Wayne Morellini
Well-known member
- Joined
- Jan 3, 2008
- Messages
- 6,157
- Reaction score
- 0
- Points
- 0
Yeah Blair, there are multiple reasons. Apart from competency issues, and just throwing rubbish solutions out to collect money (look at how the ambarella h264 solution blitz the industry by putting good competent work in, even the Japanese were caught with their pants down). Next time I pass by one of the Ambarella guys in the street, again, I'll have to ask them what is happening. If you really wanted to do RR low powdered, Ambarella is probably a great American company to contract to design it.
The thng about h264, it was that the amount of smart work you put in to making the compressed stream significantly affected the performance. Remember how around ten years ago we were getting rubbish compression, and in response I published a smart way to sacrifice image quality to sustain more quality per gigabit. That was the simple method, and image quality took off. There is more that can be done, but most don't have the competence, as is life. However, all this complexity requires more circuitry and power consumption, but if you use a GPU or other programmable solution you shift some of the circuitry to software and don't have to spend money designing it that way. I've actually come up with a solution before that should beat custom circuitry for power consumption. I've been in a scene where they developed cutting edge FPGA beating processing solutions. I am actually looking forwards to working with them on these solutions. One test project I want to do is my own codec that should beat RR in quality per unit of data.
Anyway, their is timing of roll out of the h265 standard phases, low end slow burn devcelopement cycle investments to bring out solutions in the future in a more timely manner (like when existing h264 platforms run out of steam) and the power consumption and cost of a h265 circuit is likely a big point holding it back, but I don't know yet. When the die process shrinks enough to be cheap and lower power, and the other factors arer rght they will probably come. At this stage, only cameras and other present uses, might be suitable for h265 circuites, and as usual for this this stage, solutions based on programmable solutions might be the norm for now.
I would not be too surprised if RR where based on h264\h265 with a lot of smart extras. This would make it a lot weaker, because h264's base line efficiency is less than something better, requiring a lot more smart layers to get its performance to this level.
The thng about h264, it was that the amount of smart work you put in to making the compressed stream significantly affected the performance. Remember how around ten years ago we were getting rubbish compression, and in response I published a smart way to sacrifice image quality to sustain more quality per gigabit. That was the simple method, and image quality took off. There is more that can be done, but most don't have the competence, as is life. However, all this complexity requires more circuitry and power consumption, but if you use a GPU or other programmable solution you shift some of the circuitry to software and don't have to spend money designing it that way. I've actually come up with a solution before that should beat custom circuitry for power consumption. I've been in a scene where they developed cutting edge FPGA beating processing solutions. I am actually looking forwards to working with them on these solutions. One test project I want to do is my own codec that should beat RR in quality per unit of data.
Anyway, their is timing of roll out of the h265 standard phases, low end slow burn devcelopement cycle investments to bring out solutions in the future in a more timely manner (like when existing h264 platforms run out of steam) and the power consumption and cost of a h265 circuit is likely a big point holding it back, but I don't know yet. When the die process shrinks enough to be cheap and lower power, and the other factors arer rght they will probably come. At this stage, only cameras and other present uses, might be suitable for h265 circuites, and as usual for this this stage, solutions based on programmable solutions might be the norm for now.
I would not be too surprised if RR where based on h264\h265 with a lot of smart extras. This would make it a lot weaker, because h264's base line efficiency is less than something better, requiring a lot more smart layers to get its performance to this level.