Gpu rendering Is here
Rob Redman gets to test Nvidia’s GPU powerhouse, and is left impressed
Hands-on with Nvidia’s VCA powerhouse
Nvidia’s VCA is a rendering system in a box. Essentially a box full of Quadro graphics cards, but in reality far more. You may have tried stacking a handful of PCI cards in your workstations in a bid to speed up rendering, but there are limitations to this in the form of compatibility issues, heat and power control, as well as the inherent cap on card count. This is where VCA succeeds. As a one-stop solution the chassis and underlying architecture are what counts, supported by a minimum of 192GB ram and 16 threads of CPU power. The box itself runs linux and comes pre-installed with V-ray RT and Iray. Lee is chief technology officer at Escape Technology and has over 25 years of experience in the high-end graphics business within M&E. He was awarded Maya Master status in 2007. www.bit.ly/208-danskin “It’s fast,” exclaims Escape Technologies’ CTO Lee Danskin. “I’d love to tell you more, but when you’re talking about the VCA that’s the bottom line: it’s incredibly fast.” Running V-ray RT in Maya 2016 you can get an almost real-time render at full quality of even dense, multimesh scenes, with full GI, blurry reflections and multiple transparent shaders. Once sent to the VCA the render itself is blisteringly fast. One example was a scene with 18 million polygons, which rendered clean in under 10 seconds.
A large benefit of using VCA is the scalability of its architecture. You use the network based manager to tell it what resources you want to use and that’s it. You can share the system between different artists, teams, or even locations, and as it’s server based you can share the output with anybody who has an internet connection.
Build your own solution
Nvidia’s VCA could become a service, replacing the render farms we currently utilise This means you could, if you were running a larger studio, have multiple VCAS set up, which would be a massive reduction in costs and resources. On top of that is the fact that certain other CPU tasks can be handled by VCA; all those cores can be thrown at running simulations such as Bullet physics.
One possibility too, is the DIY solution. It’s possible for smaller houses and single artists to bypass the hefty financial outlay and self-build their own solution, which could conceivably reduce the cost from the £30k to £10k, which is actually not far from the cost of many high-end workstations.
It’s early days yet but with studios adopting VCA already it looks promising, especially as developers are turning more often to harness the power of GPUS, leaving the CPUS to manage the rest of the system.