PS3 - Cell CPU explained?

[MUYA]

get ready for a technical read.

http://www.blachford.info/computer/Cells/Cell1.html

a link from the inquirer.

[Lars]

I just finished reading that "article" and now I really can't wait to see the new PS3 and how games will run on it... The part about running a SETI unit in 5 mins on one Cell and how a PS3 will have 4 kinda did it :eek: :D

The comparison to how one Cell would easily take on 5 hypothetical dual-core opterons running @ 3 Ghz also got my attention.... :p

It will be interesting to see how this PU architecture does in the future, and just how Intel/AMD/Microsoft will respond (or when/if they will have to) :rolleyes:

[Graphicmaniac]

if we add that cell has not to run a so heavy OS like winxp, winxp64 and the heavyest thing ever named Longhorn then i think my upgrade will wait much much time ehehehe

[AthlonXP1800]

if we add that cell has not to run a so heavy OS like winxp, winxp64 and the heavyest thing ever named Longhorn then i think my upgrade will wait much much time ehehehe

Indeed, the Cell processor is PowerPC 970 compatible, I think that G4 so it can use Amiga OS 4 as the OS use small footprint but it wont take full advantage of Cell but will use PowerPC compatible mode, in order to take full advantage of Cell, the answer is to port the whole Amiga OS which will take just 6 months, that probably will be the Amiga's biggest gain ever for a long time when Amiga is stuck with slow Motorola 680x0 series that cant get fast further and this time PC will face the same thing with Intel stuck at 3.8GHz but dual core and AMD's x86-64 wont defeat the Cell CPU. For the Mac OS X and Linux, it will take years to port the whole OS to the Cell.

I really think that will soon be the end of x86 and Windows, it shame to see Intel heavy defend the x86 can get as fast as 20GHz that never gonna happened and developed EPIC based Itanium that will replace both CISC and RISC based but found softwares ran too slow and it was the most expensive failure ever. Also Microsoft must had made the biggest mistake ever by abandoned Windows NT 3.51 RISC platforms and only focused on x86 platform because it much faster than RISC.

This is the same situation Commodore went through in 1990 when they made Amigas and realised that Motorola 680x0 cant get much faster than Intel x86, then they explored all the options and decided that RISC is the only way to forward, they choose HP PA RISC as Amiga's next generation CPU. That never happened as Commodore went bust in 1994 then Escom took over, decidied PowerPC is the direction they went but that never happened as Escom went bust in 1996 and Gateway took over but dont want to see the Amiga made hardware and OS ever again because they see the Amiga are serious threaten to the PC industry and Microsoft. Finally in 2000, a company bought Amiga assets and trademarks from Gateway and then the company became Amiga Inc, they created PowerPC reference platform of Amiga One and they successful ported Amiga OS to the PowerPC in 2002 then hand Amiga OS 4 development to a software house with 10 employees, the beta should be complete early this year.

10 years ago many people asked questions why didnt Amiga OS ported to the x86 as it has largest software/hardware base? Commodore knew that x86 will not survive that long and will stuck just like it did with Motorola 680x0 so 10 years on Intel did. :)

[retsam]

I really think that will soon be the end of x86 and Windows, it shame to see Intel heavy defend the x86 can get as fast as 20GHz that never gonna happened and developed EPIC based Itanium that will replace both CISC and RISC based but found softwares ran too slow and it was the most expensive failure ever. Also Microsoft must had made the biggest mistake ever by abandoned Windows NT 3.51 RISC platforms and only focused on x86 platform because it much faster than RISC. they said the very same thing when alpha came out .... im sorry to say but people dont but chips just for chips sake they buy applications, we have yet to see the "killer" ap for the cell chip. so i wouldnt count your chicken before they hatch

[Elderblaze]

the end of windows hahaha. hahahahahahaha.. hold on let me pick myself up off the floor, hahahahahaha.. yes microsoft is gonna sit on top of their multi billion dollar empire and let the almighty windows die. That's surely gonna happen. Wtf are you thinking d00d

[Sgt_Pitt]

i know its pretty funny, i can just see millions of companies trading in their pcs for PS3s :P

If anything these chips will be intergrated into the next gen pc's, i hope they do, i would like to see the end of the x86.

[Remi]

Promising figures? Sure.

So, let's see... Game developers are so busy that most of them are having trouble finding the time to just get even a simplistic handling of precision in their shaders, and now...
"This may sound like an inflexible system which will be complex to program and it most likely is..."
"The parallel processing system essentially moves a lot of complexity which would normally be handled by hardware and moves it into software."
"Writing software for parallel processing is usually highly difficult ..."That doesn't really sound like a system that'll be able to replace your desktop very soon...

[nutball]

I really think that will soon be the end of x86 and Windows, it shame to see Intel heavy defend the x86 can get as fast as 20GHz that never gonna happened and developed EPIC based Itanium that will replace both CISC and RISC based but found softwares ran too slow and it was the most expensive failure ever.

Please explain (in detail) how this will happen, in particular addressing how the reasons for Windows current position of dominance (hint: look up the following words :"backwards compatibility" and "Microsoft Office") will be overturned by a change to a non-x86 instruction set and a new operating system. Please also comment on why this PS3 thingme will be any different from the ... oooh lets see now at least half-a-dozen times now I've seen the "w00t! look at this great new idea for a processor architecture, and guess what it's got a new ISA too! it's gonna take over the world and topple Intel and Microsoft".

The following terms might be useful in your Googling: TAOS, BeOS, AmigaOS, OS/2, OS X, ARM, SPARC, Alpha, Java, Majic, Transmeta, ... OK, bored now, the rest is homework.

Essay to be handed in by Wednesday 0900GMT.

[Nv40]

*SONY* idea of FUture Computing Brilliant.... they want that in the future all electronic devices.. TV ,Printers ,Computers ,Calculators , video recorders ,DVD players .. Digital cameras ,PDas,and ALL electronic apliances *around the world* to ship with some degree of cell computer technology wich can be a cheap and with the degree they want .remember that SONY brand exist many many years before gaming consoles existed. so think about all the products they have. Sony looks to use all the electronic devices around the world as an intercontinental network of cell processors. you can be watching your TV or listening your radio and at the same time the cell processor can be computing something a real computer request in another place. :D

your are not going to notice the diference.. since a radio is a radio..a TV is A TV not a computer. but without you knowing it ,it will be helping a super computer network , or someone playing a game in another place with their console if your electronic device is conected to the network. :)

the change SOny is looking to do in the computer world.. *if they have sucess* will be as big if not more.. than the INTERNET. because not everyone have internet.. but you will have problems finding someone without a TV or a RADiO or an electronic device in his/her house. maybe they can send and receive the data through satelite.. but that will be more expensive.. or maybe using your house Electric lines .. "as a internet conection" like some COmputer Scientist already have done with sucess.


I have mixed feelings about the Cell processor.. its looks like a great idea , the way to go in the future of computers.. but its too advanced for today technology since depends heavily on networking enviroments. In the SuperWorkstation world.. that is not a problem ,since all supercomputers are networked. but as a personal PC its a diferent story.
another thing is that CELL processors are not "normal processors" they are great for some task ,but for others they aren't . can't remember which scenarios they were. i think that mathematics is one area where is weak.. CEll is vector based computing. for general purpose aplications the PSX3 will behave as a normal Apple G4 .

so if you ask me.. i think SOny is very Risky with the CELL processor for a gaming console. in the superworkstation market they will surely take the crown pretty easy . i think they are targeting too early the gaming market and even the personal Computers market with that tecnology because it sucess depends heavily on the failure of the PC industry (Intel ,AMD and MICrosoft and its millions of aplications available) to produce significative progress in the price/performance area. Right now the PC industry is close future are multiprocessors on a single core in a 64bits enviroment. and SLI or dual SLI graphics hardware is also on the horizon. SO i really wish Good luck to Sony.. they will need it ..

if it werent because Nvidia is designing their graphics chip and the *Software developers tools* , i will be very skeptic about theirs sucess in the future in the Console market. with their too advanced Cell technology. i think Nvidia will save Sony ass..this time. because even if the PSX3 is far more powerfull than the Xbox2 , not many developers will want to touch so complex console ,dificult to program. more when you have ATI in the Xbox2/Gamecube2 with an R500 with plenty of memory system and dual or more processors at hand. with so many Tools to make your work easier. it will be very interesting how it turns.... CELL idea is brilliant ,i have not doubts that is the Future... the very far one.. hehe. but i looks that is too early for it.

[bkswaney]

Very nice read. :)

[msxyz]

About Sony PS3 it's rumored that the Cell engine will perform most 3D of calculations, while the chip that NVidia is designing is nothing more that an advanced rasterizer. This means that all the calculation that on PCs are done by the Vertex Shaders / T&L on the PS3 are made by the cell GPU. Quite a comeback but it might save some die space and cut costs down.

[AthlonXP1800]

Please explain (in detail) how this will happen, in particular addressing how the reasons for Windows current position of dominance (hint: look up the following words :"backwards compatibility" and "Microsoft Office") will be overturned by a change to a non-x86 instruction set and a new operating system. Please also comment on why this PS3 thingme will be any different from the ... oooh lets see now at least half-a-dozen times now I've seen the "w00t! look at this great new idea for a processor architecture, and guess what it's got a new ISA too! it's gonna take over the world and topple Intel and Microsoft".

The following terms might be useful in your Googling: TAOS, BeOS, AmigaOS, OS/2, OS X, ARM, SPARC, Alpha, Java, Majic, Transmeta, ... OK, bored now, the rest is homework.

Essay to be handed in by Wednesday 0900GMT.

You see Windows XP has more than 40 million lines of source code now, by 2006 Longhorn could have more than 55 million lines of source code, that will be 16 years worth of 2500 people worked on all Windows projects combined. Here is no way Microsoft has the resources to port Windows XP - the heaviest operating system to other platform because of the massive source code, it will take Microsoft's 2500 people more than 8 years to port it to Cell or PowerPC processor. So Microsoft have to write from scratch to develop operating system called Windows CE for PDAs that use non x86 based CPUs but it is not backwards compatibility with XP as it wont run XP software and vice versa.

In 1990, a Amiga game programmer guy in UK called Francis Charig had a task from his publisher asked him to ported his best selling Amiga game to Atari ST, he not very happy with it, he like every programmers HATED porting because it took long and waste time to porting from CPU to CPU and operating system to operating system all the times. Then he wrote and developed a program on Amiga OS called TAOS that helped him ported it to Atari ST in very little or no time. After the fall of Commodore, he went to formed a company called Tao-Group based in Reading, UK, he continued improved TAOS over the years that time renamed it as Elate OS, a very tiny operating system that are vast technical superior to Windows CE but unfortunately no companies get it, they got Windows CE instead because of Microsoft's dominance. However lots of companies were interested in Elate OS's Intent technology and licensed it to added the middleware hosted to their own products.

In 1999, Amino Corporation researched OS partner for Amiga and found that Elate OS has the same root as Amiga OS, then after when they are the OS partner, Amino Corp went to bought out Amiga assets from Gateway and 1 year on Amiga Inc (former Amino Corp) launched Amiga DE (Digital Enviroment) for mobile and PDA, Amiga DE is actually rebadged Elate OS under license to write programs for mobiles and PDAs. When Amiga OS 4 is done and then after that Amiga DE and Amiga OS 4 will be merged, much the same way as Microsoft did with Windows 9x/Me and Windows NT/2000 created the foundation of Windows XP.

When both Amiga DE and Amiga OS 4 merged it will create the foundation of next generation Amiga OS 5, it will became cross platform 64 bit binary portable operating system with new features included improved memory proctection, J-Engine (world fastest Java engine), Intent 2, MathEngine, VP (Virtual Processor) etc. With Amiga OS 5 running native or hosted, you will be no longer need to recompile source code or binary code regardless from what operating systems, programming languages and CPUs it used. If the file source was a Windows XP paint program on x86 and run it on the target Amiga OS 5 running on Cell CPU, when the program execute and the VP will translate it on the fly, therefore it no longer necessary for Windows to be port to other platform at the massive size and also compile source codes once, it will running on all platfroms. Amiga OS 5 will be a single very small OS scalable from top (server) to bottom (mobile and PDA) on a very small memory footprint (40Mb to 5Mb).

[Remi]

A few words just to put things in perspective.

Windows today isn't 40 millions lines of source, but a lot more (about 100 millions if I remember well). These lines aren't there just because some company executive likes big numbers, but because they achieve functionalities needed, mostly to run effective business systems.

That's nothing. Yes, you read me well. That's nothing compared to the code running in application programs. The unit count isn't in million but in hundreds of billions.

Should we rearchitecture these billions of lines just to avoid changing a few millions of lines?

The key to the use of a systolic architecture (such as the Cell's) in mainstream computing are automatic tools which at least makes programming the beast simple. Ideally they would able to just take any existing code and make it benefit from this new architecture.

Unfortunately, I know of nobody who knows to do that and I'm not aware of any promising advances on that topic...

Edit: In fact I wrote that without having finished to read your post, as the direction seemed clear. A cpu-to-cpu translation technology, whatever it could be, just won't do it. What is needed is an efficient automatic parallelism extractor, a cpu-to-supercomputer translation technology if you prefer. And that's a very different problem...

[nutball]

What is needed is an efficient automatic parallelism extractor, a cpu-to-supercomputer translation technology if you prefer. And that's a very different problem...

Indeed it is, and it is a very difficult problem. Folks have been working on it for decades, and it still doesn't work effectively in the general case (certainly the automatic parallelisation features I've used on various supercomputers are hit-and-miss at best).

Think about it, if parallel programming was easy we'd all have parallel computers on our desks right now. I'd suggest that it's a fair bit easier to produce multi-processor machines than to ramp the processor speed through the ceiling to achieve the same (theoretical) performance boost.

[jolle]

A few words just to put things in perspective.

Windows today isn't 40 millions lines of source, but a lot more (about 100 millions if I remember well). These lines aren't there just because some company executive likes big numbers, but because they achieve functionalities needed, mostly to run effective business systems.


Im under the impression that a fair share of that code is legacy stuff that could serve the OS well to be cleaned out..
problem is that its quite a task to clean up 100 million lines of code I guess hehe..
In XP64 they tore out the 16bit support atleast from what I hear, but instead there is the WOW bit that prolly replaces it..
I would guess there are some pretty unecessary bits left since windows NT that could be cleaned out for everyones best.. but Im no expert on the subject..

I guess Longhorn is based on XP, so hopefully there has been a bit of "spring cleaning" going on there.. heh.