Duke Researchers Enable Graphics-Rich Cloud Gaming Without Chewing Up Mobile Bandwidth
Cloud gaming is a technology that's already here, but with high bandwidth requirements and data caps still a thing that mobile users have to contend with, there's definitely room for improvement. The question is, how? Researchers at Duke University and Microsoft Research have been working together to answer that very question, and what they've come up with is a promising new tool that could lead to console-like graphics on mobile devices with comparatively little bandwidth requirements.
The tool is called "Kahawai," which is the Hawaiian word for stream. What Kahawai does is task the mobile device -- smartphone or tablet -- with generating a rough sketch of each frame in a game, or a few high-detail sketches of certain frames. This takes some of the load off the remote server, which can then focus on filling in the missing pieces. By taking advantage of a mobile device's GPU to help with the load locally, far less bandwidth is needed.
This differs from conventional cloud gaming where the remote server does all the heavy lifting. Transmitting high-resolution graphics, processing audio, and interpreting user commands all in real-time on the server side means you'll need a fast Internet connection with plenty of bandwidth. It would also require a big data cap -- according to the researchers, games like Halo or Tomb Raider on a smartphone would chew up a 2GB monthly plan in just a few hours.
Kahawai solves that problem through collaborative rendering. To demonstrate the technique, the researchers integrated Kahawai into Doom 3 and uploaded a video to YouTube showing the experience. Running at 1Mbps, the version of Doom 3 using Kahawai was noticeably better looking than the version that relied on conventional cloud gaming methods. Have a look:
"That’s a huge win, especially if your cellphone plan has a data cap," said Duke computer scientist Landon Cox, who helped develop the approach with his graduate student Eduardo Cuervo, now at Microsoft, and Alec Wolman, a Microsoft researcher.
There don't seem to be any side effects to the gaming experience. To prove it, the researchers conducted performance trials with "hardcore gamers" -- gamers who spend an average of two hours per day playing games -- and recorded similar scores with both techniques. There was no difference in response times, either.
"You essentially get the same gaming experience, but you save a lot of data," Cox said.
The tool is called "Kahawai," which is the Hawaiian word for stream. What Kahawai does is task the mobile device -- smartphone or tablet -- with generating a rough sketch of each frame in a game, or a few high-detail sketches of certain frames. This takes some of the load off the remote server, which can then focus on filling in the missing pieces. By taking advantage of a mobile device's GPU to help with the load locally, far less bandwidth is needed.
This differs from conventional cloud gaming where the remote server does all the heavy lifting. Transmitting high-resolution graphics, processing audio, and interpreting user commands all in real-time on the server side means you'll need a fast Internet connection with plenty of bandwidth. It would also require a big data cap -- according to the researchers, games like Halo or Tomb Raider on a smartphone would chew up a 2GB monthly plan in just a few hours.
Kahawai solves that problem through collaborative rendering. To demonstrate the technique, the researchers integrated Kahawai into Doom 3 and uploaded a video to YouTube showing the experience. Running at 1Mbps, the version of Doom 3 using Kahawai was noticeably better looking than the version that relied on conventional cloud gaming methods. Have a look:
"That’s a huge win, especially if your cellphone plan has a data cap," said Duke computer scientist Landon Cox, who helped develop the approach with his graduate student Eduardo Cuervo, now at Microsoft, and Alec Wolman, a Microsoft researcher.
There don't seem to be any side effects to the gaming experience. To prove it, the researchers conducted performance trials with "hardcore gamers" -- gamers who spend an average of two hours per day playing games -- and recorded similar scores with both techniques. There was no difference in response times, either.
"You essentially get the same gaming experience, but you save a lot of data," Cox said.
