StarDestroyer.Net BBS

Newsy Woozy

The Cell processor, which will drive Sony's PlayStation 3, will run 10-times faster than current PC chips, its designers have said.



Sony, IBM and Toshiba, who have been working on the Cell processor for three years, unveiled the chip on Monday.

It is being designed for use in graphics workstations, the new PlayStation console, and has been described as a supercomputer on a chip.

The chip will run at speeds of greater than 4 GHz, the firms said.

By comparison, rival chip maker Intel's fastest processor runs at 3.8 GHz.

Details of the chip were released at the International Solid State Circuits Conference in San Francisco.

The new processor is set to ignite a fresh battle between Intel and the Cell consortium over which processor sits at the centre of digital products.

The PlayStation 3 is expected in 2006, while Toshiba plans to incorporate it into high-end televisions next year.

IBM has said it will sell a workstation with the chip starting later this year.

More...

Nice...

I was just going to post on this.

Also, read this. Excellent technical reading (though the guy overhypes a little bit).
http://www.blachford.info/computer/Cells/Cell0.html

The Cell chip is basicly a PowerPC or POWER core with multiple cell APU's (the main core can be PowerPC 970 aka G5, and for servers POWER4 or POWER5). It's amazing they've managed to get the G5 running to 4 GHz (or at least will get by 2006, I should say).

A Cell chip with 9 cores (one G5 and eight APUs) can achieve a whopping 256 gigaflops according to the article.

Wowza! This had *BETTER* go in PowerMacs by WWDC 2006. Goodbye, x86...

Some nice pictures:
The Cell chip (where the PU is the PowerPC G5 core):
Possible Cell Architecture Picture

Distributed processing:
Distributed Cell Picture

Don't inline large images -- Phong

If it's 10 times faster why don't they just adapt it to PCs and put intel/AMD out of business?

I want benchmarks. All I see right now is hype.

From what I've read, it would depend on what you're doing.

Video editting probably would be, oh, 10x faster, but something such as emulation would not be and would just use the PowerPC processor. The speed boost on the APU's depend on the task being done. It's supposed to be especially good for multimedia (3d rendering, gaming, video editting, etc).

Yeah, yeah, yeah. I remember Sony's hype machine last time. Remember how the PS2 was going to be able to render FF8 cutscenes in real time? Remember how it was so fucking powerful that it couldn't be allowed to fall into terrorist hands? Remember how the fact that it only had 4 MB of VRAM was "irrelevant" because textures could be bled directly off the DVD?

I've also heard the line about some new technology that will blow everything out of the water when it comes out and increase computing power by an order of magnitude (remember Talisman?), only it never quite seems to work out that way. Either it was massively overhyped, or by the time it does come out years after they said it would, the status quo of computing has caught or surpassed it.

So I take wild-eyed talk about PS3 and Cell with a mountain of salt.

Arthur_Tuxedo wrote:Yeah, yeah, yeah. I remember Sony's hype machine last time. Remember how the PS2 was going to be able to render FF8 cutscenes in real time? Remember how it was so fucking powerful that it couldn't be allowed to fall into terrorist hands? Remember how the fact that it only had 4 MB of VRAM was "irrelevant" because textures could be bled directly off the DVD?

I've also heard the line about some new technology that will blow everything out of the water when it comes out and increase computing power by an order of magnitude (remember Talisman?), only it never quite seems to work out that way. Either it was massively overhyped, or by the time it does come out years after they said it would, the status quo of computing has caught or surpassed it.

So I take wild-eyed talk about PS3 and Cell with a mountain of salt.

I do know what you mean, which is why I'm looking more to IBM's specs than Sonys.

That article above says 4 or 8, but what I've seen says that the PS3 will have two cells in its Cell processor (making it more powerful than the other consoles but vastly less powerful than a 'real' cell computer).

I'm expecting it to be maybe twice as fast as the XBox 2 (assuming they're going with the single G5 rumored). I'm more excited about the Cell's application in Macs than in the PS3.

After all, since it has an existing PowerPC chip (at 4 GHz) in it, it can only be faster than existing ones, right?

From my experience with ASIC chips and the quest for speed at our company, the silicon is only as good as the programming instructions that take advantage of the architecture, and that looks more true than ever with this 8-APU architecture of the Cell.

We already see some interesting possibilities with the right coding for security appliances, offloading VPN and IDS/IPS processes to specific APU's while peerforming core firewalling and task assignment on the CPU, for instance. Cool stuff.

It's quite large for a chip, so it will most likely be not very cheap.
And vector processors are a wee bit tricky to program for. They'll be pretty fast if you have a problem that normally runs quite fast in environment like that, but I don't see it replacing x86 anytime soon...

Praxis wrote: That article above says 4 or 8, but what I've seen says that the PS3 will have two cells in its Cell processor (making it more powerful than the other consoles but vastly less powerful than a 'real' cell computer).

Sony has been saying 4-8 from the beginning.

I'm expecting it to be maybe twice as fast as the XBox 2 (assuming they're going with the single G5 rumored). I'm more excited about the Cell's application in Macs than in the PS3.

It will obviously have vastly superior vector FP performance, but so what? The PS3 design isn't like the PS2, it uses an off the shelf nVidia GPU which means that unlike the PS2, the CPU on the PS3 won't be used for any of the graphics pipeline tasks. I'm not sure what Sony is hoping to use all of those vector units for, but it if it's not going to be for graphics, I doubt there will be much functional difference between the PS3 and the Xbox 2 except in certain specialized areas (background/AI processing for example).

After all, since it has an existing PowerPC chip (at 4 GHz) in it, it can only be faster than existing ones, right?

No, the PowerPC parts of the core are not performance compareable to any current iteration of PowerPC.

EDIT: One other thing, we don't know wether the entire Cell chip will run at 4Ghz or if that is just the APU's. Remember, the P4's ALU's currently run at over 7GHz on the fastest models, they are double clocked over the rest of the processor.

Let me expand upon this for a sec. IBM has created a CPU that rapidly accelerates FP ops through a series of parallel execution units, this is a similar tactic to what a modern GPU does, only it is more versatile in the types of processing it can do. This is a boon for processing tasks that are heavily parallelized, but it's not going to do all that much good in a game console where all of the highly parallelized tasks are already offloaded to another processor (the GPU).

The reason that this approach worked in the PS2 (which also uses similar vector units) is because it had a dumb graphics chip that took care of only the rendering stage of the graphics pipeline, whereas all of the setup, lighting and geometrey was handled by the CPU.

So why did Sony choose the nVidia chip for this application? Because dedicated graphics chips are still tops for graphics processing because their math units are far more specialized towards graphics. The Cell may be fast, but it can't do as good of a job as the shader and geometry units inside an NV40.

InnocentBystander wrote:If it's 10 times faster why don't they just adapt it to PCs and put intel/AMD out of business?

Emulation will be hell.

Its not that easy to switch everything from the current programming and hardware into CELL based stuff overnight, just look how hard it is to even change standerds in the PC world.

Praxis wrote:I'm expecting it to be maybe twice as fast as the XBox 2 (assuming they're going with the single G5 rumored). I'm more excited about the Cell's application in Macs than in the PS3.

The only feasible application a Cell processor would have in Macs would be as some sort of dedicated media processor for high-end workstations. You won't be seeing Cell-based PowerMacs.

Ace Pace wrote:

InnocentBystander wrote:If it's 10 times faster why don't they just adapt it to PCs and put intel/AMD out of business?

Emulation will be hell.

Its not that easy to switch everything from the current programming and hardware into CELL based stuff overnight, just look how hard it is to even change standerds in the PC world.

They already are actually, Intel has recognized the need for add on vector processing to x86, which is what the SSE units are all about. They don't do nearly as good a job as Cell though, after this expect to see new x86 add ons which add significantly to the die space and provide the sort of vector processing we see from Cell.

Durandal wrote:

Praxis wrote:I'm expecting it to be maybe twice as fast as the XBox 2 (assuming they're going with the single G5 rumored). I'm more excited about the Cell's application in Macs than in the PS3.

The only feasible application a Cell processor would have in Macs would be as some sort of dedicated media processor for high-end workstations. You won't be seeing Cell-based PowerMacs.

Cell does have general purpose applicability, but that doesn't really matter, it would be far easier to just bolt these kinds of vector units onto a modern PowerPC core and use specialized accelerators for server clusters. IBM has been doing this in select applications for years, the thing holding it back is that often the accelerator chips are far too specialized to be of much use to a wide range of HPC processing.

The Kernel wrote:
They already are actually, Intel has recognized the need for add on vector processing to x86, which is what the SSE units are all about. They don't do nearly as good a job as Cell though, after this expect to see new x86 add ons which add significantly to the die space and provide the sort of vector processing we see from Cell.

How long till X86 reachs a point where Intel and AMD Must abandon it?

Ace Pace wrote: How long till X86 reachs a point where Intel and AMD Must abandon it?

Intel and AMD have been very good at adapting the x86 instruction set to modern times. When RISC came around, everybody said x86 was doomed, but Intel simply modified the x86 core to use micro ops which let it behave internally like a RISC processor.

Over the years, Intel and AMD have introduced a load of improvements to the x86 core, and because it is so widespread, they have been able to dedicate massive resources to it. There is no reason to believe that the x86 core can't last for another two decades or even longer, there has been no critical stumbling block that has not allowed x86 to be expanded upon thusfar. There are a few inefficiencies that are inherent to x86 which make me long for its eventual replacement, but they aren't crippling the modern x86 core's ability to perform.

This will be useful:

http://arstechnica.com/articles/paedia/cpu/cell-1.ars

Feel the power of hype
"The chip will run at speeds of greater than 4 GHz, the firms said.

By comparison, rival chip maker Intel's fastest processor runs at 3.8 GHz."
And this is important, how?
-Clock isn't everything
-In more than a year, there will be faster chips from AMD and Intel
"ten times faster"
Doing what?
Faster than what current pc chips? A Pentium 4? An Athlon 64 Fx? An UltraSPARC?

Double-post nuked -- Phong

Praxis wrote:I was just going to post on this.

Also, read this. Excellent technical reading (though the guy overhypes a little bit).
http://www.blachford.info/computer/Cells/Cell0.html

ArsTechnica disagrees with the notion that OS News' article was accurate.

InnocentBystander wrote:If it's 10 times faster why don't they just adapt it to PCs and put intel/AMD out of business?

The first Cell based product will be a workstation, primarily aimed at content creation.

Sony want to put Cell chips in everything. everything.

As an aside, RAMBUS have been talking today about their next generation memory bus, which will provide all the silly bandwidth that the Cell is going to need. It's called FlexIO, and has an 8GHz data rate.

Their next gen PC bus, Yellowstone, will run at 3.2GHz, scaling to 8GHz by 2006.

phongn wrote:

Praxis wrote:I was just going to post on this.

Also, read this. Excellent technical reading (though the guy overhypes a little bit).
http://www.blachford.info/computer/Cells/Cell0.html

ArsTechnica disagrees with the notion that OS News' article was accurate.

He also wrote up a refutation to that, and though I agree with most of the points in the Ars Technica article, they're mostly just showing that some of his exclamations are hype (SETI@Home units in 5 minutes for example), but the overall article is accurate.

Durandal wrote:

Praxis wrote:I'm expecting it to be maybe twice as fast as the XBox 2 (assuming they're going with the single G5 rumored). I'm more excited about the Cell's application in Macs than in the PS3.

The only feasible application a Cell processor would have in Macs would be as some sort of dedicated media processor for high-end workstations. You won't be seeing Cell-based PowerMacs.

Um...why is this? The chip is SCALABLE, remember?

Melchior wrote:Feel the power of hype
"The chip will run at speeds of greater than 4 GHz, the firms said.

By comparison, rival chip maker Intel's fastest processor runs at 3.8 GHz."
And this is important, how?
-Clock isn't everything
-In more than a year, there will be faster chips from AMD and Intel
"ten times faster"
Doing what?
Faster than what current pc chips? A Pentium 4? An Athlon 64 Fx? An UltraSPARC?

Double-post nuked -- Phong

I believe it's referring to the PowerPC 970 core. The G5 core runs at 4 GHz.

Then ten times faster thing is mostly hype I'd guess.