SteamVR Motion Smoothing

SteamVR :: Introducing SteamVR Motion Smoothing

Although this is a relatively modest addition to VR applications, its merely one example of how they can cut the frame rate demands in half, which has enormous implications when you talking about a VR headset that requires 90fps in each eye. Valve did this in cooperation with AMD, who have also been working with Bethesda and Liquid Sky Cloud to reduce the hardware demands for 4k, VR, and even ray tracing down to a minimum. AMD is shooting for a next generation SoC, or system on a chip, that can do all of these justice using 30 watts and open source hardware and software including Vulcan and Linux, while Valve has been working towards the same goal for years now.

If you pay attention to what AMD and the developers are doing and contrast it to what Nvidia is doing, it all starts to make more sense from a hardware perspective. For example, Nvidia's new patent for "Infinite Textures", their DLSS "upscaling" for 1440p, and their ray tracing and path tracing push are all the future of graphics, but nothing about any of them is ground breaking except the fact you can now afford to buy them in a commercial product that is way overpriced. In the last few years Intel, Nvidia, and Microsoft could not have tried harder to piss more people off in the industry, by attempting to dominate the market and this is one response from Valve, who introduced Steam specifically to kick the big boys in the nuts.

So, what Steam and AMD and everyone seem to have in mind, is to focus all their efforts on dealing with bandwidth issues that are holding back lower end hardware. Today's high end laptop is tomorrow's bargain, and by merely stacking a bunch of cheaper chips together on a silicon interposer, adding a little AI circuitry, and reducing the system requirements at the same time they are going bananas. It used to be having the fastest cpu was important, but now its also the core count and AMD's new server is a cheap petaflop computer, while Nvidia's tensor cores in their RTX 2080 ti produce the equivalent of 110 teraflops, and their rasterizing engine does over 14 teraflops, making their combination equal to what previously cost about $60,000.oo.

Tensor cores, or fpga circuitry, are less accurate analog circuitry, while their cuda cores are more accurate, and the combination is fantastic for gaming and VR applications. Samsung has just begun manufacturing their new 7nm chips, which will eventually include specially modified Ryzen chips for the next PlayStation. Somewhere around 2-4nm is about as small as silicon is expected to get without adding something on top like carbon nanotubes or whatever, so it looks like we are lucky and a single cheap to manufacture chip will eventually have all the video graphics and AI that anybody in their right mind might need.

The next generation consoles should both give us an idea of what that single chip will look like, before it comes on the market, but you can pretty much say goodbye to desktop gaming ever being the same again. You'll be able to eventually cheaply upgrade your monitor and computer for almost any practical application, including running a home business, by merely plugging a new chip in the back. By reducing everything down to a single chip, they can dramatically reduce the costs of everything else including the power supply, with all of Intel's processors now having their own voltage regulators on the chip.