Cell CPU as the next thing?

[takemehomegrandma]

Maybe in 2003 a cell processor would have had a chance as a desktop CPU

Cell was never meant (or suitable) as a general desktop CPU, not then, not now.

[koaftder]

The biggest problem with Cell is Sony. You can't really approach it for use in a product without entering into a business relationship with Sony and putting a gazillion dollars on the line. I think the complexity issue folks complained about is a cop out. People grapple with AltiVec, SSE, NEON, GPU stuff and all kinds of bizarre DSP platforms successfully and to good use.

[freqmax]

I agree that Sony is product flaw.

[polyp2000]

Cell was never meant (or suitable) as a general desktop CPU, not then, not now.

Not strictly true.

"The Cell Broadband Engine—or Cell as it is more commonly known—is a microprocessor designed to bridge the gap between conventional desktop processors"

Lets not forget that most of the Amigas commodore ever sold were powered by processors designed to go in Washing Machines

http://en.wikipedia.org/wiki/Cell_(microprocessor)

In fact Cell has a lot in common with PPC and in fact the CPU in the Xbox 360 albeit with only 3 cores actually has some common heritage with the Cell.

Cell can run a desktop operating system - there were versions of Yellow Dog Linux and Ubuntu running on PS3. Sadly along with memory limitations these were crippled by Sony's efforts to sandbox other operating systems from using the GPU. I always thought someone should have made some efforts to porting AROS or tried to get AmigaOS running instead. It would have run so much nicer on the PS3 as a desktop OS! alas that never happened, and the hacks to open up the system eventually led to Sony disabling the "OtherOS" feature from the console.

With its PowerPC heritage its already proved that Cell can work as a desktop processor if it is given the right resources.

Still - as many have already noted - Cell has a question mark hanging over its head in terms of future - although PPC looks like it still has some sort of roadmap for the future.

IMHO a new Amiga needs to look at whats happening with games consoles in terms of architecture. This kind of design is just like what the Amiga was all about. Custom chips helping out the Processor all in a dedicated , targeted platform enabled commodore to produce a computer that was both cheap and blew away the competition at the time. Thats the philosophy a new amiga needs to embrace.

In all honesty - if i had plenty of cash i might pony up for an X1000 but i dont really get what that Xena chip is all about or what kind of niche its all about?

So I pose another question..

Forget the CPU for a moment - What Custom chips should a new Amiga have that could make it stand out from other platforms on the market. Ill throw one into the  mix .

Howabout a chip that can specialise in encoding / decoding / transcoding Video / audio at many times faster than real time?

Any other ideas?

N...

[freqmax]

Forget the CPU for a moment - What Custom chips should a new Amiga have that could make it stand out from other platforms on the market. Ill throw one into the  mix.

I agree on this!

[commodorejohn]

Saying something outperforms a Pentium IV is not saying much at all. Pentium IIIs outperform P4s, clock-for-clock. There's a reason Intel dropped Netburst completely and went back to the P3 as the basis for designing the Core architecture.

[ferrellsl]

Cell was never meant (or suitable) as a general desktop CPU, not then, not now.

Actually, quite a few Linux fans would dispute your comment.  I know quite a few of them that ran Linux on their cell powered PS3s.....including me.  It ran quite well as a matter of fact but Sony wouldn't provide the details for hardware assisted graphics so after a while I decided to move on as better hardware (i.e. nVidia) and software came along, and now I just use my trusty PS3 as a dedicated Blu-ray player and media streamer.

[vox]

Saying something outperforms a Pentium IV is not saying much at all. Pentium IIIs outperform P4s, clock-for-clock. There's a reason Intel dropped Netburst completely and went back to the P3 as the basis for designing the Core architecture.

I do believe Cell is not as much performant, as parallel core computing CPU and there its more advanced then G5 chips.

So its great PPC chip, and it would be nice to see MorphOS and AMigaOS for it. Linux is already there.

[Duce]

Cell was never designed to go into desktop machines, it's simply unsuitable.

[freqmax]

I read that the Motorola 56000 @ 33 MHz that the Atari Falcon used enabled it to play mp3 in realtime, doing spectrum analyse using the sound input with 4096 point FFT. The performance is like 1,8 ms for a 1024 point FFT.

Maybe that's a hint as to what kind of periphial circuits that could be useful?

[takemehomegrandma]

@polyp2000, ferrellsl

Never ever have I claimed that the Cell couldn't be used as a desktop CPU, only that it never (not even when it was new) has been suitable for this, thus a poor choice when better alternative exist(ed). The main feature of this piece of silicon were the numerous SPE's, with a PPC mostly there to tie it all together. Thus you will pay a very high price for a CPU of which main features will go largely unused in a normal desktop context, while you only utilize the PPC (PPE) part of it for your needs, which is an insane waste of resources. This is why it's not suitable, and this is why it was never used in desktop configurations. Better alternatives that actually made sense existed, so those were used instead. No brainer really, and it's somewhat strange that it's even being discussed. Maybe in the aftermath of the X1000 and similar, people has lost all track of concepts like "suitability", "viability", "waste of resources", etc? Why else are we regularly having these kind of pointless discussions about Cell and whatnot?

In the spirit of this: What I think is long overdue by now, is a discussion about Power 7! I mean, just imagine if "A-eon and the guys over at Varisys" would create a $10,000+ "AmigaOne X2000" desktop system for OS4 based on Power 7? Oh, wait... :rolleyes:

[Crumb]

What do you think about using a Cell CPU that has 30x FPU performance to that of a Intel PC as the CPU for a computing platform?

I would prefer x86+OpenCL since medium-range gfx cards probably have that power and much more :-)

[freqmax]

OpenCL, as in 0wn3d by Apple for use with closed hardware like Nvidia?

The power of Amiga comes from powerful auxillary chips?

[BlackMonk]

The only non-PS3 consumer application of the Cell that I know of is in some Toshiba laptops like this one:  http://www.xbitlabs.com/articles/mobile/display/toshiba-qosmio-f50_5.html#sect0

Heck, used more than I thought: "Besides Toshiba’s Qosmio F50, G50 and G55 series notebooks, the SpursEngine can be found on standalone PCI Express cards offered by Leadtek and Thompson as well as in the new 42- and 46-inch TV-sets Toshiba Regza."

Not much more info:  http://en.wikipedia.org/wiki/SpursEngine

But just to reiterate what everyone else said, it ain't easy to program fer and it don't like running an OS, really.  It has one generic PPC core and then 7 vector units and you really really have to have stuff that is highly parallelizable to wring the best performance out of it.  It took the PS3 crowd several years to do so for games.

[blakespot]

What do you think about using a Cell CPU that has 30x FPU performance to that of a Intel PC as the CPU for a computing platform?

The Cell is a poor general purpose microprocessor. This is why Apple passed it by. It has notable power when its SPU's are coded into motion for the media-intensive purposes for which it was designed.



bp

[blakespot]

I highly recommend this book for anyone interested in the Cell or the PowerPC in the Xbox 360 --- and what goes into design decisions that make those processors right for game machines (but not for general purpose computers).

http://wraltechwire.com/business/tech_wire/venture/story/4497770/



bp