sorry, but I don't have time to experiment, I'm kinda busy, working:-)
As you said it yourself, you don't know, if two FX4000 cards will saturate 32 lanes of bandwidth. If you don't, then you can't claim, that it will work properly. If you still insist on trying, then I, personally, would rather go the Hackitosh route, where there are more than 40 lanes available, on the same hardware as Linux Resolve and you can run multiple GPUs properly, with no expansion chassis needed. BTW, someone had mentioned, that multi GPU on Linux is the same as on Mac. Unfortunately, it doesn't, for the reason I just had mentioned.
Now, with new Intel architecture in the upcoming Macs it's different story altogether...

I do agree a hackintosh is a viable route if you know what u r doing. I keep meaning to learn the intricacies of this myself but until I have mastered it I wouldn't want to try for something for pro use.

I strongly disagree about the gpu expander though and think that any limitations of mac+expander vs. linux is just not about PCI-E bandwidth. If BM own tests show the GTX285 performing on par if not better than the new Quadro then it seems to me highly unlikely that they are saturating the 16x lanes.

As often happens in technology, various components leapfrog each other in terms of capacity and performance. Hence the term bottleneck. External I/O is a good example of how performance can be limited - we all know how important esata is despite its outrageously poor connection design and lack of power. But just as even the fastest SSD is just now coming on par with esata's theoretical 3Gb/s capacity, GPU cards have and will continue to lag far behind the interconnect.

BTW congrats on a busy schedule - for myself this is the slow time of year and after two features and a bunch of other stuff now might be the time to jump into the hackintosh world. But if the new macs are even close to what we all hope that they will be I won't be crying over a missed opportunity. But sadly I'm getting resigned to the fact that the Apple of my heart has sold herself to become a mass consumerist megabeast akin to the bad old microsoft days.

-noel
 
I'm sorry, but I don't have time to experiment, I'm kinda busy, working:-)
As you said it yourself, you don't know, if two FX4000 cards will saturate 32 lanes of bandwidth. If you don't, then you can't claim, that it will work properly.

8 GB/s is a lot of bandwidth. It's hard to think of that Resolve could be doing that would come close to using it. We're not talking about a high-end 3D game that might have to load huge textures from main memory in a fraction of a second when the user whips around a corner. Resolve is processing frame data that comes it at a nice reliable rate -- under 250 MB/s when working in 2K. Even if you take into account that the data has to be transferred both in and out, and factor in some absurdly generous overhead, this probably uses well under a quarter of the available bandwidth provided by a 16x slot.
 
As you said it yourself, you don't know, if two FX4000 cards will saturate 32 lanes of bandwidth.

There is a number of configurations in the Config Guide that comes with the 7.1 installer download. Some of which have three FX4000 cards in the Cubix expander or two FX4000's and two Rockets.

My guess is that Blackmagic tested these setups and it works.
 
My guess is that Blackmagic tested these setups and it works.

There is plenty of bandwidth on a Gen 2.0 x16 PCI-E link.

On a setup here, we are using 2 GTX285s and 1 Red Rocket in the Cubix expansion chassis, 4096x2304 Red Half-res premium decode, and more than 20 nodes of grades at 24fps. We also have a Decklink Extreme 3D+ and a ATI 5770 for GUI monitoring in the Mac Pro. There is still a slot open for a FC card in the Mac Pro.
 
Just a question: are there plans for a configuration tool for the Wave panel?

While it is nice that now so many functions can be accessed through the Wave, it also gets very 'deep' in the sense that you have to press the top right arrows many times in order to get to the menu that you need. And I am still trying to find out what the F-keys are doing...
 
Just a question: are there plans for a configuration tool for the Wave panel?

While it is nice that now so many functions can be accessed through the Wave, it also gets very 'deep' in the sense that you have to press the top right arrows many times in order to get to the menu that you need. And I am still trying to find out what the F-keys are doing...

Try holding both the up and down buttons together.
 
Try holding both the up and down buttons together.

Oh, hey, that's cool! (This brings up a master menu where you can choose between the different menu sets)

Now the menu sets make more sense to me as well.

Another hint on the F-keys?
 
A great update indeed!

Is anybody having issues with resizing, softening, moving their power windows (except with circular - this one works as it should) when you Alt + Add Node With a Power Window (Rectangle etc etc). This doesn't seem to be the case when you add them separately. This is when using the wave panel.

Audio from the r3ds would be great!
 
will the multi-GPUs and multi-Rockets increase the transcoding of files into ProRes, DNxHD and others? If so, how much? Anyone tested this yet?

As for audio syncing, Resolve should have audio syncing to the double system WAV files done properly, and not just rely on passing sound through from the in-camera RED files.

We need to be able to pass soundroll and sound TC through to ProRes (XML) or DNxHD (ALE) files.
 
PROBLEM

PROBLEM

OK, installed the upgrade and it stopped to recognize the quadro fx 4800 card, just the GT120. I'm on a 2009 nehalem macpro with no issues before this upgrade.
Any ideas?
 
Back
Top