David Kosslyn and Ian Thompson are the founders of Angle Reality, Inc. Angle is a secret project supported by US$8 million in funds. The project has been running for two years and they will not disclose too much about the constructed virtual world. At least Will not speak openly. But they would say that the virtual world they are creating has, to some extent, changed the relationship between computer hardware and software. When using a PC (personal computer) or a game console to run this virtual world, the GPU chip has played an unexpectedly large role and has taken a lot of burden on the CPU.
GPUs are the acronyms for graphics processing units. Chips were originally designed to better create graphics for games and other software, but they still play an important role in the development of the virtual world created by Angle. In addition, Kosslyn and Thompson transfer countless other tasks to these chips because the GPUs excel at parallel computing. A machine can hold hundreds of GPU chips, and each chip is largely auto-run. Kosslyn and Thompson said that by simultaneously processing hundreds of digital tasks, these chips can bring you closer to the illusion of being in another universe.
When they first started building their own virtual world, Kosslyn and Thompson ran most of the tasks on the CPU (the computer's central processor). But that won't work. As Kosslyn explained, they use the CPU to create new objects as you see in this world - trees, shrubs, rocks, and anything else they need to imitate. It takes about one-fifth of a millisecond to load each object into the chip - it takes too long to think about how many objects they need to load.
Kosslyn said: "If you zoom in on 10 million objects, you will suddenly be squashed by the magnitude. Just loading the world is very slow. We thought, 'Oh my God, if we can do a lot of independent operations It's really fast at the same time." And that's what GPUs are good at."
In hardware and software around the world, their approach is part of a broader shift. For decades, according to the frequently cited Moore's Law, the processing power of personal computer chips has been upgraded once every 18 months or so. But in recent years, this trend has begun to slow down, even if modern software applications require more processing power than ever before. Google chip engineer Norm Jouppi said: "Someone called it the end of Moore's Law, and I prefer to call Moore's Law retreat because it hasn't completely ended, but it's just not working harder." The result is : Companies and coders are now shifting their work tasks from the CPU to a wider range of alternative processors. If a single chip cannot provide them with enough processing power, they need many processors (processed together).
These changes have driven a huge data center that supports Google, Facebook, Microsoft, and Amazon. Because their comprehensive online services can no longer rely solely on the CPU to handle all tasks, these companies are loading the main processing tasks into GPUs, programmable chip FPGAs, and custom chips, such as the artificial intelligence (AI) used by Jouppi in designing Google. processor. Neural networks and other forms of artificial intelligence are often the driving forces behind this transformation.
This change is so great that it has sent ripples to the global chip market. The wealth of the world's largest GPU manufacturers is steadily rising. Intel, which does not have much access to GPUs, has spent billions of dollars acquiring companies that manufacture FPGAs and various artificial intelligence chips.
Today, as Kosslyn and Thompson showed, this trend has moved to the other end of the Internet, pushing PCs and game consoles. Neural networks also help drive this change, but VR may be a bigger driver in the long run. In virtual reality, much processing must be done at the client rather than at the data center. VR must operate in real time, it can't save the millisecond time that sends and calculates through the wire, this will reduce the efficiency. Likewise, it is not possible to save milliseconds on the client's central processor for all calculations. Kosslyn said: "Some tasks are completely difficult for the CPU to handle."
For their virtual world, Kosslyn and Thompson are trying to take advantage of these additional GPU processing capabilities that can be used on high-end PCs and game consoles. Thompson said: "We mainly want to try to fill in the milliseconds that can be done in addition to the GPU drawing graphics." But in the long run, as the virtual world becomes more popular - and even more complex - this may drive Hardware vendors install more GPUs into PCs and consoles. Device manufacturers may also promote GPUs or other replacement chips into VR headsets so that they can operate without PCs and consoles. And this trend may extend to smart phones that are often used instead of dedicated headsets.
Zvi Greenstein, general manager of consumer VR at NVIDIA, said that the company believes there is a huge market for virtual reality. Kosslyn and Thompson reminded that the tools for running code creation in GPUs are still a bit rough on the edge, but companies such as NVIDIA are trying to change the status quo.
Frank Soqui, general manager of Intel’s Virtual Reality and Gaming Group, wonders how many developers actually move workloads to the GPU. But keep in mind that GPUs are primarily in the NVIDIA arena and NVIDIA is one of Intel’s main competitors. Despite this, Soqui does believe that the market will continue to shift to alternative processors. Intel recently acquired a startup called Movidius, which manufactures professional chips that enable devices to sense the world around them—including inventions such as robots and driverless cars. But Soqui said that Movidius' chips can also make VR headsets aware of what's happening in the real world, such as pointing out where you are headed.
At the same time, Microsoft has set up a dedicated processor for its Hololens-enhanced reality headset that, among other things, helps the device track your motion trajectory. This is another example of a computing task moving from the CPU to something else.
The result? Developers like Kosslyn and Thompson must think differently about how to develop their own software. But this is only one of the consequences of this change. It also means another shift in the global chip market, further to the GPU and other processors. And, most notably, it means that virtual reality will evolve so much as the coder stores most of his tasks on the CPU, and VR will become faster and better.
Transparent Oled,Transparent Oled Display,Transparent Oled Display Module,Transparent Oled Lcd Display
Shenzhen Newvision Technology Co.Ltd , https://www.newvisionlcd.com