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 :)

RED + NVIDIA Announce Realtime 8K R3D GPU Decoding

Misha Engel said:
A lot of people would love to know how you do that without any issues (I got it working with issues).
Click to expand...

Once you have a generic PC motherboard (I always use Gigabyte brand but others like Asus work) - with USB C I/O ports working in OS X, and NVidia Web Drivers installed - it works. I have RRX in a Sonnet eGFX box now running on a USB C/Thunderbolt 3 port. If people are having graphic card troubles with this config then I think that not having the latest NVidia Web Drivers installed might be the root of their problem with hardware recognition. On a hackintosh motherboard (logic board in AppleSpeak) you have to initialize the USB C/3.1 ports with Windows drivers from a Windows 10 installed hard drive, then they are active when OS X is installed to another drive or a dual boot system.

The NVidia Web Drivers are downloadable from their website and come in a OS X installer package.
 
Bruce Schultz said:
Once you have a generic PC motherboard (I always use Gigabyte brand but others like Asus work) - with USB C I/O ports working in OS X, and NVidia Web Drivers installed - it works. I have RRX in a Sonnet eGFX box now running on a USB C/Thunderbolt 3 port. If people are having graphic card troubles with this config then I think that not having the latest NVidia Web Drivers installed might be the root of their problem with hardware recognition. On a hackintosh motherboard (logic board in AppleSpeak) you have to initialize the USB C/3.1 ports with Windows drivers from a Windows 10 installed hard drive, then they are active when OS X is installed to another drive or a dual boot system.

The NVidia Web Drivers are downloadable from their website and come in a OS X installer package.
Click to expand...

Does this mean that you also got the NVidia card working in an eGPU box without attaching a monitor to the GPU in the eGPU-box?
 
Misha Engel said:
Does this mean that you also got the NVidia card working in an eGPU box without attaching a monitor to the GPU in the eGPU-box?
Click to expand...

Yes, using a 1080Ti as a CUDA rendering GPU for DaVinci Resolve in eGPU. I have a 980Ti in the computer case for GUI. No external monitors attached to eGPU box.
 
Bruce Schultz said:
Yes, using a 1080Ti as a CUDA rendering GPU for DaVinci Resolve in eGPU. I have a 980Ti in the computer case for GUI. No external monitors attached to eGPU box.
Click to expand...

Still wonder if it will work with a modern iMac(Pro) or MacBook(pro) with an AMD GPU inside, but I will give it a shot (with a GTX970), what version of MacOS are you using?
 
Everything Misha & Brian post seems true. I've had several MacVid hacked (they don't like that term but that's what they are) cards and they always work in OS X when the generic off-the-shelf version of the same card doesn't. Neither of the NVidia cards I'm using now are MacVid cards, but they are working fine in OS X Sierra 10.12.2.

Can't speak to the iMac issues, I've always found iMacs way too limiting in terms of hardware expandability and always build desktop MacPro style units.
 
Could we get a hint on what CPU is needed?

Currently I am deciding between a 1070 MaxQ or a 1080 GTX card in a laptop. Could possibly be worth waiting for Turing cards in laptop but that could take some time.

With current hardware I think the CPU will be the bottleneck so the 1080 would probably not give that much extra, but with the new optimisations it would probably be very different, or?
The 1080 is about 70% faster for calculations than 1070 MaxQ but of course needs a bigger laptop.

Best,
/Andreas
 
I just purchased AMD's Threadripper 2950x for my CPU. I believe it is going to help with my Quadro RTX card ;)
 
A heads up to Mac users wanting to have access to the newer NVIDIA GPUs.
Apple is essentially blocking NVIDIA from releasing updated drivers on the latest OS X Mojave.

https://www.macrumors.com/2018/11/01/nvidia-comment-on-macos-mojave-drivers/

This situation has been in limbo for the last month and a half.

For concerned OS X users, when you get the "Apple sucks" rant out of your system take the time
to sign a petition like below and perhaps hit also Apple at the email links listed on the page.

https://www.change.org/p/tim-cook-a...-work-with-nvidia-on-drivers-for-mac-os-10-14

Takes less time than typing a rant and more effective.

Brian Timmons
BRITIM/MEDIA
 
By the time this RED+NVidia software is out of beta you can get a CPU for less than a $1,000 that can do the same so your cheap non RTX Quadro cards can fully focus on the de-bayering.

For the time being you're stuck to atleast an overclocked i9-9760x for $ 1,600 or 2970wx for $ 1,300 and some hustle with process lasso to de-crypt, -compress and -code ALL current flavors of .R3D in real-time.

When this new software works on 1/2 or 1/4 resolution in real-time with 8k it is still an added value for laptop users.

For Resolve users it makes no difference wheter you use NVidia or AMD cards, they more or less have the same speed.


Lets also start a petition that MacOS works on ARM, Power9 and RISC-V processors.
 
Any updates on this new cool tech. I'm wavering between getting an MSI laptop with 1080 card or an Alienware with 1070 QT. 1080 is about 70% faster but laptop is quite a bit bigger as well.
Cheers
/Andreas
 
AndreasOberg said:
Any updates on this new cool tech. I'm wavering between getting an MSI laptop with 1080 card or an Alienware with 1070 QT. 1080 is about 70% faster but laptop is quite a bit bigger as well.
Cheers
/Andreas
Click to expand...

December... release is tied to Nvidia’s RTX schedule.
 
It is genuinely very exciting what is happening here... ...but I do hope we see a really useful implementation/execution off this approach/technology from a gradeing POV ...

I do worry about completely loading up GPU with de-wavelet duties. Tools like resolve are (out side of compression related duties) completely dependent on GPU... resolve can run out of VRAM memory even on on a 12gb card. Therefore I worry about how much GPU memory is left post decode... this may make the coming gforce versions less useful if doing proper colour work... but obviously great for rushes dev!?

Big question for me: Will 2 identical GPUs work together in parallel or series with new SDK ? One card decodes then hands over to the other for colour work? If it has to push back through the mother board rather than over a Nvidia bridge then this is probably less ideal...or will it be in Parallel? With both cards decoding and doing effects?.... this is probably more likely? But then when you do Temporal NR things fall apart?... TNR can only ever use one card and it needs lots of memory!! So will enough computation ability and memory be left over after the decode? This is a big question that needs asking?

Anyway I hope there is a user visible switch so we can pick which approach works best for us?... at least until this is both super stable and super lean ?
 
nVidia is promoting a multi GPU architecture called NVLink that might be well suited to multitasking situations like Michael notes. I'll leave it to proper propeller heads like Jeff K. to address what is, and is not, supported. In my head, I keep thinking one GPU for decrypting/debayering - another for computational work - and a third for driving the hero display. No idea if that makes any sense in a modern workstation with HBM/fabric/etc. Count me curious to know more...

Cheers - #19
 
With our implementation you can select which card or cards you want to do GPU decode with, and the people that implement the sdk can declare that as well.

For those concerned about memory saturation, Nvlink does allow memory combining across multiple
Cards but only on Linux right now.
 
Thanks Jarred...Great to hear!

Let's hope memory combining comes to windows...

Jarred Land said:
With our implementation you can select which card or cards you want to do GPU decode with, and the people that implement the sdk can declare that as well.

For those concerned about memory saturation, Nvlink does allow memory combining across multiple
Cards but only on Linux right now.
Click to expand...
 
You must log in or register to reply here.
Back
Top