[Majost]

Here's the IBM release for the cell stuff: http://www-03.ibm.com/chips/news/2004/1129_cell1.html

IBM, Sony Corporation, Sony Computer Entertainment Inc. (Sony Corporation and Sony Computer Entertainment Inc. subsequently referred to as Sony Group) and Toshiba Corporation today unveiled, for the first time, some of the key concepts of the highly-anticipated advanced microprocessor, code-named Cell, they are jointly developing for next-generation computing applications and digital consumer electronics. The four companies also announced that they would reveal technical details of Cell at the International Solid-State Circuits Conference (ISSCC) to be held from February 6-10, 2005 in San Francisco. Specifically, the companies confirmed that Cell is a multicore chip comprising a 64-bit Power processor core and multiple synergistic processor cores capable of massive floating point processing. Cell is optimized for compute-intensive workloads and broadband rich media applications, including computer entertainment, movies and other forms of digital content. Other highlights of the Cell processor design include: Multi-thread, multicore architecture. Supports multiple operating systems. Substantial bus bandwidth to/from main memory, as well as companion chips. Flexible on-chip I/O (input/output) interface. Real-time resource management system for real-time applications. On-chip hardware in support of security system for intellectual property protection. Implemented in 90 nanometer (nm) silicon-on-insulator (SOI) technology. Additionally, Cell uses custom circuit design to increase overall performance, while supporting precise processor clock control to enable power savings. ... Cell provides a breakthrough solution by adopting a flexible parallel and distributed computing architecture consisting of independent floating point processors for rich media processing. Cell supports multiple operating systems, including PC/WS operating systems, as well as real-time CE/Game operating systems. In addition, the Cell processor is scalable and can be utilized in a variety of applications — from small digital CE systems within the home to entertainment applications for rendering movies, to scientific applications, such as supercomputers. A team of engineers from IBM, Sony Group and Toshiba are collaborating on the design and implementation of Cell which is expected to deliver vast floating point capabilities, massive data bandwidth and scalable, supercomputer-like performance. The design work is taking place at a joint development lab the three companies established in Austin, Texas, after the project was announced in 2001. IBM plans to begin pilot production of Cell microprocessors at its 300mm wafer fabrication facility in East Fishkill, NY during the first half of 2005. The first computing application IBM plans for Cell is the Cell processor-based workstation it is developing with SCEI.

Not only that, but IBM, Sony and SCEI power-on Cell processor-based workstation prototype, too. Here's a particularly interesting quote about that:

IBM, Sony Corporation (Sony) and Sony Computer Entertainment Inc. (SCEI) announced today that they have powered-on the first Cell* processor-based workstation. The prototype workstation is the first computing application planned for the highly-anticipated Cell processor. The companies expect that a one rack Cell processor-based workstation will reach a performance of 16 teraflops or trillions of floating point calculations per second.

16 TF in one Rack!!!

But this all happened yesterday... and people have hinted at it before. So, what's new today?

[micmoo]

Is there any technical reason (or business reasons) that would prevent Apple from using this technology? Windows cannot run on PPC, so would that leave Linux for the Sony workstation? It seems if Apple can use this tech, OSX would be the ultimate OS to run the cell!

[Barto]

Because Cell is heavily distributed, software will have to be modified to run well on the platform. It will require some effort, but hey, the 68k to PPC transition worked out fine, as did the Classic to X transition.

Anyone else up for another "short term pain, long term gain" move by Apple?

[DrGruv]

Cell supports multiple operating systems, including PC/WS operating systems, as well as real-time CE/Game operating systems. In addition, the Cell processor is scalable and can be utilized in a variety of applications — from small digital CE systems within the home to entertainment applications for rendering movies, to scientific applications, such as supercomputers.

Didn't directly see apple listed... would apple WANT to use this?

Why? (with power5's around the corner and of course 970mp)

-but perhaps for media creation and testing on a g5/g6?

Would this technology make its way into the next gen apple chips?

morpheus?

[Henriok]

Originally Posted by DrGruv Didn't directly see apple listed... would apple WANT to use this?

Apple usually don't show up on ANY lists hinting to new and exotic hardware. It's not their way to pre announce stuff like this. And.. Mac OS X _is_ a PC/WS operating system if I'm reading the acronyms correctly.

[DrGruv]

16TF in one rack -

I wonder how many chips?

$$$$$

20 chips X let's say $400 per chip (very high) = $8000 + $4000 to $8000 for subsystems and box?(i have no idea) = $16000+- for a 16TF computer!

WOW

Talk about a shift in power, the virginia tech computer with all of the xserves does a 12 to 14TF or in that range...

[micmoo]

this is killing me not knowing if Apple will take advantage of this. Anyone have any info? I'll take wild speculation at this point...

[Quagmire]

That cell thing was released yesterday so it wouldn't be what morpheus is talking about. Also, I wonder if the supply problems will only be with the 970fx and not the new 970GX,970MP, and G5 mobile. Morpheus?

[Programmer]

I wonder if the supple problems will only be with the 970fx

You think the 970fx was too easily bent?

[thegelding]

could somebody explain the cell processor for us tiny brains?

thanks

g

[Quagmire]

Originally Posted by Programmer You think the 970fx was too easily bent?

fixed. I meant supply.

[DrGruv]

cell - it's going to be 4 processors in one and cost under $100 according to an article i just read in USA today

[FallenFromTheTree]

More than one UNIX based OS like LINUX & LONGHORN?

And this too!


The processor, code-named Cell, will handle vastly more memory than today's consumer chips as well as
<<enable hardware-based copyright protection>> and allow multiple operating systems to run at the same time. It also will feature multiple cores, or logic engines, on a single die.

[Morpheus]

Ok, try now:http://power.org/

"Coming together to form the Power.org community are AMCC, Bull, Cadence Design Systems, Chartered Semiconductor Manufacturing, Culturecom, IBM, Jabil Circuit, Novell, Red Hat, Sony Corporation, Shanghai Belling, Synopsys, Thales, Tundra Semiconductor and Wistron."

mmm... where is Apple? Shouldn't they be in?

Morpheus

[Kurt]

Originally Posted by DrGruv cell - it's going to be 4 processors in one and cost under $100 according to an article i just read in USA today

IBM is going to do all this for $100 but they can't make enough 970fx chips for Apple's needs. I find this cell chip a little hard to believe. In fact with the performance of IBM in producing 90 nm chips for Apple, I don't see how they can do half the things they are promising. Cell processors, Xbox 2, and I thought Nintendo too. In some ways it doesn't seem possible.

[Quagmire]

Originally Posted by Morpheus Ok, try now:http://power.org/ "Coming together to form the Power.org community are AMCC, Bull, Cadence Design Systems, Chartered Semiconductor Manufacturing, Culturecom, IBM, Jabil Circuit, Novell, Red Hat, Sony Corporation, Shanghai Belling, Synopsys, Thales, Tundra Semiconductor and Wistron." mmm... where is Apple? Shouldn't they be in? Morpheus

Apple will begin to use Cell processors? Or apple might go back to Moto?(shivers)

[Jay]

Originally Posted by Undertaker Apple will begin to use Cell processors? Or apple might go back to Moto?(shivers)

A third possibility is that Apple and IBM are working on something different than Cell.

[Henriok]

Originally Posted by Morpheus mmm... where is Apple? Shouldn't they be in?

What about Freescale, Nintendo, Microsoft and Toshiba? Don't they want to play with the other Power people?

What does this mean for us? Will we get more information from IBM regarding the future of PowerPCs?

[iSteve]

from Power.org, december 2 news
"In a related product development, TimeSys, headquartered in Pittsburgh, PA, announced the availability of its TimeStorm Linux Development Kits for IBM PowerPC 750 FX and GX. Thousands of embedded Linux developers worldwide from a range of industries have chosen TimeSys solutions to streamline the development, customization and validation of homegrown or commercial Linux-based embedded systems."
750 GX PM in january? Morpheus....

[MCQ]

Huh? Isn't a 750GX something of a G3 type variant?

[iSteve]

Originally Posted by MCQ Huh? Isn't a 750GX something of a G3 type variant?

That's right... I find it strange too... the news is dated dec 2 2004.
In fact I wrote "750 GX PM" but it would have been better "750 GX Mac".
I've always tought the GX variant of the 750 was nothing but vaporware!

[Henriok]

Originally Posted by iSteve I've always tought the GX variant of the 750 was nothing but vaporware!

The 750GX has been shipping for some time AFAIK.
You might think of the 750VX that's pretty vaporous.

Apple skipped the 750GX and wen't directly to the G4, and that's probably why we havn't seen much of it in the Mac related press. It has larger cache than the 750FX and is clocked a bit higher and supports a 200 MHz bus. It's still a 130 nm part though and I'm amazed that we havn't seen a 90 nm 750 class processor yet. It'd be a killer in the embedded market.

[iSteve]

Originally Posted by Henriok The 750GX has been shipping for some time AFAIK. You might think of the 750VX that's pretty vaporous. Apple skipped the 750GX and wen't directly to the G4, and that's probably why we havn't seen much of it in the Mac related press. It has larger cache than the 750FX and is clocked a bit higher and supports a 200 MHz bus. It's still a 130 nm part though and I'm amazed that we havn't seen a 90 nm 750 class processor yet. It'd be a killer in the embedded market.

yep... my fault. with all this fx, gx , vx stuff I messed up the processor family referring the GX step of the 750 to the GX step of the 970...
Not bad... I've made 4 posts and 3 of them are total bullshit... :-))) nice record!

[Henriok]

Originally Posted by iSteve Not bad... I've made 4 posts and 3 of them are total bullshit... :-))) nice record!

Haha No worries. For what it's worth the "GX" moniker seems to indicate some fabrication enhancements and increased cache and frequency . That's pretty much what we are hoping the 970GX will bring us.

[thegelding]

nice site...lots of hype...kinda hard for me to find the info i want...too dim i guess

here

g

[CoreMac]

There is some speculation at a certain place (sorry name escapes me ) that Cell tech could work with the G5 as a sort of next generation Altivec. Possible?

[Programmer]

Considering we know practically nothing about the Cell except that at least some part of it uses the POWER ISA, it is a possibility. The Cell's vector processors aren't really like AltiVec -- they are seperate cores with a (presumably) different ISA. AltiVec is an extension to the PowerPC ISA and therefore runs as part of a PowerPC core.

If that isn't clear, try this: a processor core (or SMT thread) follows its own stream of instructions that controls what it does. The instructions that it understands are its ISA (instruction set architecture), i.e. a list of specific instructions that tell the core what to do. Adding AltiVec created the ability to insert a new set of instructions into the stream for the G4/G5 PowerPC core. Adding Cell vector processors adds additional cores that each have their own stream of instructions which they execute at the same time as the POWER core. The instructions found in the streams for the vector processors are different than those found in the POWER core's stream, just like a POWER core and an x86 core have different instructions in their streams. It is possible that IBM/Sony/Toshiba used some portion of the existing POWER instructions, but I doubt it as that would serve little purpose -- either a core understands the whole stream, or it will crash while trying to execute it (or emulate it really slowly).

What Cell vector processors and PowerPC AltiVec (and Intel's SSE/SSE2/..., and most modern GPUs) do have in common is 128-bit registers which can hold, among other things, 4 32-bit floating point numbers at a time. These registers are used with SIMD instructions, which stands for "Single Instruction Multiple Data" -- i.e. a single instruction does the same thing to all the data elements found in the register (i.e. the 4 32-bit floating point numbers). This is as opposed to conventional instructions which typically do one thing to one piece of data, i.e. a register that holds 1 value.

I ask for the pardon of technical readers for glossing over details and over-generalizing things.

[Chinney]

Originally Posted by Programmer ...SMT thread... ISA (instruction set architecture)... instructions found in the streams for the vector processors...Intel's SSE/SSE2.... 128-bit registers which can hold, among other things, 4 32-bit floating point numbers...SIMD instructions, which stands for "Single Instruction Multiple Data"...the 4 32-bit floating point numbers.... I ask for the pardon of technical readers for glossing over details and over-generalizing things.

Do you also ask for the pardon of the non-technical readers, who don't know what the hey you are talking about? I rememer early on in AN history, Mr. Gelding saying how much he missed Programmer: although g never really understood what he was saying, he felt smarter after reading his posts. I second that.

In any case, Programmer's post here is certainly more comprehensible (to me) than some others he has posted and, for that matter, some of unixguru's posts. I note that one of the 'guru's recent messages concluded as follows

ohci_hcd 0000:21:00.1: OHCI Host Controller ohci_hcd 0000:21:00.1: irq 35, pci mem e000000080021000 ohci_hcd 0000:21:00.1: new USB bus registered, assigned bus number 2 usb usb2: Product: OHCI Host Controller usb usb2: Manufacturer: Linux 2.6.5-7.97-pseries64 ohci_hcd usb usb2: SerialNumber: 0000:XX:00.X hub 2-0:1.0: USB hub found hub 2-0:1.0: 3 ports detected NET: Registered protocol family 10 IPv6 over IPv4 tunneling driver st: Version 20040318, fixed bufsize 32768, s/g segs 256 eth0: no IPv6 routers present eth1: no IPv6 routers present

And this is just a small sample of that post.

I suppose that we have people of all ranges of technical knowledge here on AN. I've been meaning to post a poll to try to get a better idea of "who we are", at AN, in that regard. I'll get around to it one of these days.

[thegeek]

Ehh, the guru post was just a cut-n-past of the output from the dmesg command or boot.log file (some Linux distros have it). Nothing special except that this is some interesting hardware that just about nobody else has their hands on. Most of it isn't terribly relevant to anything we'll see in a Mac.

[dglow]

IBM and AMD have found a way to improve transistor performance by up to 24 per cent - without increasing the power draw - using a tweaked implementation of Big Blue's 'strained silicon' process. The process is not only going to be applied to upcoming AMD64 and PowerPC chips, but is likely to underpin the production of Sony's 'Cell' CPU and future Macintosh computers.

They're calling the technique 'DSL' - you can read the full Register article here.