How to save a smartphone charge?

Original author: Miran Lee, Yunxin Liu
  • Transfer
Since February 23, gentlemen! Yes, the day of defender of the fatherland has long ceased to be only for defenders, but they have a separate salute!

On the occasion of the holiday, we have prepared an article about what most guys love - about games. More precisely, about our new development, codenamed RAVEN, which will help maintain the treasured interest that disappears so quickly when playing on the phone. It renders only those game frames that are noticeably different from those nearby. Look under the cat!

Over the past ten years, mobile games have become a huge industry in the entire gaming industry. According to Newzoo , the global mobile gaming market reached $ 46.1 billion in 2017, up 19.4% from a year earlier.

Gamers can enjoy stunning mobile games due to the ever-expanding capabilities of modern mobile GPUs. However, everything has disadvantages, and in this case, a lot of energy consumption. The power consumption of mobile GPUs is increasing with the increase in the number of graphics computing. As a result, high-quality mobile games with advanced graphics are demanding and drain the battery very quickly.

To solve the above problem, researchers from Microsoft Research Asia (MSRA) and Korea Advanced Institute of Science & Technology (KAIST) have developed a new system called RAVEN, which can reduce the power consumption of mobile games without harming the user.

RAVEN is based on FPS in mobile games: many frames continuously displayed in the game are either perceived the same or very similar. The differences in these frames are too small to be noticeable to players. However, mobile games always render frames at 60 FPS, no matter how similar they are. Based on a measurement study by scientists, these excess frames can account for more than 50% of the total number of frames in many games. Obviously, eliminating the rendering of these redundant frames can significantly reduce power consumption.

RAVEN is a new system that relies on the visual perception of a person to scale the speed of rendering frames. To achieve this, RAVEN uses perception-aware scaling (PAS) technology. This technique reduces the rendering speed of frames in games when subsequent frames are predicted to be similar to previous ones. At the same time, the similarity should be at that level so that the user does not notice a “subsidence”.

RAVEN uses a side channel to track the displayed frame sequences in order to adapt the user's perception of graphics changes during the game. In this way, RAVEN adapting reduces GPU power consumption.

The RAVEN system consists of three main components that collectively scale the rendering speed of game frames: Track Difference Tracker (F-Tracker), Rate Regulator (R-Regulator) and Rate Injector (R-Injector). All these components work in the form of a pipeline, in order. First, the F-Tracker measures graphic similarity between two frames. Then the R-Regulator predicts the level of similarity between the current and the next frame (frames). Prediction is based on how similar the current frame and the previous frame (frames) are. If subsequent frames (predicted) are quite similar to the current one, R-Injector limits the frame rendering speed by introducing some delay in the rendering cycle and skipping graphics processing for an unnecessary frame (frames). Currently, RAVEN can skip up to a maximum of three frames and thus

The main problem of RAVEN is that it is not clear how to determine the similarity of frames at low cost of computing power. The direct method for comparing similarity is the level of structural similarity of frames (SSIM). Determining SSIM is a complex calculation and therefore uses a lot of power, especially for wide frames. Today’s mobile devices, including smartphones, usually have a high screen resolution of 1920 × 1080 pixels or higher, which makes calculating each SSIM level using RAVEN a pointless exercise.

To solve this problem, our colleagues used two innovative methods

First , they developed an energy-saving method for measuring graphic similarity, based on the sensitivity of the human eye to color differences. This method uses the difference in brightness (that is, the Y component in the YUV color space ) between frames. Researchers positively evaluated this method by comparing it with SSIM under various conditions, and the results showed that it effectively measures graphic similarity at low processing power costs.

Secondly, the researchers built a virtual display cloned from the main display of a mobile device, but with a much lower resolution (for example, 80 × 45). The system reads the graphic contents of the virtual display to measure the level of similarity. Since the resolution of the virtual display is much lower, the computational requirements are also much lower.

Thus, the two methods described above effectively reduce RAVEN power consumption.

As a next step, the researchers introduced the RAVEN system into the Nexus 5x smartphone. In a study of 11 users, colleagues conducted comprehensive experiments using various gaming applications to evaluate the effectiveness of RAVEN. On average, power consumption per game session decreased by 21.8%, and at the peak by 34.7%. At the same time, the quality and sensations of the game remained at the same levels as before the use of RAVEN.

Demonstration at MobiCom 2017:

RAVEN is the first system for mobile games designed to scale frame rates and save energy, as well as based on graphical similarity of frames. Documentation describing the RAVEN system “RAVEN: Optimizing Energy Consumption for Mobile Games” was published and shown at MobiCom 2017. The technology authors are Chanyou Hwang, Saumay Pushp, Changyoung Koh, Jungpil Yoon, Seungpyo Choi and Junehwa Song from KAIST and Yunxin Liu from MSRA.

Also popular now: