Mobile VR is a chip factory game?

Some media had previously stated that “the virtual/augmented reality is behind the flames, and it is more like a chip manufacturer war.” The influence of the chip factory is so great that it has reached the point where the industry is concerned.

Which vendors are doing VR-optimized chips?

In foreign countries, in addition to Qualcomm, Samsung Exynos drives Gear VR and Big Brother VR all-in-one; HoloLens, Storm Devil and 3Glasses Blubur W1 use Intel chips; ARM, which is responsible for designing chip architecture, is also optimizing for VR. The domestic Ruixin Micro launched the RK3399 chip; the Quanzhi will soon release a VR hardware solution based on its own chip on the 21st.

A Jun, head of hardware research and development of a VR all-in-one startup company, said that although there is this arrangement, Samsung and Qualcomm are actually moving. He said:

At present, the Android platform Qualcomm and Samsung Exynos are the most suitable, Samsung's power control is better, Qualcomm is relatively strong in the GPU. MTK is in a behind-the-slate state; Intel is in the layout, but it is too late to move this block, and it is hard to say now; Huawei's GPU is still not very good, far below the level of the "pass line" of 7420 (unicorn 955 package is T880MP4, The 7420 is MP8. From this aspect, it can be understood that Huawei has not begun to pay attention to VR.

What is the so-called "optimizing for VR" in the chip factory?

At present, Qualcomm, which has the largest motion, provides a low-level SDK to facilitate developers to invoke sensor data. It also optimizes the GPU rendering aspect and power consumption.

As for the results of Qualcomm's optimization, VR live service provider NextVR once developed a live broadcast app using Qualcomm-optimized 820. Founder Dave Cole said in an interview with RoadtoVR, “This new technology has raised the threshold for mobile VR, and Make the 820-equipped mobile phone ahead of the best mobile VR on the market."

"...820 has many advantages for an external IMU (inertial measurement unit) introduced by older SoCs. Qualcomm's Compute DSP (digital signal processing) and more targeted HVX acceleration enable high-speed IMU sampling, low latency, and display driver optimization. "Cole continued.

However, in the eyes of A, the optimization effect of Qualcomm remains to be seen, after all, currently equipped with 820 VR one machine has just been listed.

A Jun studied the ARM SDK, said the SDK is still relatively basic, basically all things CardBoard. "(ARM) is mainly to do their own thing, improve performance and reduce power consumption, optimization is mainly Google and the work of the plant. ARM SDK to do that, Google will certainly do." He said.

Most of the current vendors are only optimizing VR based on SoCs for smart phones and tablets. It is not impossible for the VR market to grow in the future. It is not impossible to launch a dedicated chip, and even an independent GPU will appear. Google’s Project Tango has a design dedicated to 3D. The processed chip.

VR hardware factory also has to optimize

After the VR integrated hardware solutions based on various chips are released, we will see a big wave of integrated machines in the market. In fact, this wave of products has already begun. But you will find that there are few really good products.

The reason is that the optimization of the chip factory cannot solve all the problems. Just like highly-standardized smart phones, the real difference lies in what hardware vendors can do on their own.

A Jun said that simply relying on Qualcomm to solve all the problems is unrealistic, such as VR's special need for rendering speed, if you die of hardware performance, it will be very difficult.

There is also power consumption, and it has a great relationship with software optimization. It will be more and more difficult to reduce power consumption by using process technology. It is also a matter for Qualcomm and Intel. Therefore, Chinese companies still have to optimize. Content determines hardware and failings?

Mobile VR Game Lamper VR

Many manufacturers that Lei Feng contacted have stated that mobile VR is suitable for more fragmented and lighter applications. In their view, this makes mobile VR less demanding on the performance of the chip, coupled with the appropriate optimization, it is estimated that the product is almost the same. Many vendors also pursue a strategy of occupying pits. Even though they can only do so, mobile processors are progressing so quickly that they will become better and better in a few years, just as the first generation of iPhones released. WANG Longfei, Shenzhen Virtual Reality Virtual Reality Solution for Shenzhen VR One Machine, told Lei Fengwang that “content determines the hardware and fails”. This means that if you use a mobile VR to experience a big game, you naturally fail, but if you just play some light applications, then it is a pass.

This point of view has its own rationale, but A Jun thinks, "The primary VR has the performance requirements, which will not be short." Then returned to the battle of mobile VR and grid line, according to Noriti CTO Dai Ruoli's point of view, mobile VR pass line should be Gear VR. In fact, A Jun also believes that the VR pass line on the chip is the Samsung Exynos 7420. This means that most of the domestic products have not yet reached this pass line.

For many vendors' pit strategy, A Jun thinks this is too naive. "The hardware is fair to everyone, but it's not that we make chips anyway. A program, you can use others to use it, the key is to optimize the ability , combination of hardware and software capabilities."

Finding a solution from Shenzhen is not a technology. Shenzhen's planners started as low-end machines, and high-end machines had no experience at all, so they couldn't pass VR products. Most of the big manufacturers are low-end machines outsourcing, high-end are their own. This is one reason why I am so pessimistic about the hardware factory.

to sum up

Although chip makers are trying to deploy VR, the real leaders in the industry are the chip giants Qualcomm and Samsung on smartphones and the top-level ARM. Domestic chip makers either have no action or are still taking the low-end route like the smartphone era.

For domestic VR helmet hardware startup companies, simply relying on a chip solution obviously cannot win the competition in the industry. According to the Lei Feng network reporter, the group A who has done a VR all-in-one has revealed the truth about the industry—it is not a core competency to use a big chip company. It is not realistic to expect the chip factory to solve all problems. Must have its own optimization capabilities and software and hardware integration capabilities.