f13.net

http://www.blachford.info/computer/Cells/Cell5.html





Just in case anyone's interested about the PS3...It's a semi long/boring read, but still dumbed down enough for the layman. Perhaps it's just hype, but I found it interesting because every other explanation of the Cell is either too technical (for me) or too dumbed down.

I'll believe the PS3 kicks ass when I see it. There were plenty of great technical write-ups on the power of the Emotion Engine....but then they hobbled it with a paucity of RAM.

that article seems very optimistic regarding price and availability.  ten cores on one chip is going to make a single chip bigger, i don't care what he says.  this means fewer per wafer, then there's the new fabrication process to worry about.  multiple cores on one chip are already out there, though, and we use a few where i work; they actually are quite a bit cheaper than the previous models.  so maybe he's not full of shit.

but we are talking about Sony.  best to wait and see.

Glad to see you capitalized Sony. How about buying a few caps for those words that start the sentences, too? It just makes your informative post more readable. Good content, poor structure.

Due to popular request, Yegolev is now using capital letters in strategic locations near you.

Sponsored by Sony.

How has Souk gone this long without getting grief titled "English Teacher" or something more smarmy?

:)

I'm the new guy, I'm not going to say anything.

On-topic, I refer the class hyuh: http://www.ibm.com/chips/products/powerpc/processors/

Apologies for the overabundance of information there.  I'm sure no one is interested in this crap other than me.  But hey... this is the internet.

I vote "Schoolmarm".

How does Sony expect those chips to function inside the confined space of a console casing without setting fire?

A magical .0079 micron process?  These are the questions I have.  Our servers sound like a hairdryer concert, is the PS3 going to have heatpipes and 120mm fans?

Refer to PSP technical retractions.  Again, I must click Wait-And-See button.

I think that the author of the analysis is overly optimistic in thinking that Cell will sweep aside the x86, critical to that happening is going to be how well it can run x86 applications for the transition.  Cell is going to require *radically* different software design, so different everyone will be starting from scratch (most programmers have *never* coded in ASM, Cell essentially allows nothing else).  Until languages and compilers catch up, there's going to be a serious shortage of coders.

And the Cell architecture may have insanely high theoretical capability, but when emulating an x86 it's likely to lose all of those advantages and have to essentially operate as if the PU and one APU is all it has.  Since much of the power of the system is the parallel operation of the APU's (and the rest is in the parallel operation of multiple Cells, also something it won't be able to take advantage of when emulating x86 unless the applications were already built for multiple CPU's) the performance of a PS3 emulating a PC is apt to be less than stellar.

And as he pointed out, the x86 architecture has defeated technically superior platforms before, through the power of the installed base and the software library to service it.  Cell's best chances are in pushing x86 out of markets where it's not doing so well at pushing in (consoles, PDA's, embedded devices).  Once it has done that, and your console, HDTV, PDA, and smart refrigerator all have processing power to loan to your Cell-based PC, and Linux/WINE/Open Office to provide the basic functionality is available....  Then the rules could change.

On the other hand, the DRM built into the system may kill its chances there before it even gets started.  PC's are general purpose computers that can be used to violate copyright because they can be used to do *anything*.  A DRM-integrated system that refuses to play nice with Linux burns a whole slough of the potential bridges between the console and the desktop.

Sony may have to decide, once and for all, if they are a technology company or a content generator.

--Dave

Well said.  Emulation with good performance is entirely possible, however it requires a tightly-integrated virtualization environment, assuming that the OS does not handle this itself.  Our machines handle this well by keeping the virtualization environment entirely in RAM.  This means Sony (or someone) will need to develop an VE/OS for their hardware well before displacement of the x86 happens.  The cost of this environment, as well as the necessary hardware that exists outside the processor unit, will drive the cost of such a machine out of the range of the common citizenry.  Besides, you can run a virtual Linux server on this sort of hardware right now.  I refer the class hyuh (http://www.ibm.com/servers/eserver/openpower/press/pr_091304.html).  The thing to take away here is that The Future Is Now, but Joe Jackass can't afford the hardware.  Not sure how Sony is going to make this affordable, other than continuing to let IBM lead this area and feeding off of the scraps.

As long as it does progressive-scan, I'll consider that a leap.

That cat from Metal Gear...you know, the japanese bigwig who's name I should know...was bitching in a similar fashion in the post-mortem to MGS, that it was tough to come to grips with programming for the EE in the PS2.

MGS was the first PS2 game I saw that I actually liked the graphics.

Forgive my ignorance, but wouldn't part of the emulation shell perform some automagical feat to trick the x86 software into thinking it's running on a single cpu, but then (again, using the magic of emulation) distribute that across it's multiple cells? With the emulation layer handling the sychronizing, coordination, and all that stuff I know little about? I know it's obviously a lot more complex than I'm painting it, but at a high level, it makes sense to me.

I don't see this thing conquering the desktop any time soon.  The deal is their impressive numbers seem to depend upon massive parallelism... when the vast majority of applications of personal computers do not require, nay, are not really applicable to parallelism.  Most users get confused when they have more than one application open in Windows, even purportedly intelligent users, people with Ph.D.'s.

Most of end-user level applications is user interface and storage access, and most users can only manipulate one UI at a time.  One-track minds in computers are just fine for the one-track minds in their users.

Where such an approach works well is when you need to apply regular algorithms to large data sets.  They mention SETI@Home on that website... SETI@Home is _perfect_ for parallelism.  It might work out well for producing higher-quality CG images...  It'd be nice if you could get Shrek-quality animation out of a game, at a useful framerate.  But other than graphics processing, the normal processing of user input and game state and whatnot isn't going to benefit from parallelism.

As for programing for it...  There are plenty of people with experience in parallel programming.  And plenty of languages designed to work nicely with it.  Erlang (http://www.erlang.org/), for one example.

I just don't see it replacing desktop systems any time soon.

Alkiera

I'm going to read the article when I have some time but I want to point out that that last sentence sucks.  Seriously, it gives me little faith for the technical validity of the rest of the paper.


You would be surprised then, I know I'm not the only person on this board with a working knowledge of HSPICE and CADENCE.

I skimmed through the article.  To quote the Princess Bride, "I do not think that word means what he thinks it means."

He really mangles some basic architecture concepts so thoroughly that I really have to wonder what makes him think he's qualified to read a patent specification and understand it well enough to be able to decipher it for the masses.

He's probably got the general gist of it close but I would hesitate to quote him on any specifics.

As much as techheads and whoever drool over tech specifications for next gen shit - there's only 2 questions the real gamer has.

Is it backwards compatible natively - without running an emulator of some type.

Are the games fun and do they LOOK next-gen. I.E. The Gamecube (except for Resident Evil 4) still doesn't look as good as the Dreamcast, imo. Nor does it have as many fun games.

Real gamers have unlimited funds? Some of us also want to know how much moolah this shiny will cost. Three questions... I'll start again.

These are REAL gamers. (http://www.wimp.com/nut/url)

(found on the board that must not be named, so I didn't link it.  I forget the name of the show, however)

Reno 911 (http://www.comcentral.com/mp/play.jhtml?reposid=/multimedia/reno911/206_wizard.html)

Cell technology sounds good, though I imagine it'll take awhile for the developers to wrap their heads around it.   I'm just hoping Sony keeps the reverse compatibility that the PS2 had.   That's one of the main selling features, IMHO.

They could put giant chrome tubes out the back and make it look like a motorcycle.   Vroom!

...

Or not.