Jump to content
kaysadeya

New chip decodes 48 video streams

Recommended Posts

Forwarded from a friend....

 

From Tuesday's "Good Morning Silicon Valley" newsletter:

 

Analysts revise rating of Cell chip from "hypeware" to "Hey, would you look at that!" No microprocessor since Transmeta's Crusoe has inspired as much hype and bloviating as the IBM/Sony/Toshiba Cell. Though it's been in development for some four years now, very little is known about the Cell, aside from the grand pronouncements of its designers: 10X performance for many applications; supercomputer on a chip. These promises have been made for months with no real supporting evidence. Now, it finally appears we have some. At the COOL Chips VIII conference in Yokohama in late April, Toshiba showed a video of a Cell processor simultaneously decoding 48 video streams. "In the film, 48 MPEG-2 streams stored on a HDD were read, decoded and projected onto a 1,920 x 1,080 resolution display divided into 8 x 6 cells, each of which showed a different video in each cell," Nikkei reports. "The company expects the technology to be used to display thumbnails for a video list. Of the eight synergistic processor elements (SPE) used in the Cell, six are used for decoding 48 MPEG-2 streams and one is used for scaling the screen. The remaining SPE can be used for a completely different processing function." I've no idea how large or small the video streams were; a poster at Beyond3D claims they were 720x480 MPEG-2 streams with 6 SPEs. But whatever their size, 48 streams is a lot of video, and the fact that it was all seemingly decoded with power to spare is pretty astonishing.

Share this post


Link to post
Share on other sites

If Cell works out it will be insanely powerful. But note that in that example it was simply displaying already encoded video.

Share this post


Link to post
Share on other sites

It probley will. From what I've been reading Cell looks like it's going to be a beowulf cluster kind of tech. Great for processes that writen in parallel, less useful for some other things. But I don't expect to see it in PC's anytime within the next few years.

Share this post


Link to post
Share on other sites

Maybe what the CCTV industry needs is a standardized (and I have to emphasize STANDARDIZED) operating system designed from the ground up for DVRs. It’s ironic that we depend on operating systems like Linux (let’s just say Unix) and Windows, which were designed long before anyone even heard of a DVR.

Share this post


Link to post
Share on other sites

yep, GE (Kalatel) has been using it from day 1.

Except their DIGI-4 is now a linux DVR (i think), network software is not that great though, plus it only comes in a 4 channel.

Share this post


Link to post
Share on other sites

Did you say STANDARDIZED or BASTARDIZED?

Linux is solid as a rock (though, not yet ready for the desktop yet-reguardless of what anyone tells you). The best if you are running a server or a DVR.

Share this post


Link to post
Share on other sites
Did you say STANDARDIZED or BASTARDIZED?

Linux is solid as a rock (though, not yet ready for the desktop yet-reguardless of what anyone tells you). The best if you are running a server or a DVR.

 

Yes, Linux is better than Windows for a general-purpose operating system. But the question is what is it built for? Can a single OS leverage all of the new hardware and host more “esotericâ€

Share this post


Link to post
Share on other sites

The problem with a stanardised OS is it will never be perfect. It will always be missing something. And developing an OS is expensive. Who will pay for it? Will it be OSS? Will it be licenced in a friendly manner?

Share this post


Link to post
Share on other sites

For the standalones, anyone in the embeded market would be a player. But here is a monkey wrench for your works. What if I want to put it in a windows domain? With Linux I can use Samba, Windows has native support for it, but other RTOS's don't. You're perfect OS is going get bloated quickly if you start supporting everything.

Share this post


Link to post
Share on other sites

I’m referring to the dedicated market, not add-on DVR cards. The question how can all of the DVR manufactures leverage new hardware, such as the ‘Cell chip,’ if they each have to individually upgrade their proprietary OSs, or wait for upgrades from MicroSoft or the Linux community?

 

With a standard OS, DVR manufactures could save money and focus their efforts on efficient manufacturing and distribution and adding value elsewhere. It’s really just the same model currently used in the PC market.

Share this post


Link to post
Share on other sites

Again, in the embeded market, this is alot more important. And how does GE/Bosch/Pelco compete if the guy from China is saying it's the same OS? That's the point where people's eye's glaze over.

Share this post


Link to post
Share on other sites
Again, in the embeded market, this is alot more important. And how does GE/Bosch/Pelco compete if the guy from China is saying it's the same OS? That's the point where people's eye's glaze over.

 

Features, usability, stability, security, and quality. I doubt many people buy a DVR based on its OS. It’s the quality and usability of the hardware and the software on top of the OS that matters to the customer.

 

The guy in China has just as much access to Linux, Windows, or any other OS as GE/Bosch/Pelco, so I don’t see how leveraging a general-purpose OS gives them any advantage. Based on what Rory says about the quality of GE/Kalatel DVRs, it may be that GE is a few steps ahead of the rest of the industry by leveraging an OS that is more suitable for deployment in a DVR.

Share this post


Link to post
Share on other sites

You'd be amazed how effect that sales techinque is. If someone thinks they are gettin X at X-100 cost....well who reads spec sheets? But why choose a standardized OS and reinvent the wheel when you have a large number of options already? Linux comes in embeded and RTOS embeded. Then there are the 47+ other embeded OS's out there. Linux will be for sure ported over. I'm sure at least some of them will be ported to Cell. But what does an industry standard OS offer to the companies, where is the money in it?

Share this post


Link to post
Share on other sites
But what does an industry standard OS offer to the companies, where is the money in it?

 

No doubt that’s why there isn’t a standard OS for DVRs. Whether it’s worth someone’s effort to create one is beyond me. I can certainly see why it would be of benefit to DVR manufactures and customers. I’m just saying “wouldn’t it be greatâ€

Share this post


Link to post
Share on other sites

What benfits? They are selecting off the shelf OS's to start with. They aren't doing the R&D. I'm curious to know what the benfits are to the end user?

Share this post


Link to post
Share on other sites
What benfits? They are selecting off the shelf OS's to start with. They aren't doing the R&D. I'm curious to know what the benfits are to the end user?

 

Let’s back up and look at the big picture….

 

The first question to ask is what are you gaining from the use of a general-purpose OS?

 

• Maturity, yes

• Lots of functionality, yes

• A degree of standardization, yes

• Low cost, yes

 

Then what are the tradeoffs?

 

• Large footprint and complexity -- many of the OS features have no use or are not optimized for a relatively simple, video-oriented, real-time environment like a DVR. Sure you can get everything to work just fine, but the question is, would everything work better, faster, and more reliably with an OS designed for a DVR environment?

 

• Security – A general-purpose OS is a target for hackers and leaving it hanging on a network is a potential liability.

 

• Upgrades -- What drives future upgrades? Is it the latest chip that can efficiently decode video, or to play host to the latest spreadsheet or word-processing applications?

Share this post


Link to post
Share on other sites

Security through obscurity is a poor arguement. Apache has proven that being number one does not make you the biggest target. The footprint arguement applies to windows but linux can be run at under a 1 MB. And all of your arguements apply to the current crop of embeded systems.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×