Virtual Reality (VR) has been courting AEC professionals for years. The technology is still inchoate, but it’s getting more refined by the day. Engineers, designers and architects are using VR and beginning to benefit from the novel perspective it offers their industries. It’s been a slow trickle, but the VR dam is poised to burst.

Recognizing this mounting pressure, graphics company NVIDIA made VR a key component of its latest GPU microarchitecture, Turing. Turing is the basis of NVIDIA’s new RTX graphics cards, encompassing both the professional market with Quadro RTX and the consumer market with GeForce RTX. RTX is shorthand for “real-time ray tracing,” a computationally intensive rendering technique that Turing can now achieve in real-time.

“In VR, much of the time you probably need to use ray tracing,” said Carl Flygare, Product Marketing Manager for NVIDIA Quadro at PNY.

That’s because ray tracing puts the R in VR by providing photorealistic renders. “In the past, for interactions like light passing through a glass windshield and showing the interior of a car, traditional rendering wasn't really terribly realistic,” Flygare continued. “If you use ray tracing it can be completely realistic.”

Automotive designers and their counterparts in other industries can clearly benefit from the enhanced realism of real-time ray tracing. But this eponymous feature of RTX cards is just one of the ways that Turing takes aim at VR applications. Let’s take a closer look at how Quadro RTX cards in particular are optimized for VR.

Quadro RTX: Shade for VR

The Turing microarchitecture enables four main shading improvements in Quadro RTX cards, all of which substantially benefit VR. The first is mesh shading, a new shading model which greatly improves the efficiency of geometry calculations. It removes CPU draw call bottlenecks by offloading certain tasks onto parallelized GPU programs.

The bottom line for VR is quicker graphics, which isn’t just a convenience—it’s a necessity. If there’s a latency of more than 20ms in VR applications, users are prone to motion sickness.

Another huge shading improvement in Quadro RTX cards is variable-rate shading (VRS). This feature allows developers to finely control the shading rate of their graphics, from once per 16 pixels up to 8 times per pixel. In this way, developers can focus the bulk of their graphics power to areas of the screen that are most important, and minimize the work needed elsewhere. This helps improve frame rate and allows designers to prioritize where they want the most detail.

“I'll use another automotive example,” Flygare said. “Maybe you're working on a supercar and you're using carbon fiber. And you're actually displaying the carbon fiber as part of the industrial design and you've set up a specific weave pattern, and you don't want that weave pattern to fall apart or start to show moirés. The ability to make this section of the scene a priority, to always apply maximum GPU resources to it, is a great thing.”

VRS enables an algorithm called foveated rendering that’s especially useful for VR applications. For head-mounted displays (HMDs) that support eye tracking, foveated rendering applies VRS to the areas of the screen where the user is directly looking. Much like how our own ocular system prioritizes that which is within direct sight, foveated rendering gives the highest resolution to the VR objects a user is actively focusing on. Today, few HMDs currently support eye tracking. Nonetheless, it’s clear why Flygare refers to VRS as “a standout feature for VR.”

A third shading improvement in Quadro RTX cards is texture-space shading. With this technology, pixel shaders sample from a private coordinate space (called a texture space) that’s saved to memory. This allows shading results to be reused and resampled.

“Texture-space shading improves shading efficiency so you can have more elaborate, more sophisticated textures for better realism,” Flygare explained.

Finally, Turing extends a feature first offered in the Pascal microarchitecture: single-pass stereo. This feature allowed cards to render two views, differing by an x offset, in a single pass. Now, Turing allows cards to render up to four views in a single pass, differing in as many coordinates as necessary. It’s called multi-view rendering, and it enables smoother VR experiences with wider fields of view.

“You can think of multi-view rendering as single-pass stereo on steroids,” Flygare said. “It can render up to four geometries and have them cached and ready to present, rendered, shading, lighting, all in a single pass. And it can also work for directional audio. So, it allows for a more fluid and immersive experience if you're moving around rapidly. It's not only rendering what you're specifically looking at, but making a best guess at the next part of the scene you may want to look at, and renders it as well.”

Memory, VirtualLink, and VR SLI

We’ve seen many ways in which Quadro RTX cards are suited for VR, but one of the biggest (literally) is the cards’ memory. The previous top-spec Quadro card, the Quadro P6000 (based on the Pascal microarchitecture), offered 24GB of GDDR5X memory. The top-spec Quadro RTX card, the Quadro RTX 8000, has 48GB of more efficient GDDR6 memory. Quadro RTX cards can also be coupled with the second-generation NVLink connector, so two Quadro RTX 8000s can be linked to offer users a massive 96GB of collective memory.

Memory is a huge boon for VR applications, and it sets Quadro RTX cards apart from their consumer counterparts.

“GeForce RTX has similar capabilities, but the fact of the matter is Quadros have a lot more GPU memory,” Flygare said. “And having the entire [VR] environment within GPU memory means they get maximum performance.”

Quadro RTX cards also support an emerging VR standard called VirtualLink, developed by an industry consortium including NVIDIA, Oculus, Vive, Microsoft, Valve, and AMD. VirtualLink defines a USB-C mode to transmit power, data, and display information to an HMD over a single cable. Most current HMDs require three or four different cables, making the user “Borg-like,” in Flygare’s words. VirtualLink simplifies HMD setup and enhances user comfort. No HMDs currently support VirtualLink, but Flygare is confident support is coming.

Quadro RTX cards also support VR SLI, where two cards can be used with a single HMD, one card per eye. “If you have two Quadro boards in the system, you can assign one to the left eye, one to the right eye, and start doing insanely high-resolution scenes and not notice any latency,” Flygare said.

Pushing for VR

The additions to its latest GPUs clearly reveal NVIDIA’s confidence in the future of VR. But the company hasn’t left it there. NVIDIA has also developed VRWorks, a collection of APIs, libraries and engines for developers to build the applications and hardware necessary for VR. VRWorks includes tools for professional VR applications as well as games and other consumer-oriented apps.

“NVIDIA has spent a ton of money and put a lot of engineering resources into SDKs and APIs to make it easy for third parties to build VR enabled apps,” Flygare said.

To learn more about the VR technology in Quadro and how it can accelerate design workflows, watch the PNY webinar on November 12^th: Take your Product Design to New Heights with VR. 

PNY Technologies has sponsored this post.  All opinions are mine.  --Michael Alba.