IBM Streams HD Video at 3Mbps

http://www.pcworld.com/article/id,144581-c,video/article.html#


A new encoding system that offers the ability to stream HD video at 3Mbps is being launched by IBM.

The system, which was developed by IBM and Broadcast International will be on display at the National Association of Broadcasters conference, which takes place in Las Vegas this week.

more info on the processor
http://domino.research.ibm.com/comm/research_projects.nsf/pages/multicore.CellBE.html

Both Sony and Toshiba had a hand in this.
http://domino.research.ibm.com/comm/research.nsf/pages/r.arch.innovation.html

The Cell architecture grew from a challenge posed by Sony and Toshiba to provide power-efficient and cost-effective high-performance processing for a wide range of applications, including the most demanding consumer appliance: game consoles. Cell - also known as the Cell Broadband Engine Architecture (CBEA) - is an innovative solution whose design was based on the analysis of a broad range of workloads in areas such as cryptography, graphics transform and lighting, physics, fast-Fourier transforms (FFT), matrix operations and scientific workloads. As an example of innovation that ensures the clients' success, a team from IBM Research joined colleagues from IBM Systems Technology Group, Sony and Toshiba, to lead the development of a novel architecture that represents a breakthrough in performance for consumer applications.

One of the guys out at NAB just checked out for that booth for me. His off-hand comments tonight:

"The quality was good, not great, but it was decent. It sure doesn't meet the expectations of "HD". It was impressive though."

- Tom

Is there any point at which processing power can play a larger role than the codec itself? Or is the codec always the limiting factor?

I guess a better way of asking the question would be, if there was virtually no CPU limitation, what would the lowest bit rate be of an optimized codec to still produce HD content? Can processing power ever overcome today's internet bandwidth limitations?

Scott

is this part of/related to the 'spursengine cell processor' concept that is supposedly/allegedly going to be used in future toshiba dvd players? sounds like it, a little.

TRBerry, I am glad your friend reported back on what he saw at NAB, this is good news. It was running on a small IBM Bladecenter machine( a person can buy one for home). I would not think IBM would bring in a Z10 mainframe to show it's true power or a IBM P server. A Cable company or (NETFLIX,BLOCKBUSTER would use the Pserver or mainframe) at a couple of datacenter, FYI I am a Mainframer at a large Fortune 500Electric Utility Company)

IBM Z10 mainframe
http://www-03.ibm.com/systems/z/

IBM PServer
http://www-03.ibm.com/systems/p/hardware/midrange_highend/index.html

IBM Bladecenter
http://www-03.ibm.com/systems/bladecenter/hardware/servers/power.html

Is there any point at which processing power can play a larger role than the codec itself? Or is the codec always the limiting factor?
Improving processor power gets us closer to the max of what the codec can do, but it's not even close to linear.

For example, with our VC-1 implementation, going from the highest-speed mode to the highest-quality mode, might slow the codec down by 8x, while letting you drop bitrate by 20%.

I guess a better way of asking the question would be, if there was virtually no CPU limitation, what would the lowest bit rate be of an optimized codec to still produce HD content? Can processing power ever overcome today's internet bandwidth limitations?
It certainly makes it better, although computational requirements go up with frame size, and frame size is mainly limited by bandwidth, so with lower bandwidths we get more MIPS/pixel.

There were tons of demos of low-bandwidth live streaming at NAB this week. I was particularly impressed with the Inlet Spinnaker HD with VC-1. They even did a great looking 960x540 @ 1.5 Mbps, which wasn't enough bandwidth for SD only a few years ago. And there are still improvements for VC-1 for live that we're working on.

But we're certainly never going to get 1920x1080 @ 2 Mbps with existing codecs. Maybe H.265, but mature implementations of that are at least a half-decade away.

But we're certainly never going to get 1920x1080 @ 2 Mbps with existing codecs. Maybe H.265, but mature implementations of that are at least a half-decade away.

I'm looking forward to h.265. Will it be a quantum leap like MPEG-4 was over MPEG-2?

I'm looking forward to h.265. Will it be a quantum leap like MPEG-4 was over MPEG-2?
Hard to say until it's done and we've had a couple of years for the implementations to mature.

Remeber that the original MPEG-4 part 2 codec was actually a huge dissapointment, and really wasn't enough better than MPEG-2 in efficiecy to make it worth the switching costs. Hence the scramble to make H.264. Which itself was pretty disappointing at higher resolutions until the last-minute addition of the High Profile.

Improving processor power gets us closer to the max of what the codec can do, but it's not even close to linear.

For example, with our VC-1 implementation, going from the highest-speed mode to the highest-quality mode, might slow the codec down by 8x, while letting you drop bitrate by 20%.


It certainly makes it better, although computational requirements go up with frame size, and frame size is mainly limited by bandwidth, so with lower bandwidths we get more MIPS/pixel.

There were tons of demos of low-bandwidth live streaming at NAB this week. I was particularly impressed with the Inlet Spinnaker HD with VC-1. They even did a great looking 960x540 @ 1.5 Mbps, which wasn't enough bandwidth for SD only a few years ago. And there are still improvements for VC-1 for live that we're working on.

But we're certainly never going to get 1920x1080 @ 2 Mbps with existing codecs. Maybe H.265, but mature implementations of that are at least a half-decade away.Thanks for the response, Ben. I kind of thought that's how it would work, but just didn't know.

It's fascinating to see just how much can be extrapolated from smaller and smaller data streams. I'm glad to see there's still room for improvement.

Scott

Thanks for the response, Ben. I kind of thought that's how it would work, but just didn't know.

It's fascinating to see just how much can be extrapolated from smaller and smaller data streams. I'm glad to see there's still room for improvement.

Yep. By comparison, MPEG-2 is arguably about 3x more efficient than when the first implementations were made. All codecs that are in wide use show progressive improvements over the years. Certainly what we can do with VC-1 Main Profile today blows away the WMV 9 implementations first demoed back in 2002, even though our latest and greatest bitstream today works perfectly in those old decoders.