PDA

View Full Version : News from the Cell


Coliba
05-03-2005, 02:06 AM
________________________________________


Toshiba Demonstrates Cell Microprocessor Simultaneously Decoding 48 MPEG-2 Streams


Toshiba demonstrated that its Cell microprocessor, jointly developed with the Sony Group and IBM Corp., can simultaneously decode 48 SDTV format MPEG-2 streams. At the COOL Chips VIII event held in Yokohama from April 20 to 22, 2005, the company showed a film demonstrating the decoding process.

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. 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.

In the demonstration, Toshiba used an operating system environment it had developed to increase the efficiency of Cell software development. One of the environment's key features is that application software developers can program software without considering which threads will be allotted to each of the different SPEs, because the environment allows the automatically scheduling software to SPEs.

_____________________________________

http://techon.nikkeibp.co.jp/english/NEWS_EN/20050425/104149/?ST=english

There's also an interesting article with Ken Kutaragi from Sony (don't know what he does there, the article doesn't say):
http://techon.nikkeibp.co.jp/english/NEWS_EN/20050407/103542/

"The network of today still remains a means to access information. The time will come, however, when the network itself becomes a computer. So far, we have pursued how to improve the performance of a single chip, but we are now becoming aware of the limits that loom there. Once you suppose that the network is a computer, what you have to do is to improve the computing performance of the network as a whole. This is s paradigm shift. Be it on-chip, on-board, or on-network, chips will be infinitely connected -- and that is what this "Cell" is all about."

"I intend to stick to 90nm technology. I know some people who suggested going into mass production using such outrageous technologies as 65nm or 45nm, but you can't make something only by dreaming. In my management role, I'm determined to start mass production with the 90nm. "

"I'm not going to give a green light unless we attain a certain yield rate level. Just think of the amount of chips we are to produce - 20 million for game consoles alone, an estimated 10% of TV sets, as well as home servers."

Beamtracer
05-03-2005, 03:44 AM
In the demonstration, Toshiba used an operating system environment it had developed to increase the efficiency of Cell software development.
It's interesting that Toshiba didn't elaborate very much about what operating system they used. Maybe it was just some simple mock-up OS that can't really do anything else but decode MPEG-2 videos.

It'll be interesting to see if Cell makes it to the desktop computers we all use.

IBM is a big promoter of Linux. I wonder how IBM will modify Linux to take advantage of Cell. I assume all the applications would have to be rewritten or modified.

In the story above, Sony is quoted as saying that "The network is the computer". I thought that was Sun Microsystem's line, from about 10 years ago!

subcast
05-03-2005, 06:43 AM
I think the cell stuff is fascinating. All lame marketing stuff aside this tech is going to rock with a **** on its ****. :bowdown:


*edit - When i came to this morning I censored myself.

Para
05-03-2005, 07:40 AM
IBM is a big promoter of Linux. I wonder how IBM will modify Linux to take advantage of Cell. I assume all the applications would have to be rewritten or modified.

While Linux will very likely to be ported to Cell by someone (after all Linux is also available even for iPod! (http://sourceforge.net/projects/ipodlinux/)) I wouldn't be sure that they'd just port some other OS to use the new platform since Cell is completely new in every aspect a chip can be. Basically it'd be like trying to fit a square block into a triangular hole if they'd tried to port another OS to Cell instead of making their own.

Oh and I still don't understand that network computing thing they're promoting. Doesn't that require an insanely fast LAN between the chips to actually work in an efficient way?

Geta-Ve
05-03-2005, 07:50 AM
I think the cell stuff is fascinating. All lame marketing stuff aside this tech is going to rock with a sock on its cock. :bowdown:


er.. I don't think you can say that..

DrFx
05-03-2005, 10:09 AM
I'd like to see some Cell-based renderer and PCI card, sort of like Gelato or Pure. It would be easier than adapting a whole OS to it.

Coliba
05-03-2005, 05:02 PM
Do renderers need double precision floating point calculations or are they mostly satisfied with single precision? Even with dp the Cell would be so fine for rendering.....mmmm.....if Sony can put a Cell in a 400$ Playstation, why couldn't we have a PCI card with 2 Cells on it, for 1000$? Would be the equivalent of 10-15 Opterons 240 maybe? And it could be used for other things besides rendering.

BigJay
05-03-2005, 07:02 PM
From the article they probably built an OS to develop apps for the cell. Would be great if that is something that comes on the PS3.


Will be interesting to see what comes out of all that power.

novadude
05-03-2005, 07:11 PM
Would be great if that is something that comes on the PS3.

An OS that lets you do anything else (like use the console as a PC) would be great for every game console, but I doubt we'll ever get what we want.

js33
05-03-2005, 07:24 PM
.....if Sony can put a Cell in a 400$ Playstation, why couldn't we have a PCI card with 2 Cells on it, for 1000$? Would be the equivalent of 10-15 Opterons 240 maybe? And it could be used for other things besides rendering.

If you think 2 Cell processors would be equivalent to 10-15 Opterons I think you will be severely disappointed when they come out.

Cheers,
JS

Coliba
05-03-2005, 07:49 PM
Well, I'm no expert on cpu's and system architecture but from what I'm reading in different articles and forums, a 2.2ghz Opteron has a peak gflops performance of around 5? I assume that's double precision? The Cell was announced from 256 gflops sp and around 26 gflops dp. So at least 5 times more than an Opteron. But it's not just the raw power of the Cell that's exciting to me, but this whole concept of a "system on a chip", the extremely large memory bandwidth and fast connection possibilites that makes it in my view perfect for rendering. That's why I wanted to know if renderers really need dp floating point calculations or if sp is enough? Even using exclusively dp the Cell is still very powerful, imagine if only sp is needed? Damn I hope somebody notices this CG market and develops some sort of PCI card or an box where you can just add more single Cell chips instead of a completely new box along with motherboard, case, ventilation, psu etc etc etc....

chadtheartist
05-03-2005, 08:29 PM
I think in the future there may be in fact on board processors designed for rendering. Longhorn, and OS X both use, or will use GPU's for rendering the GUI, so why can't other apps be used to do the same thing, but use something like the Cell instead?

subcast
05-03-2005, 08:38 PM
The simplest explanation of the cell is they are scaleable CPUs. Meaning you could have 1 or 10 or 100 or 1000 cells all dedicated to a single task. that could virtually eliminate rendering as we know it now.

rendermania
05-03-2005, 08:40 PM
There already is prototype real-time time raytracing/GI hardware, like what these folks are working on:

http://www.openrt.de/

Beamtracer
05-03-2005, 09:44 PM
As darkness falls over the city, and the office workers leave their skyscrapers to commute home, all those thousands of PCs in those towers lie idol. This is wasted computing capacity.

A Cell network could keep these processors busy. After hours they could be combined for some serious number crunching, either by the company that owns them, or they could be placed on the national CPU grid for others to use.

At the moment, if someone has electricity generating ability they can sometimes sell their excess power back to the national electricity grid. In the future the same may apply to computing. Cell processing makes it all possible.

Brave new world!

Para
05-04-2005, 10:38 AM
As darkness falls over the city, and the office workers leave their skyscrapers to commute home, all those thousands of PCs in those towers lie idol. This is wasted computing capacity.

Usually yes, but I actually know a certain small ad company that uses even their laptops and handhelds to render when the deadline comes a bit too close.

I still haven't heard a clear answer where people assume the bandwidth required for nation wide multi-Cell computers is going to be taken from. The amount of needed bandwidth isn't trivial, it's more like a few gigabytes per second per core.

halo
05-04-2005, 10:54 AM
ok, so the output device was 1920 pixels wide...what res was each piece of footage?

Coliba
05-04-2005, 11:41 AM
Well, if they decoded SDTV rez, that would be 720x483 I guess.

Thalaxis
05-04-2005, 04:56 PM
Cell is completely new in every aspect a chip can be.

That's actually not true. There isn't really much about it that's truly new. The biggest "new" thing about it
is that it's a single-chip implementation. The PPC element is based on an older design, and even the
Amiga had a host + DSP's architecture.

None of that changes the fact that the Cell brings to the table an unprecedented amount of both bandwidth and computing power to the living room, and at an unprecedented price point, if you compare it to where
that sort of muscle was just five years ago (Cray Y-MP, NEC Earth Simulator, etc).

That has a LOT of advantages, of which manufacturing cost is only one. A major one is internal
interconnect bandwidth. Take that out, and those 48 streams would have been 2... which by itself is
impressive. For some applications, that's the important thing, and the rest of the computing power is just
icing on the cake :)

Frankly, I'd be surprised if IBM didn't already have teams working on a Linux port.

I'm guessing that the PC will be the niche gaming platform by the end of the decade, especially if
the XBox360 is competitive with the PS3.

Thalaxis
05-04-2005, 05:03 PM
A Cell network could keep these processors busy. After hours they could be combined for some serious number crunching, either by the company that owns them, or they could be placed on the national CPU grid for others to use.


True, but has absolutely no relationship to Cell.

richcz3
05-04-2005, 06:14 PM
Until a company actually comes to market with a Cell based product everything is pure speculation. The big press event last year with it's ominous (hype) overtones are all we have to deduce a parctical consumer end use. Toshiba will use the Cells in their next Gen TVs.

The best bet is that they (Sony, Toshiba, and IBM) will produce modular components (like daisey chained computing) for the living room. The same direction Microsoft is looking to market their next gen console. It's all about media centers.

Coliba
05-04-2005, 06:49 PM
Media centers will surely be the Cell's biggest market, but I just don't want to give up hope we'll be able to use it for CG as well :) There are various indications that IBM and Sony won't limit it to media centers and hdtv's:
"The Cell-based workstations we are creating with Sony and SCEI will deliver scalable, supercomputer-like performance to the media, entertainment and video game industries," said John Kelly, IBM's senior vice president and group executive.

IBM will develop the Cell-based workstations while the Sony group develops the operating environment."


This article also has some interesting points:
http://neasia.nikkeibp.com/neasia/001090

"We wanted to make an architecture that would be valid for at least a decade," said James Kahle, IBM fellow, Broadband Processor Technology, Microelectronics Div, IBM Corp of the US, emphasizing the future-oriented design of the chip. The prototype chip is merely the first step in realizing this goal, merely a starting point."

"Each company is involved in its own software development project, and it appears, for example, that multiple varieties of Linux running on Cell already exist."

PhilOsirus
05-04-2005, 09:50 PM
I'm guessing that the PC will be the niche gaming platform by the end of the decade, especially if the XBox360 is competitive with the PS3.

It already is a niche gaming platform.

richcz3
05-04-2005, 11:19 PM
It already is a niche gaming platform.
Having played PC games since way back before Multimedia PCs (1992) were the big thing. I would agree with Phil. The PC is niche gaming platform. As a gaming platform it is its own worst enemy.
Try keeping up with the Jones's for 14 years. You learn to see the real value of a Console system.

Beamtracer
05-05-2005, 12:01 AM
Frankly, I'd be surprised if IBM didn't already have teams working on a Linux port.

Yes, IBM is pouring huge amounts of money into Linux, so you'd expect that a Cell port would be part of it, which they'll hand over to the open source community when they have finished it.

This makes me wonder whether Apple would have been better off using the Linux kernel (and putting their Aqua GUI renderer on top of it) rather than using the BSD UNIX kernel as they do. Then they could have got a free ride on all this development.

Thalaxis
05-05-2005, 02:28 AM
Yes, IBM is pouring huge amounts of money into Linux, so you'd expect that a Cell port would be part of it, which they'll hand over to the open source community when they have finished it.


That last part is a big maybe. They're not in it for the open source community, they're in it
for the consulting business.


This makes me wonder whether Apple would have been better off using the Linux kernel (and putting their Aqua GUI renderer on top of it) rather than using the BSD UNIX kernel as they do. Then they could have got a free ride on all this development.

The difference would have been having access to IBM's compilers, rather than trying to
roll their own. Other than that, it wouldn't have changed much.

It's also possible that it would have delayed their launch considerably since NeXTStep was
based on BSD, so staying with BSD might have made the port easier.

Coliba
05-05-2005, 10:37 PM
Sorry to bring this up again, but I thought it might interest the techies ->news about an IBM developer releasing "preparation" patches for linux kernel to run the BPA(Broadband Processor Architecture (Cell)) on the PPC64 platform:

http://lkml.org/lkml/2005/4/19/191

"This series of patches adds a bit of infrastructure in preparation of
getting the Broadband Processor Architecture (BPA) into the kernel as
a new platform type of ppc64.
BPA is currently used in a single machine from IBM, with others likely
to be added at a later point.

None of these preparation patches are really specific to the
architecture itself. Hopefully, I will be able to send the actual
platform code really soon now.

BPA and pSeries can share some code, mostly because they are both
using rtas. The first two patches are splitting out the common
code from the pSeries_pci implementation into a generic rtas_pci
base.
The nvram and watchdog drivers are pretty generic and are first used
in the new machine."

And here an early version of the SPU file system:
http://groups-beta.google.com/group/fa.linux.kernel/msg/9849e6f8455f0e57

This is looking good :) :)

Beamtracer
05-06-2005, 09:52 PM
->news about an IBM developer releasing "preparation" patches for linux kernel to run the BPA(Broadband Processor Architecture (Cell)) on the PPC64 platform

Does this mean that the first desktop OS to run on a Cell processor will be Linux?
Would regular Linux applications run on it?

BigJay
05-07-2005, 01:57 AM
The great thing about linux is that you can recompile alot of apps to work on any processor. Look at fink and x11 on the Mac. It is downloading code people are using on x86 processors and running it on G4s and G5s. I am not saying all of it is possible but if IBM releases the neccessary libraries for their processor and related hardware there is no reason linux won't run on it.

They have linux running on iPods...lol it's a pretty portable OS.

Thalaxis
05-07-2005, 02:35 AM
Does this mean that the first desktop OS to run on a Cell processor will be Linux?
Would regular Linux applications run on it?

Once the libraries are ported, most applications should be compile compatible with Cell.

Beamtracer
05-07-2005, 04:40 AM
Once the libraries are ported, most applications should be compile compatible with Cell.
Well, there's a lot of speculation about whether or not Apple will adopt Cell, considering that IBM is their main supplier of processors. Some have said that one reason that Apple may not do it is because all their apps would need to be rewritten for Cell.

Maybe that wouldn't be the case after all. Maybe it just needs work done to modify the OS, and then existing apps would run. I imagine that whatever applies to Mac OS X and its BSD UNIX underpinnings would be the same situation that exists for Linux.

Thalaxis
05-07-2005, 04:50 AM
Well, there's a lot of speculation about whether or not Apple will adopt Cell, considering that IBM is their main supplier of processors. Some have said that one reason that Apple may not do it is because all their apps would need to be rewritten for Cell.


One thing the Cell is not is a general-purpose processor. It's a clock-speed optimized
design specifically built for gaming and multimedia.


Maybe it just needs work done to modify the OS,

Just? As if modifying the OS is a simple task...

Beamtracer
05-07-2005, 06:40 AM
One thing the Cell is not is a general-purpose processor. It's a clock-speed optimized
design specifically built for gaming and multimedia.
I thought IBM had "bet the farm" on Cell. They must have intentions way beyond gaming and multimedia.


Just? As if modifying the OS is a simple task...
It's still easier to modify the OS rather than the OS + Apps.

In Apple's case, they made developers write for OS X. It would be a tough call to make developers modify their apps again for Cell.

I wonder if Apple is big enough to undertake such a major rework of the operating system to use Cell. Then again, larger companies making operating systems seem to have more difficulty upgrading them in a reasonable time frame.

What will Cell mean for Microsoft? I guess they are too wedded to Intel and Dell to ever switch to something else.

But imagine if MS doesn't go with Cell, and Cell turns hype turns out to be true, and it really is a real break though in speed. I think there would be a lot of Windows users around here who would be willing to switch to Linux if it meant they could gain huge speed increases with Cell.

It has the potential to leave MS behind for heavy rendering tasks.

Thalaxis
05-07-2005, 01:18 PM
I thought IBM had "bet the farm" on Cell. They must have intentions way beyond gaming and multimedia.


That doesn't change anything about the current implementation. The current memory
limit, for example, is physically hard-wired. For IBM's bread and butter, it's also going to
need MUCH more cache and won't have nearly as much use for all those vector engines.
In one of IBM's server systems, that would just be a heat generator.


It's still easier to modify the OS rather than the OS + Apps.


Slightly.


In Apple's case, they made developers write for OS X. It would be a tough call to make developers modify their apps again for Cell.


It probably wouldn't do wonders for the availabiliity of third party software.


I wonder if Apple is big enough to undertake such a major rework of the operating system to use Cell. Then again, larger companies making operating systems seem to have more difficulty upgrading them in a reasonable time frame.


Not likely. It would be far easier to add Cell to personal computers through add-in boads,
Microway-style.


What will Cell mean for Microsoft? I guess they are too wedded to Intel and Dell to ever switch to something else.


They're too wedded to the software that's already out there.


But imagine if MS doesn't go with Cell, and Cell turns hype turns out to be true, and it really is a real break though in speed. I think there would be a lot of Windows users around here who would be willing to switch to Linux if it meant they could gain huge speed increases with Cell.


Around here, it's possible. But 95% of the people who buy computers don't need or want
more power than they already have. Those people aren't going to buy anything new until
their existing machines give up the goose, or the new machines offer them something
they want enough to pay for it. Speed only matters to the high end.


It has the potential to leave MS behind for heavy rendering tasks.

It's possible, but it's not likely to happen quickly, and definitely not with the PS3 version
of the Cell. It's going to be a wonderful gaming system, but it's so specifically optimized
for that that it isn't likely to be able to make a dent in anything but the gaming market,
though between it and the XBox360, I think the gaming market will be turned on its ear.
It's going to be an exiting year for gaming.

CGTalk Moderation
05-07-2005, 01:18 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.