The biggest hurdles when dealing with high-resolution content such as 8K is usually when it comes to playing it back and working with it on a computer. Capturing 8K is the easy part, but you need some pretty serious computing muscle to deal with it in post. So RED have teamed up with NVIDIA to try and ease the pain. Using NVDIA’s Turing GPU architecture and Quadro RTX GPUs, content creators dealing with 8K footage no longer have to be slowed down and hindered by poor playback performance.
An 8K camera captures at 8192×4320, or more than 35 million pixels per frame. If you break that down, five minutes of 8K footage captured at 24 frames per second equates to 250 billion pixels. If you have hours of footage for your project that number will jump past 100 trillion very quickly.
To handle 8K resolutions you need to use very powerful, expensive workstations, and high-end custom hardware. Even with all of this technology, you can still run into bottlenecks. Software and NLE solutions such as Adobe Premiere Pro, Blackmagic’s DaVinci Resolve and Autodesk Flame are already capable of working with 8K footage, but depending on the processing power of the computer, a lot of people end up viewing their 8K files at significantly reduced resolutions. Attempting to play back the footage at full resolution can cause the application to drop frames or stop playback while buffering, forcing you to choose between smooth playback and working in full resolution.
Well, that’s all about to change. NVIDIA’s Turing GPU architecture which was announced back in August now makes it possible for anyone to work with 8K footage in full resolution in real time. Even playback of frame rates greater than 24 frames per second can be done on a single-processor PC with one Quadro RTX GPU. Not only is the performance good, but it comes at less than half the price of CPU-laden workstations, putting 8K within reach of more content creators. According to NVIDIA, “Turing is NVIDIA’s most important innovation in computer graphics in more than a decade.”
By offloading all of the computationally intensive parts of the REDCODE processing to a Turing GPU, NVIDIA and RED are giving post-production professionals access to 8K footage at full resolution in real time. And it’s not just for Turing, this acceleration will also substantially increase REDCODE processing performance on other NVIDIA GPUs. New capabilities will also be possible with the NVIDIA RTX Tensor Cores and RT Cores available with Turing. Editors will gain from new functionality like AI-enabled upscaling, which will let them intermix archival footage or zoom in beyond 8K resolution with the best possible results.
Last night RED and NVIDIA held an event at the Linwood Dunn Theatre which was the official reveal of this new technology. At the event, they demonstrated full 8K REDCODE RAW being played back in real time. On a system featuring two Titan RTX GPUs, they were able to playback 8K at 48-60fps.
Quadro RTX 8000 with 48GB memory: $10,000 estimated street price
Quadro RTX 6000 with 24GB memory: $6,300 ESP
Quadro RTX 5000 with 16GB memory: $2,300 ESP
All this tech does come at a cost. The Quadro RTX 8000 with 48GB memory is going to be around $10,000 USD and you are going to have to spend at least $6K to get real-time playback performance in 8K. If this all sounds too expensive for you, don’t worry, NVDIA is planning to release more affordable consumer cards soon.
The target release is December 2018 to NLEs for incorporation, but that could be pushed into early 2019. All NVIDIA GPUs running CUDA will see a significant improvement. Whether Apple’s rumored new MacPro will offer NVIDIA as an option will be interesting to see. Read more about that at the bottom of the article.
Do you have to spend $10K to achieve 8K 24fps playback performance?
According to RED, they have been able to achieve 8K 24fps real-time playback performance using a P6000 ($4,385 USD). They are confident that the lowest end Turing will still work well, and the even 1080 Ti ($699 USD) cards should see a massive improvement in performance.
How does this affect R3D transcoding?
All transcoding requires decoding of the R3D first, speeds will be greatly improved but the encode itself will take the same amount of time. Overall, the time required to export media will be significantly decreased.
Is RED allowing GPU-based wavelet decode?
Yes, they have enabled the option to move entropy and wavelet decode to the GPU as well as debayer.
Can you achieve real-time playback performance with effects layers as well?
According to RED, they have been able to achieve full real-time decoding and color grading at 8K. They expect there to be headroom in the decode performance to allow for effects layers as well in real-time, depending on your system specifications.
Apple and NVIDA, it’s hardly the garden of Eden
With Apple’s release of Mojave, a lot of professional users have complained that they experiencing slow performance rendering regressions using NVIDIA hardware when they upgraded from 10.13 to 10.14. Apple controls all the drivers for Mac OS, so NVIDIA cannot release a drive until it is officially approved by Apple.
This has lead to an online petition that was started by Hester Adams that states:
“The community of people who have chosen to use Nvidia products with their Apple hardware demands that Apple permit Nvidia to make web drivers that work with all Mac OS versions going forward. Being able to utilize the hardware acceleration provided by Nvidia is vital to many of us, as are the bug fixes and features introduced with the application and OS updates. These drivers have always been available, but Apple now blocks Nvidia from supporting OS 10.14 Mojave. We demand that Apple publicly commit to a support plan to work with Nvidia to provide continuous support for their customers.
By signing this petition, we are making our voices heard. If issues persist and we are feeling neglected, we will“:
Contact Apple Customer Support until it is fixed
Email Apple Execs
Tweet @tim_cook using hashtag #UnblockNVIDIA: https://twitter.com/tim_cook
Register bugs at https://developer.apple.com/ bug-reporting/
So far 4,376 people have signed this petition.