In a recent post, JC Fletcher talks about the service promoted by onlive.com gaming platform. In this post he talks about some subjects on which the Cloud Gaming could have problems, namely the video encoding process and the latency on communications. In previous posts I already have talked about some of these questions and also about the projects developed by AMD on this field, but the information provided by the JC’s post will help introduce some new thoughts on this matter.

In my perspective the latency problem is the most difficult to control by Cloud Gaming companies. They will always depend on the end-user’s Internet connection, thereby they must provide technologies that dynamically adjust the gaming conditions to the Internet topology. They will also have a very restrict time frame to accomplish their work. JC has reported that 150ms time frame would be the limit to have quality gaming. Each Cloud Gaming company must include in this 150ms time frame all the time used with the gaming processing (3D render and encoding), the time need to transmit to the client, and also some time for buffering. Buffering is the simplest way to reduce the impact of the jitter (the latency variation).  As Jules Urbach says, and as I’ve talked about in a previous post, we must realise that the latency problem analysis should be divided in two flows: the upstream (user inputs), and the downstream (gaming visual output). As he also said, we must use, as example to follow, the technical solutions found on nowadays online gaming to solve problems on the upstream users’ input.  It’s difficult to say that 150ms time frame would be enough, but using the previous examples, we could say that the energy of the Cloud Gaming companies must be concentrated on the second flow – the downstream or the gaming output.

JC Fletcher advocated in this post that the gaming processing and the gaming output encoding are things difficult to  achieved, and therefore that we wouldn’t hav solution to the second part of the problem.

“The bottom line here is that OnLive’s ‘interactive video compression algorithm’ must be so utterly amazing, and orders of magnitude better than anything ever made…”

I agree with this statement in one aspect: this kind of encoding  would be definitely something that hasn’t been done until now. But I believe, as opposed to JC Fletcher belief, that this is something that nowadays CPU/GPU’s power will be able to do. We must realise that this kind of encoding is very different from the regular live video encoding. In a traditional live video encoding process we always have two separate parts:(1) the acquisition process, usualy done by video cameras; (2) the encoding process.  In 3D gaming, these two parts could (should) be integrated, creating a new, innovative and high performance encoding/render mechanism, able to reduce the time need for the whole process, and also increasing the quality of the encoding result.

Another thing that’s important to notice is that in this kind of service, as we already have in traditional console gaming, the hardware architecture isn’t necessary the same as the one currently present in our PC (CPU+GPU). The Cloud Gaming provider should deploy a gaming platform that would take in all the power of existing GPUs. This is exaclty what AMD is proposing with the “Render fusion”. At the same time they must provide a developing platform that would “constrain” the gaming developer. This model is the one already used by traditional gaming consoles like XBOX or Playstation. This is the only way that gaming platforms have to obtain the maximum return from its platforms. Create a new gaming platform for developers could be a tricky question, but it doesn’t have to be completely new, a Cloud Gaming could easily use existing platforms and add to it small amount of changes…