The Next Generation of GPUs are Upon Us

The modern VR revolution needed quite a lot of different branches of technology to reach a certain points before something like the Oculus Rift or Vive is possible. The main ones that we usually point to are smartphone components. The motion sensors and high resolution displays in modern premium VR systems all only here because smartphone market competition drove things in that direction.

What’s easy to forget is how important the improvement in GPU technology was at the time. When Oculus first announced the minimum requirements to make one of their headsets work, people were pretty shocked at how high the numbers were. You needed a high-end GPU costing nearly a grand just to get in the door.

It was really dismaying, but then something magical happened. Nvidia released the 1000 series of cards. These cards represented a major leap in both performance and power consumption. Now all of a sudden a mid-range GPU was fast enough to drive VR and used little enough power that even laptop could now easily be VR ready. I honestly believe that if Nvidia hadn’t brought out these superlative chips, then the adoption of VR would have been much slower. Now the company has begun reading a replacement for the 1000 series of cards. So what do we know so far?

Say Hello to Turing

Every GPU architecture generation from Nvidia is named after a famous name in science. Previous generations have been named Maxwell and Pascal for example. The upcoming cards have been named Turing, after the father of computing Alan Turing.

Turing is bringing a whole raft of new features to the GPU space and many of them are going to be very important for the advancement of VR.

The Quadro RTX Lineup

Nvidia’s Quadro cards are incredibly expensive professional workstation cards that are used for things like CAD design and 3D rendering, to name a few use cases. Apart from promising considerably more raw GPU power, we know that the Quadro RTX cards come with cutting-edge AI-enhancing hardware, VirtualLink, and, most importantly, real-time ray tracing ability.

Ray tracing is an incredible 3D rendering method that actually traces the path of simulated light rays through a scene. This method is commonly used for high-end CG films and effects. The problem is that it takes hours just to render a few seconds of footage this way. Nvidia has been working on this problem for a long time and Turing is bringing real-time ray tracing to the market.

From a VR perspective, the introduction of real-time ray tracing could be a game changer. It’s certainly set to be a game changer for desktop graphics as a whole. Unfortunately, these Quadro cards are not consumer hardware. You’re looking at anything between $2500 to $10 000. Which is par for the course when it comes to Quadro cards. For most people, the consumer cards based on Turing will be much more interesting.

Geforce RTX – What We Know

The problem is that we don’t actually know much about the next generation of GeForce cards. Nvidia’s consumer line aimed at gamers.

While there hasn’t been a detailed announcement, the company did release this teaser video:

In it, they strongly hint that the flagship card will be called the RTX 2080. Moreover, it’s clear that the consumer implementation of Turing is also built around ray tracing.

However, good old WCCF Tech published a rumour about the RTX 2080 Ti model’s specifications. Obviously, none of this is confirmed, but like all believable rumours, the numbers are not outside of the realm of reality.

This card is rumoured to sport an astounding 4352 CUDA cores, which handle graphics and general purpose GPU operations. It may also have 576 TENSOR cores, which handle AI operations.

The full-fat Turing core that has been revealed can run at up to 1.7 GHz, which means all those specialized cores will also really go like stink. While the non-TI 2080 will be a little cut down from this, all of these high-end Turing cards are likely to decimate the Pascal-based cards they are replacing.

Why It Matters

Ray tracing probably still won’t be a mainstream feature, but just the fact of real-time ray tracing in a consumer GPU is a major step forward. In VR it could mean a significant improvement in how realistic things look. In an AR context, it means rendering objects that really look like they are blended in with what’s really around you.

The overall shift in power per Dollar hopefully means that we’ll start to see VR-ready cards at least as fast as a 1050 Ti at the same price point for something like the MX150. A chip more at home in tablets.

These cards could also mean better support for the next generation of VR headsets that will have as much as 8K resolution displays. We have no idea when the consumer versions of Turing will launch, but clearly now is not a good time to buy a VR laptop or PC.