Beta
×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

IBM Adds Videogame Console Chips to Mainframes

Zonk posted more than 7 years ago | from the this-story-has-about-everything-i-love-in-it dept.

IBM 103

GoIBMPS3 writes "Soon the powerful 'Cell' microprocessor that fuels Sony's PlayStation 3 console will be available in IBM mainframe computers. The intent is to allow high-performance machines to run complex online games and virtual worlds. 'The integration initially will be accomplished by networking the mainframe with IBM's Cell blades, but eventually the Cells will be plugged more directly into the mainframes via PCI adapter cards, IBM said. It's the latest twist in IBM's years-long effort to keep mainframes not only relevant but also cutting-edge. IBM is touting the partnership as an example of hybrid computing--a trend sweeping the high-performance computing industry as companies augment general-purpose servers with special-purpose chips that to accelerate particular tasks.'"

Sorry! There are no comments related to the filter you selected.

Also in the works: (5, Funny)

TinBromide (921574) | more than 7 years ago | (#18887877)

IBM scrapped a wii-mote enabled server consoles for the management of online worlds.

However the prototype was destroyed in a freak bowling/mountain dew/pizza accident.

Barry Bonds, dead at 42 (-1, Offtopic)

Anonymous Coward | more than 7 years ago | (#18888091)

I just heard some sad news on ESPN talk radio. Barry Bonds, world-famous baseball player, has passed away at the age of 42 from a brain aneurysm. He was close to breaking Hank Aaron's all-time homerun record, which now will sadly live on. Even if you didn't appreciate his endless skills and uplifting attitude towards everyone around him, there's no denying his contributions to modern-day sports. Truly an American icon.

Re:Barry Bonds, dead at 42 (0)

Anonymous Coward | more than 7 years ago | (#18889357)

Why are you saying that? I have found nothing of the sort on any news or sports sites........

Re:Barry Bonds, dead at 42 (1)

suckmysav (763172) | more than 7 years ago | (#18892713)

It is a common trolling tactic used on /.

You must be new here etc etc

Re:Also in the works: (1)

DrEldarion (114072) | more than 7 years ago | (#18889169)

I heard that the project was shut down due to the sudden increase in broken monitors.

Re:Also in the works: (1)

noidentity (188756) | more than 7 years ago | (#18893637)

"IBM scrapped a wii-mote enabled server consoles for the management of online worlds.

However the prototype was destroyed in a freak bowling/mountain dew/pizza accident."


Add my monitor and keyboard suffering a similar fate.

IBM has not been relevent in years (-1, Flamebait)

Anonymous Coward | more than 7 years ago | (#18887915)



IBM has not been relevent in years. RIP already!!

Re:IBM has not been relevent in years (0, Flamebait)

suckmysav (763172) | more than 7 years ago | (#18892743)

Spoken like a true teenage dingbat who thinks the PS/3/Wii/XBox360 (pick your fanboi allegience) is the pinnacle of the worlds computing achievements.

Congratulations on being an ignorant toolbox.

A number of BOFH episodes come to mind (3, Interesting)

arivanov (12034) | more than 7 years ago | (#18887933)

Some venerable BOFH episodes come to mind (though there it was VAX, not a mainframe).

For the 30+ crowd... (1)

creimer (824291) | more than 7 years ago | (#18887939)

It's about time that video game programmers have a career path when they face mandatory retirement at the age of 30. After all, Space War was created on a mainframe.

Worst. Headline. Ever. (-1, Flamebait)

Anonymous Coward | more than 7 years ago | (#18887967)

Come on Slashdot. News for nerds?

A microprocessor from an IBM alliance has been added to an IBM mainframe. Nice idiot spin. There may be something to talk about here, but they are hardly putting a game console chip into a mainframe any more than a Power PC is a game console chip or a Pentium or a 68K is a "game console chip" because someone somewhere used it in a game console.

 

Try reading the synopsis (4, Interesting)

spun (1352) | more than 7 years ago | (#18888247)

That was my first thought too. Then I actually bothered to read the fricken' synopsis and realized that this was IBM's actual spin on it, and they were touting the cell-on-mainframe thing as a server for virtual worlds. It's still kind of a dumb spin, obviously geared for non technical types who may be decision makers on large MMORPG projects. IBM's thinking is probably that they will associate the cell processor with games and therefore feel it would be a good thing to use in running an MMORPG. Which it probably will be, as an MMORPG server is inherently a massively parallel application.

PCI? (3, Interesting)

MSTCrow5429 (642744) | more than 7 years ago | (#18887981)

They're going to plug in Cell CPUs into the mainframe via just plain PCI? Don't they mean PCI-X or PCI-E? I'd think using a Hypertransport bus would be perfect for co-processors, too.

Re:PCI? (3, Informative)

drinkypoo (153816) | more than 7 years ago | (#18888087)

PCI is doubtless the cheapest option available. Many systems have multiple PCI buses, for example I used to have a system with a SiS chipset whose slots were split between two PCI buses, and it had an AGP8x and at least one additional internal PCI bus; it had both 32 and 64 bit PCI devices onboard (a RAID controller was on a 64 bit, 33MHz bus internally.) If the IBM mainframes that use PCI buses have enough of them, then it might be worth it. They might be envisioning this on a "mainframe" cluster, where you could distribute them throughout the network. Or of course, the person who wrote the article could simply be ignorant enough to not differentiate between multiple flavors of PCI. If we look at some other Stephen Shankland articles on C|Net [com.com] we find that they tend to be very simple and formulaic, without any technical detail whatsoever. Just putting those letters "PCI" into the article must have just about killed him. Normally if you see a "word" in an article by him that isn't plain English, it's the name of some company.

Informative parent post. And what about the FAH? (1)

mollog (841386) | more than 7 years ago | (#18889809)

Nice post.

I'd be very curious to see when they create some Folding-at-Home (FAH) clients. The PS3 clients are kicking butt, and I'll bet that hasn't escaped the attention of IBM. Distributed computing is an unappreciated upcoming technology.

In fact, I have to wonder if IBM's work on the WCG isn't part if its effort to develop this sort of technology and to create some high visibility track record.

PCI?-(Pretty Clear and Insightful) (1, Funny)

Anonymous Coward | more than 7 years ago | (#18890311)

"If we look at some other Stephen Shankland articles on C|Net we find that they tend to be very simple and formulaic, without any technical detail whatsoever. "

Ummm. We're talking about a Cnet audience there. If we wanted nonformulific, and complex stories filled with technical details, we'd read slashdot.

Re:PCI? (2, Informative)

Jherek Carnelian (831679) | more than 7 years ago | (#18891575)

PCI is doubtless the cheapest option available. Many systems have multiple PCI buses
One of the key features of the IBM mainframes is high-availability - the machines never go down. That means they have to support the ability to replace failing cards while the system is up. For PCI cards, that almost always means a dedicated PCI bus per PCI slot - so you can power down and reboot individual cards without affecting any other cards (because there are no other cards on the same bus to be affected).

I would be very surprised if the IBM mainframes did not implement each PCI slot with a dedicated bus, and thus also making the full bandwidth of the bus available to each PCI device.

Re:PCI? (1)

Amouth (879122) | more than 7 years ago | (#18893033)

they don't. atleast on the small end of the stuff..

the run 1 bus per back plane and if my memory is right they can freeze a slot for hotswap and the buss will buffer commands to it until it is resumed..

never had a chance to mess with much of it..

Re:PCI? (1)

jdray (645332) | more than 7 years ago | (#18899573)

Ten years ago I was working with Compaq servers that had the capability to shut down power to a single slot. We (I was working at Intel at the time) were developing a dual-port NIC in partnership with them, and hot-swap PCI was the big new tech. Several times I got distracted and pulled cards without powering down the slot (which was done with a little push button on each slot). Pulling the cards was no problem, and so long as the power was off to the slot when you put the card in, you could power it up, reload the driver (NLM; I was testing Netware) and keep running. It was a nice setup.

Re:PCI? (1)

FlyingGuy (989135) | more than 7 years ago | (#18896979)

Dude!

I think just MAYBE IBM knows what its doing hardware-wise. There are a handfull of manufacuters of main frame computers that you can plunk into a room and hook 100,000 users up a single machine and have them all banging the same DB2 or Oracle database without the thing even blinking an eye and IBM is if not THE best at it then they are certainly in the top 2.

Your thinking in PC terms. These guys invented the notion of throughput.

It would be something to see a Z series loaded up with cell processors though.

And I am pretty darn sure it runs Linux.

Flying Cars! (5, Funny)

psydeshow (154300) | more than 7 years ago | (#18887993)

Someday, in the future, a computer the size of a small room will be possible!
And it will be as powerful as today's most advanced videogames...

Makes quite a bit of sense to me (3, Insightful)

jimicus (737525) | more than 7 years ago | (#18888011)

Let's face it, nobody buys a mainframe unless they've already got a very specific use planned for it. It's not like an x86 server where it's cheap enough that you might think "We'll use it for X, but even if we don't, we'll use it for Y".

And the Cell isn't really intended for general-purpose use - it's far more appropriate to use it in a system where the code has been written and designed specifically for it.

What better market than one which is composed almost entirely of people with reasonably specific, defined needs?

Re:Makes quite a bit of sense to me (1)

Volante3192 (953645) | more than 7 years ago | (#18888109)

And the Cell isn't really intended for general-purpose use

Really? I could be wrong about this but I thought they were pushing the Cell as the next great awakening in chips, sure, game consoles and mainframes but they were also talking about cell phones and other portable electronics.

Just what I remember though, I didn't devote much memory to it.

Re:Makes quite a bit of sense to me (4, Funny)

Tim Browse (9263) | more than 7 years ago | (#18890735)

Just what I remember though, I didn't devote much memory to it.

That's ok - the PS3 designers didn't devote much memory to the Cell processors either.

Re:Makes quite a bit of sense to me (1)

jimicus (737525) | more than 7 years ago | (#18891317)

By "general purpose use", I meant "sticking in a PC, running OS_OF_CHOICE and APPLICATION_OF_CHOICE".

Which is exactly what doesn't happen with cellphones and portable electronics. Generally, the manufacturer and network has quite a bit of say over what runs on them so code can be developed specifically for them.

Re:Makes quite a bit of sense to me (1)

twbecker (315312) | more than 7 years ago | (#18889931)

People don't buy mainframes period. I think IBM offers them only for lease at this point. Which works out fine because you really can't run one without a support contract from IBM anyway.

Re:Makes quite a bit of sense to me (1)

cmdrbuzz (681767) | more than 7 years ago | (#18891879)

People don't buy mainframes period. I think IBM offers them only for lease at this point.
Not really, as the Bank I work for currently has ~25 zBoxes and we're adding a few more.
We generally purchase them outright and then pay the frame charge when / if we want to upgrade.

Re:Makes quite a bit of sense to me (1)

donaldm (919619) | more than 7 years ago | (#18897633)

Try working for a Telco or bank. I am not talking a few thousand dollars for blades I am talking 100's of millions of dollars and that is just on mainframes with AIX, HP-UX, Solaris and even Linux operating systems. In the majority of cases these companies actually own their mainframes. Don't get me started on Storage Area Networks and Enterprise Backup Systems because these don't come cheap either.

All business (at least those that want to stay in business) have support contracts be it low (PC's and blades) through to high end (Mainframe) hardware and their respective operating systems.

mainframs are perfect for MMORPGs (2, Insightful)

geekoid (135745) | more than 7 years ago | (#18888027)

You can create new servers on the fly,
Really high performance
Easily scalable,
and virtual worlds will never go down for any reason outside the code.
To try and replicate those efforts on PC servers is a waste of money.

Properly done, those issues that arise in many MMORPGs when a large percentage of their population goes to on area for an event....Blizzard I'm looking at you.

Re:mainframs are perfect for MMORPGs (4, Insightful)

blincoln (592401) | more than 7 years ago | (#18891183)

You can create new servers on the fly,
Really high performance
Easily scalable,
and virtual worlds will never go down for any reason outside the code.

Don't forget these added bonuses:

So expensive you could buy an entire data center's worth of x86 servers instead of leasing a single mainframe for a year.
Requires members of the dwindling cult of mainframe experts to administer.

Re:mainframs are perfect for MMORPGs (1)

Amouth (879122) | more than 7 years ago | (#18893105)

not always.. they could always lease space and time on one from IBM - it isn't that bad..

that and i don't think it is that expensive compared to alot of 1u rack mount servers.. that and nothing can touch the i/o of a mainframe

Re:mainframs are perfect for MMORPGs (1)

Biggerveggies (517226) | more than 7 years ago | (#18891811)

Actually, IBM has been working in this field already. Check this: http://redmonk.com/cote/2006/10/06/redmonk-radio-e pisode-27-games-and-mainframes-hoplon/ [redmonk.com] and IBM's press release: http://www-03.ibm.com/industries/media/doc/content /news/pressrelease/1551338111.html [ibm.com] Essentially, Z Series Mainframes + Linux + MMORPG. Personally, I've always wondered why more MMORPGs are not run on mainframes, considering all the points you illustrated: Scalable, parallel, huge uptime, etc.

Re:mainframs are perfect for MMORPGs (1)

Grimbleton (1034446) | more than 7 years ago | (#18893275)

Don't forget EA and Ultima Online.

For sale on Ebay (5, Funny)

MillionthMonkey (240664) | more than 7 years ago | (#18888045)

I have a JCL script for a character I've been working on for months. He has five hundred hit points, six spells, a sword, a long sword, a battle axe, a print spool, and a suite of DB2 utilities.

Pay for shipping and I'll send you the punch cards.

Re:For sale on Ebay (1)

crunchly (266150) | more than 7 years ago | (#18890425)

Hmmm... I've been working on a character for years and years and all I have to show for it is a gecko corpse that I am afraid to eat.

@

accelerate particular tasks? (1)

tuffy (10202) | more than 7 years ago | (#18888111)

Perhaps the Cell will result in a more optimized Game Grid so the MCP can better manage programs by putting them through multi-core Light Cycle matches.

Networks Ground to a Halt! (4, Funny)

willie_nelsons_pigta (1006979) | more than 7 years ago | (#18888157)

IT productivity took a plunge today on this announcement as millions of IT professionals stopped work dead in their tracks and began day-dreaming of the "after-hours" activities that they could get involved in.

More news at 11.

Re:Networks Ground to a Halt! (2, Interesting)

OriginalArlen (726444) | more than 7 years ago | (#18889061)

Very amusing, but seriously - when will low-end Cell CPU-powered boxes arrive? I'm thinking of the ProLiant / PowerEdge type servers, 1-6U commodity servers of the sort I'm planning to install in the new rack in my bedr^h^h^h E_TOO_MUCH_INFORMATION

(Come to that, what about middling-high end workstations of the sort most of us probably spend most of our day? I've the germ of a nice computer collection here, with a 68000, a Z80, various Pentium "0" and upwards intel boxes and couple of Sparc machines. Not enough architectures!! I won't be happy until I have one example of every compile target in the Perl makefile :)

Re:Networks Ground to a Halt! (2, Funny)

CthulhuDreamer (844223) | more than 7 years ago | (#18889801)

Only one Slashdot would people naturally assume that the term "new rack in my bedroom" involved computer equipment...

Re:Networks Ground to a Halt! (1)

MillionthMonkey (240664) | more than 7 years ago | (#18890529)

> Only one Slashdot would people naturally assume that the term "new rack in my bedroom" involved computer equipment...

Not me. I assumed it involved sadomasochism as soon as I read it.

Re:Networks Ground to a Halt! (1)

shannara256 (262093) | more than 7 years ago | (#18890993)

> Not me. I assumed it involved sadomasochism as soon as I read it.

E_TOO_MUCH_INFORMATION

Re:Networks Ground to a Halt! (1)

OriginalArlen (726444) | more than 7 years ago | (#18898683)

It's a comms rack in fact - 19" wide but not properly deep enough for full-size pizza-boxes; their arses hang out the back. The server's arses, that is. It was TMI because what kind of sad bastard has not just a PC in their bedroom, but servers - lots of servers...

Actually they're mostly powered down most of the time. But I find it hard getting to sleep without the soothing whirr of the fans...

Re:Networks Ground to a Halt! (1)

jdray (645332) | more than 7 years ago | (#18899723)

I can't believe I've made it this far down the thread without someone imagining a Beowulf cluster of these things. Has that little gem of Slashdot culture gone so far out of style?

Cell is not a "Console Chip" (5, Interesting)

ArcherB (796902) | more than 7 years ago | (#18888159)

IBM Adds Videogame Console Chips to Mainframes

The Cell is no more a console chip than the x86 (used in XBox) or the PowerPC (used in the 360). Yes, it is used in a console, but I hate to see such a powerful chip "type-cast" to the console. I'm glad IBM is cutting the Cell loose by actually using it for something other than console gaming. However, I wish they would have used a better example than "Virtual Worlds" for its uses. Something like Medical Imaging, 3D Rendering or even Weather Forecasting would have been so much better towards breaking the Cell from its gaming niche.

Re:Cell is not a "Console Chip" (1)

raftpeople (844215) | more than 7 years ago | (#18890357)

Cell already appears in devices from IBM and from Mercury for medical, defense, etc.

Re:Cell is not a "Console Chip" (1)

Torvaun (1040898) | more than 7 years ago | (#18892401)

Virtual Worlds aren't just for gaming. They're also great for simulations, especially physics and collisions. Military uses this to do training when live-fire might not be possible or would be prohibitively expensive. This would also be pretty awesome for engineering, both civil and hardware.

Yes, I do realize I've used VR gaming and world design for my two examples. So sue me.

Re:Cell is not a "Console Chip" (1)

Nullav (1053766) | more than 7 years ago | (#18896211)

At least according to a Wikipedia entry, IBM kinda' designed the chip for supercomputing. Sony just got a scaled-down version of it (7/8 SPE's actually functioning).
http://en.wikipedia.org/wiki/Cell_processor [wikipedia.org]

As the parent said, the Cell CPU was by no means designed specifically for the PS3.

CELL isn't a video game processor! (5, Insightful)

Theovon (109752) | more than 7 years ago | (#18888209)

IBM isn't the first or last to come up with the idea behind the CELL processor. With the advent of programmable shaders, the GPU industry was headed that way like a freight train. When people started using these highly pipelined, highly parallel SIMD stream processors that we know as GPUs to do supercomputing, people with a clue took notice and decided that it would be sensible to strip out the video components and generalize the compute engine. However, this kind of compute engine only works well for stream processing, requiring some kind of general-purpose CPU to supervise. And this is exactly what the CELL processor is: A PowerPC supervising the operation of an array of stream processors.

So, while the CELL is inspired by GPU design, I think it would be more appropriate to say that CELL is a supercomputing architecture that, being what it is, is also highly suitable for graphics applications. As such, I think what the slashdot article says is silly. What IBM is doing is putting a supercomputing architecture into a mainframe. This isn't weird. It's sensible and a wise move, technically and competitively.

Re:CELL isn't a video game processor! (4, Insightful)

SnowZero (92219) | more than 7 years ago | (#18888727)

That's not a particularly accurate characterization. GPUs are very wide, shared memory multiprocessors, while the Cell is a somewhat wide local memory multiprocessor. That is why the Playstation still needs a GPU; The Cell isn't really the best type of processor for current graphics approaches. The Cell design was started quite some time ago (new CPU architectures are not designed overnight), so I don't think it would be accurate to say the Cell is inspired by modern GPU design. Rather, it would be more accurate to say both were inspired by DSPs, which have been used for signal processing tasks for decades.

Also, I don't think I'd use the word "supercomputing" to describe the distributed signal processing for which GPUs are now being used. GPUs won't be used on "real" supercomputers until they decide to implement proper 64-bit IEEE 754 (floating point) support. In that sense, the Cell is more immediately useful for HPC, and along with IBMs presence in that area, you'll see a supercomputer using Cell chips first.

Re:CELL isn't a video game processor! (0)

Anonymous Coward | more than 7 years ago | (#18890751)

Much supercomputing is done by scientists and most scientists need double precision. Cell works awesome for single precision and isn't anything partiuclarly special for double precision.

Cool... (4, Funny)

R2.0 (532027) | more than 7 years ago | (#18888235)

I always wanted to use a pickup truck to bring my rig to a LAN party.

(Actually, that's all bullshit. Don't play games, never been to a LAN party, don't know where to find one. But that's never stopped posting here before.)

Re:Cool... (0)

Anonymous Coward | more than 7 years ago | (#18888849)

You play Big Mutha Truckers too?

The Cell Isn't a "Game Chip" (4, Informative)

_xeno_ (155264) | more than 7 years ago | (#18888249)

The Cell has never been intended solely to be a "game chip." It was always intended to be useful in large supercomputer type environments.

The Folding At Home client is an example of a large clustered-based application that uses the Cell as a math processor, as is the recent "real-time raytracing" demo. Both are applications of the Cell in a "mainframe" type environment.

So it's not surprising that IBM would be releasing Cell-based machines - that's been the plan all along. It was never intended just to be used in the PS3.

Re:The Cell Isn't a "Game Chip" (0)

Anonymous Coward | more than 7 years ago | (#18888379)

Congratulations, notice that all on your own, or did the five previous comments give it away?

Idiot.

(By the way, if you had bothered to read the article, you would know that they were talking about using the Cell to run MMORPG servers, meaning that it is still a game chip. But I guess you were in too much of a hurry to rehash what's been said by practically every other comment to bother reading the article.)

Re:The Cell Isn't a "Game Chip" (1)

yoprst (944706) | more than 7 years ago | (#18891497)

The Cell has never been intended solely to be a "game chip."
Yea, right. Also, I'd like to use this opportunity to welcome our new single-precision overlords...

Serious uses my arse.... (1)

Dogtanian (588974) | more than 7 years ago | (#18888353)

...IBM might be trying to put a "serious" spin on this, but let's be honest, we all know that it's aimed at Los Alamos scientists who want to play Motorstorm and Resistance: Fall of Man during their tea break.

Re:Serious uses my arse.... (1)

Hoi Polloi (522990) | more than 7 years ago | (#18888737)

Someone rich guy will buy it for their home so he can play online with 5000 FPS. "Dude, I pwned you!"

Fail (0)

Anonymous Coward | more than 7 years ago | (#18888363)

Someone should really make sure that IBM gets the "your product failed" memo every once in a while. Unless they're planning to sell these things to projects like SETI they don't seem to have much use.

Any chance this will drive down the manuf. costs? (2, Interesting)

Rev Jim (AKA Metal F (1004571) | more than 7 years ago | (#18888371)

So now that IBM plans to use the same cell processors found in the ps3 what effect will it have on manufacturing costs or even supply issues? Anyone? It sounds like good news for Sony at least PR wise. ANd if it drives costs down, even better news for them.

Re:Any chance this will drive down the manuf. cost (1)

mindwhip (894744) | more than 7 years ago | (#18888967)

What it *could* mean is IBM are beginning to think that, after seeing the sales figures, using these in just the PS3 won't be the cash cow they had hoped for and are trying to boost sales of the cell processor in any way possible.

PCI Card with a Cell (2, Interesting)

everphilski (877346) | more than 7 years ago | (#18888375)

What I want: a hobbyist card with a Cell chip to do hobbyist things on. There is a company that makes a dual-cell card ... for $8000, 'coming soon'. Anyone know of a cheaper way to get into Cell, besides Playstation?

I would think there would be a healthy market for a 'cell accelerator card', especially in the world I come from (Modeling and Simulation)...

Re:PCI Card with a Cell (1)

seebs (15766) | more than 7 years ago | (#18889347)

The companies that have Cell systems haven't even bothered to return my calls and emails with queries.

So, I got a PS3, and I got a Quad G5 to run the sim on, and it was cheaper and works about as well.

Re:PCI Card with a Cell (1)

CompMD (522020) | more than 7 years ago | (#18889759)

I'm certain the Cell could do wonders for modeling and simulation. I'd love to see the performance of programs like Unigraphics (NX), NASTRAN, or Fluent on a Cell compared with the multi-cpu Xeon boxen I have.

Re:PCI Card with a Cell (4, Interesting)

rockmuelle (575982) | more than 7 years ago | (#18890411)

The best hobbyist platform is definitely the PS3. Linux runs great on it and HDMI->DVI adaptors let you use it with a DVI monitor (we run them at 1920x1600 on Dell 24" LCDs). As long as PS3s are in the $500-600 price range, there's no real incentive for another low-cost Cell platform. Of course, access to the graphics pipeline is limited, but the SPUs are much easier to program and more flexible than GPUs for general purpose computation.

If you do bite the bullet and go with a PS3, we've developed a Python library -- CorePy [corepy.org] -- for programming the SPUs (and PPU) directly. It replaces assembly/intrinsics with Python function calls and provides components for building and optimzing SPU programs. It takes the sting out of using the C-based tools and gives you more flexibility with how you use the SPUs.

-Chris

Re:PCI Card with a Cell (1)

oGMo (379) | more than 7 years ago | (#18891209)

Of course, access to the graphics pipeline is limited

...and it should be noted that on the blade servers or any embedded board you happen upon, you're not going to have a GPU at all. ;)

How much is a full PS3 devkit anyway? Compared to an IBM mainframe, the PS3 route may be still be cheaper...

Re:PCI Card with a Cell (1)

everphilski (877346) | more than 7 years ago | (#18892087)

I'm down with C, my problem with the playstation is the limited graphics capabilities ... visualization of the simulation is as improtant as the simulation itself, if it means bad visualization methods don't let you see the true results ... so I'm personally wary about buying a PS3 for that reason until I can find some good hard facts on the graphics capabilities with that hypervizor in place...

PCI Card with a CUDA (0)

Anonymous Coward | more than 7 years ago | (#18890493)

"I would think there would be a healthy market for a 'cell accelerator card', especially in the world I come from (Modeling and Simulation)..."

Or you could get a DX10 card and use CUDA [nvidia.com]

Re:PCI Card with a Cell (1)

oGMo (379) | more than 7 years ago | (#18891171)

Anyone know of a cheaper way to get into Cell, besides Playstation?

You think someone's going to release a Cell board for under $500 (or even $600)? Heck, embedded dev/"hobbiest" boards for stuff like ColdFire [steroidmicros.com] are $200 and ARM [littlechips.com] are $399+ (most places don't even have quotes on the site), and those are cheap, low-end systems.

Re:PCI Card with a Cell (1)

yoprst (944706) | more than 7 years ago | (#18891647)

That's because of very small volumes. The chips themselves (and they sell well, unlike dev boards) are dirt cheap. Make your own board, solder chips to it and you're set

Re:PCI Card with a Cell (1)

oGMo (379) | more than 7 years ago | (#18893119)

This is irrelevant. The poster was looking for a hobbiest solution for Cell, not a mass-market platform. Yes, you can get a contract for 100k XScale boards for a good rate: this is why people use them. 50 $400 devkits aren't a big expense for such an endeavor, either.

But for a hobbiest developer looking for a cheap way to hack around on the platform, $500-$600 for a Cell system is dirt cheap, even compared to the lowest-end devkits for embedded systems, as shown.

Re:PCI Card with a Cell (1)

yoprst (944706) | more than 7 years ago | (#18894453)

When I say make your own board, solder chips to it, I do mean it. Make it, do not order it. Chips are sold at whatever quantities you wish. If you want to buy one ARM CPU instead of 100, you'll end up paying $11 for an item instead of $9.

Re:PCI Card with a Cell (1)

yoprst (944706) | more than 7 years ago | (#18891691)

There are cheaper ways to inflict programming pain on yourself. What's wrong with old trusty mpi-2 sugared clusters? They'll give you more bang for the buck, and they're almost as painful to program, aren't they?

Mostly for streaming data, not gaming. (2, Informative)

Animats (122034) | more than 7 years ago | (#18888465)

IBM has been talking this up for a while. The idea is to offload some "streaming stuff" onto the Cell processors. The phrase "XML acceleration" has been used [ibm.com] , which probably means the Cell gets the job of taking some DB2 result and pumping it out in XML. It's also useful for SSL encryption and other related streaming-type tasks.

This is a traditional IBM transaction processing approach. The mainframe is surrounded by lesser machines which handle the communications and formatting, extract the transaction which needs access to the data, ships that to the mainframe, gets a result back, and then formats a reply to the requestor. In the green-screen terminal era, that was done by dedicated hardware. In the web era, too much of that work moved onto the mainframe itself.

Think of this usage of the Cell as offloading the front half of Apache to peripheral processors. When your AJAX app makes an XMLHttpRequest, the idea is that the front-end machines get the request, decode it, wait until it's complete, then pass one single transaction to the mainframe. A single reply comes back, is reformatted as XML, and is shipped out to the client. The number of events processed by the mainframe goes way down, and all the protocol work is offloaded to the low-cost Cell machines with tiny memories.

Has nothing to do with gaming, though. They're not putting the PS3's GPU (from NVidia) on mainframes.

Still only 256KB (not MB) per Cell CPU, though. That's too small. Just cramming the whole protocol stack in there will fill most of the memory. I think this thing will really start to fly when IBM gets up to a 2-4MB per Cell CPU. Then you'll be able to fit the front-end processing for a web server in the Cell. Until then, it's a niche product.

256mb global, 256k local (1)

raftpeople (844215) | more than 7 years ago | (#18890569)

Still only 256KB (not MB) per Cell CPU, though

256k is for an SPE. The Cell CPU itself is currently configured for 256mb.

Re:Mostly for streaming data, not gaming. (0)

Anonymous Coward | more than 7 years ago | (#18897581)

IBM already has special purpose XML processors called datapower [ibm.com] .
  • XI50: transforming between disparate message formats, including binary, legacy, and XML,
  • XS40: WS-Security, XML encryption, XML/SOAP firewall filtering, XML digital signatures, XML schema validation, two-way SSL, XML access control, XPath
  • XA35: processing XML, XSD, XPath and XSLT

Re:Mostly for streaming data, not gaming. (1)

Animats (122034) | more than 7 years ago | (#18902361)

The WebSphere DataPower XML Accelerator XA35 is more like a web page template engine front end. The front end faces the HTML clients, and the back end faces some system that delivers results in XML. That's a potential Cell application, because it can be done as a streaming operation, with the template and the transaction results merged in a Cell SPE. Might make sense for catalog applications, where the user does a search, and after the search is complete, a nice-looking catalog page needs to be built.

(The killer app: off-loading the insertion of contextual ads.)

Finally! (4, Funny)

backbyter (896397) | more than 7 years ago | (#18888641)

A machine that *might* be able to handle Aero.

IBM Z9 - Now with BluRay! (1)

TG-Apophis (1006535) | more than 7 years ago | (#18888783)

So when will they start adding BluRay and dual HDMI? :)

Gotta sell them somehow (0)

Anonymous Coward | more than 7 years ago | (#18888881)

I guess someone has to buy up all of those leftover PS3s.

maybe it will come to pass afterall (0, Offtopic)

FudRucker (866063) | more than 7 years ago | (#18888883)

Duke Nukem Foreaver...

Sweeping New Trend (2, Insightful)

vux984 (928602) | more than 7 years ago | (#18889049)

trend sweeping the high-performance computing industry as companies augment general-purpose servers with special-purpose chips to accelerate particular tasks.

As I recall my 286 had a Math Coprocessor.
Years later I bought a hardware MPEG decoder card so that I could watch DVD's without skipping on my old Pentium ii.
And over the last several years I've installed GPU boards to accelerate some particular video rendering tasks.

Its nice to see the idea of special purpose chips for hardware acceleration is finally catching on in high performance computing.

Re:Sweeping New Trend (1)

yoprst (944706) | more than 7 years ago | (#18891819)

Special purpose chips for hardware acceleration were used all along for high performance computing. Once the gate count was good enough to make capable chips the trend was simple - ditch those bastards. Of course, if there's pile of cash burning your pocket you can always find an expensive low FLOPS/$ hardware that either allows you to pack lots of FLOPS into small space, or just plain waste money... The idea is always there, it's just that people are not buying into it.

Not a Console Chip (2, Informative)

AikonMGB (1013995) | more than 7 years ago | (#18889131)

The title is misleading. The Cell processor is not a "console" chip, it is a microprocessor. Period. So what if Sony decided to use the Cell processor in the PS3? They could have selected from any number of processors: AMD64, x86, PPC, Motorola 6800.. whatever!

The Cell processor is and always has been designed for shipping out complex calculations to sub-processing units (I believe their latest term is Synergistic Processing Units [SPUs]?), it was not designed for purpose of Sony bragging about it.

Aikon-

Needs a different headline,... (0)

Anonymous Coward | more than 7 years ago | (#18889397)

How about, "Cell Processor Used Outside PS3"?

Am I the only one who remembers the talk about how the Cell was going to create a paradigm shift in general computing, and we'd be seeing them in all sorts of devices? Now they put it in an industry product and its news...

Isn't the processor (2, Insightful)

gelfling (6534) | more than 7 years ago | (#18889883)

Faster than the PCI bus? Seems like they need another way to plug it into the mainframe.

this is irrelevant. (1)

DragonTHC (208439) | more than 7 years ago | (#18890149)

mainframes are no longer necessary.

the only reason people still use the AS/400 or any other IBM mainframe is because it was too expensive to justify scrapping for something modern.

the AS/400 or iSeries as it's been renamed, could be replaced in a heartbeat by a LAMP server with an AJAX frontend for 1% of the cost.

you can run LAMP on an iSeries now if I'm not mistaken.

Re:this is irrelevant. (4, Informative)

raftpeople (844215) | more than 7 years ago | (#18890865)

the only reason people still use the AS/400 or any other IBM mainframe

An AS/400 is a midrange, not a mainframe. Despite having a large span of scalability, the AS/400 only overlaps the bottom end of the mainframe in performance.

Also, the reasons people buy AS/400's and mainframes are as follows:
Extremely high reliability and security
Performance and scalability
Protection of software investment

justify scrapping for something modern

Do you realize that the AS400 hardware and operating system is more "modern" than Unix? Did you realize that the as400 hardware and operating system have key features that other OS's lack today but most people are moving that direction? Do some research on security and the as400, for example.

the AS/400 or iSeries as it's been renamed, could be replaced in a heartbeat by a LAMP server with an AJAX frontend for 1% of the cost.

Have you ever seen an AS/400 that required even an operator? High end, sure, but small to medium business the controller puts in the backup tape and that's about it. Hardware/software, you're right, but unless you include support salaries then you are comparing apples to oranges, although you could make the exact same argument about Sun/Oracle being more expensive than LAMP.

Re:this is irrelevant. (0)

Anonymous Coward | more than 7 years ago | (#18897867)

>>mainframes are no longer necessary.

Hey hotshot! You'd better tell all the worlds major banks, insurance companies, utilities etc.

All of them store and process over 80% of their data on mainframes, not on a friking, LAMP stack!

Kids nowadays!

Where is? (3, Funny)

SnarfQuest (469614) | more than 7 years ago | (#18890527)

Where is the "imagine a beowulf cluster os these" comment?

Oh, here it is.

Marketing Cell Games (2, Interesting)

Doc Ruby (173196) | more than 7 years ago | (#18890921)

The intent is to allow high-performance machines to run complex online games and virtual worlds.

The intent is to make the high-end Cells (with all 8 SPEs working) cheap by selling millions of PS3s with lower-grade (6 SPEs) Cells, a scale economy that big machines ("mainframes") couldn't achieve on their own.

They're not going to be running "games" like VirtuaFighter on mainframes, especially not without the 9x as fast RSX video chip the PS3 includes. But they will be allowing us to run supercomputer-fast Monte Carlo simulations on PS3 under Linux.

So I guess if their marketdroids keep lying to us about making IBM mainframes into game consoles, it's worth it if they keep delivering the reverse, which is much more interesting.

I think PS3 Cells have 7 working SPEs? (0)

Anonymous Coward | more than 7 years ago | (#18891281)

I think the Cell in the PS3 has 7 SPEs enabled (one is disabled to increase manufacturing yields, similar to what often happens with GPU pipelines).

Of course, the PS3 operating system permanently reserves one of the SPUs to itself. (The Xbox360 does a similar thing, reserving half the CPU time of one of the 6 PPC instruction pipelines it has to the OS).

That's what the world needs, a $1,000,00 console (2, Funny)

Anonymous Coward | more than 7 years ago | (#18891039)

I thought the PS3 was over priced at $600.

Misleading headline (1)

asninn (1071320) | more than 7 years ago | (#18891261)

You got it all mixed up, folks. If anything, the headline should've been "IBM Adds Mainframe Chips to Videogame Consoles" when the PS3's architecture was first announced - it was always clear that the Cell would be used for more than just a console.

Brilliant, gaming is important! (1)

m3talocasnica (247894) | more than 7 years ago | (#18891273)

Yeah, babe, IBM pwnz!

OS/400 on cell? (1)

charlieo88 (658362) | more than 7 years ago | (#18891297)

What I really want to know is when am I going to see OS/400 on a PS3? Remember the OS/400 on a PS2 [slashdot.org] post?

way to go, IBM (0, Troll)

yoprst (944706) | more than 7 years ago | (#18891403)

Yet another increase in number of suckers that IBM managed to fool into buying its clumsy CPU...

At last! (2, Funny)

OldManAndTheC++ (723450) | more than 7 years ago | (#18892167)

I can write an RPG [wikipedia.org] in RPG [wikipedia.org] !

"GameFrame" (1)

almondjoy (162478) | more than 7 years ago | (#18892693)

Ok - now I've said it... start using it!

Speed up the slow mainframe please.... (1, Informative)

richman555 (675100) | more than 7 years ago | (#18894245)

The only reason the cell processor is on the mainframe is to accelerate certain things which perform poorly on the mainframe such as Java or web serving. Anyone who has worked on the IBM mainframe with Java will know this. Java works there, it just doesn't perform well at all. With this in mind, I don't see this used much for gaming. Hopefully, the cell is used for virtual mainframe Linux instances which seems to make more sense.

Cell and mainframe, truly weird combination (1)

Tom Womack (8005) | more than 7 years ago | (#18897757)

This isn't where I would have expected IBM to put Cells; from the first announcements several years ago of Cells and of the Blue Gene architecture, everyone's asked 'when do we get a Blue Gene made of Cells?'

That may be something that has to wait for the 65nm 'Cell 2' which IBM described at Cool Chips X ... I wasn't there, I've only got the one-paragraph description from the program, but the major features are that double-precision processing is now pipelined so you get 100GFLOP/chip (two flops per fmul instruction * 3.2GHz * 8 SPEs * 2 doubles per vector-register) and that the memory system is now four-channel DDR2 of up to 16GB, rather than RDRAM of up to 2GB.

[but this chip uses 100 watts, and a 16GB DDR2 memory system would be 64 chips, so it wouldn't fit in the racking or remotely in the cooling of current bluegene]

Just out of curiosity, where is the forum for interesting things done using Linux-on-PS3? I am expecting truly wonderful demos to crop up at scene.org in the medium-term future.
Load More Comments
Slashdot Login

Need an Account?

Forgot your password?