The Rise and Fall of the PlayStation Supercomputers (theverge.com) 50
"On the 25th anniversary of the original Sony PlayStation, The Verge shares the story of the PlayStation supercomputers," writes Slashdot reader jimminy_cricket. From the report: Dozens of PlayStation 3s sit in a refrigerated shipping container on the University of Massachusetts Dartmouth's campus, sucking up energy and investigating astrophysics. It's a popular stop for tours trying to sell the school to prospective first-year students and their parents, and it's one of the few living legacies of a weird science chapter in PlayStation's history. Those squat boxes, hulking on entertainment systems or dust-covered in the back of a closet, were once coveted by researchers who used the consoles to build supercomputers. With the racks of machines, the scientists were suddenly capable of contemplating the physics of black holes, processing drone footage, or winning cryptography contests. It only lasted a few years before tech moved on, becoming smaller and more efficient. But for that short moment, some of the most powerful computers in the world could be hacked together with code, wire, and gaming consoles. "The game consoles entered the supercomputing scene in 2002 when Sony released a kit called Linux for the PlayStation 2," reports The Verge. Craig Steffen, senior research scientist at the National Center for Supercomputing Applications, and his group hooked up between 60 and 70 PlayStation 2s, wrote some code, and built out a library.
"The PS3 entered the scene in late 2006 with powerful hardware and an easier way to load Linux onto the devices," the report adds. "Researchers would still need to link the systems together, but suddenly, it was possible for them to imagine linking together all of those devices into something that was a game-changer instead of just a proof-of-concept prototype."
"The PS3 entered the scene in late 2006 with powerful hardware and an easier way to load Linux onto the devices," the report adds. "Researchers would still need to link the systems together, but suddenly, it was possible for them to imagine linking together all of those devices into something that was a game-changer instead of just a proof-of-concept prototype."
Re: Saddam Hussein (Score:2)
Re: (Score:1)
In the USA the top nuclear experts only needed a slide rule.
Re: (Score:3)
Re: (Score:1)
That was a load of bullshit and so is this story. No PlayStation was ever used for anything like that.
5 Faked Claims by Console Manufacturers [youtube.com]
Re:Saddam Hussein (Score:4, Insightful)
Mod up.
The end of PlayStation clusters was brought about when Sony pushed out a firmware patch to explicitly disable the Other O/S [wikipedia.org] feature. And it wasn't just Hussein. Iran and the NORKs could get their hands on PlayStations by the container load and lash them together.
Of course, the Evil Empires could simply have disconnected their supercomputers from the Internet and not downloaded the disabling firmware update.
Re: (Score:2)
Of course, the Evil Empires could simply have disconnected their supercomputers from the Internet and not downloaded the disabling firmware update.
And I'm sure they did. Wouldn't help when you needed to replace failed units though. If they cared enough they probably could have worked with Sony to get some privileged access, but by the time it was an issue this brief window where the cell beat typical servers in performance per watt per dollar had closed.
Fuck sony. (Score:5, Informative)
Re: Fuck sony. (Score:1)
Re: (Score:2)
Cyberpunk 2077 will rock on PC
I can't say I blame them (Score:3)
Re: (Score:2)
Re: (Score:2)
I would not be surprised if the console being a loss leader for games may have also played a factor in why they removed the ability to boot a second OS.
Isn't that what the GP just said?
Re: (Score:2)
Re: (Score:2)
What's worse: I think that the fact that they needed to sell the software to recoup costs on the hardware was the main reason that they took Linux off the machines.
Re: (Score:2)
Re: (Score:2)
oops... apparently, that's the comment you replied to. Ignore the preceding remark.
That's what I get for scrolling by comments too quickly.
Re: (Score:2)
Re: (Score:2)
they were heavily subsidizing the cost of the PS3 with software sales and then large organizations took notice and built clusters of the things with no intention of ever buying games. You can't have a loss leader that's all loss.
Too bad they didn't make a console they could sell at a profit, like Nintendo. Or, for that matter, Microsoft. The PS3 hardware was just too special, and it didn't actually translate into better games. Many games were actually better on the 360, because it was so much easier to get the majority of the performance out of the hardware, and they were cross-platform.
Sony learned from that experience, but not until they took away functionality with which the hardware was sold — alienating Linux users, many
LOL (Score:1)
Found the bitter Xbox fanboy.
Cry more bitch.
Re: (Score:1)
Found the bitter Xbox fanboy.
You couldn't find your own asshole with both hands, a flashlight, and an inspection mirror. Microsoft can die in a fire. But Sony can die of ass cancer in a fire.
Re:I can't say I blame them (Score:5, Informative)
they were heavily subsidizing the cost
That's not why they pulled OtherOS support. They were sold at a loss on initial release in 2006. Their costs fell and they were probably breaking even in mid-2008, and making a profit on each sale by 2009.
The security on the PS3 was cracked in January 2010, allowing access to the hypervisor and GPU from OtherOS. Sony didn't want to allow that because it might enable running copied games. The patch to block OtherOS was pushed a few months later, in April 2010.
Re: (Score:2)
> They were sold at a loss on initial release in 2006. Their costs fell and they were probably breaking even in mid-2008,
Sony was selling the PS3 at a loss for FOUR years due to backwards compatibility of having to have the PS2 EE CPU onboard. Hence the need to dump the "fat" PS3 for the "slim" ones.
Re: (Score:2)
Most fats sold couldn't play PS2 games. When the market balked at the price tag PS2 compatibility was ditched in the 40GB model and the 20/60 models were discontinued. By 2008 any PS3 sold with back-compat was new-old stock.
No, the losses came from trying desperately to compete with Microsoft's aggressive pricing on the 360 and the lack of non-ps3 interest in the cell preventing costs coming down. However microsoft's aggressive cost cutting famously caused the RRoD which cost them a lot more in the long run
Re: (Score:1)
What a hilarious load of fanboy revisionist history drivel.
The PS3 outsold worldwide the RRoD plagued Xbox 360 every single year on the market right from the very start.
The Xbox 360's RRoD was present not only before any price cuts, it was causing pre-launch demo units to fail everywhere.
And the PS3 was the top selling console worldwide after the the two and a half year Wii fad died off. The RRoD plagued Xbox 360 came in last place in worldwide sales - even with its tens of millions of duplicate consoles so
Re: (Score:2)
You can't have a loss leader that's all loss.
And that's where Sony should fire the MBAs who came up with that loss leader model and come up with a viable strategy. I was in university last time I owned a playstation. And at $300 for the console and $80/game (Canadian pricing), as a poor student, I bought a mod chip and then "bought" my games for $3 each at the rental store.
When Sony killed PS3 linux, I decided I'd never buy another playstation. So now as an adult with disposable income, I actually pay for my games and all those games are for
Re: (Score:2)
Everyone plays the game now, even Nintendo cut their throat to get better hardware. The days of consoles being sold at a profit relied on chips developed in-house in Sony's case and aggressive cost cutting and overpricing for Nintendo. Sega used off the shelf hardware to save on R&D costs but they got hammered on build costs. Now everyone uses off the shelf hardware and sells for a loss in the hope that game sales cover the shortfall until process shrinks bring the cost per unit down. The Cell was suppo
Re: (Score:2, Informative)
Yeah all seven or eight of you were furious...
Re: (Score:2)
But as long as customers line up to hand Sony their money, the companies will rightfully understand they can do whatever they please.
Re: (Score:2)
By early standards, smartphone = supercomputer (Score:3)
As a former game dev... (Score:4, Interesting)
And one who's written a shitload of PS3 SPU code, I can tell you that they were a pile o' shite.
In every metric except RAM bandwidth the SPUs were worse than the VUs they replaced.
Re: (Score:2)
And yet capable enough and priced in a way that made them very attractive for supercomputer applications. My university had one cluster at the time. It wasn't as performant as the actual supercomputer bought from IBM at our campus but it also had at least one zero less on the price tag.
Wait, what? (Score:3)
Researchers investigating certain classes of problems seek out low cost commodity hardware to do the work? I'm shocked. Shocked!
The PS2 and PS3 had powerful CPUs (well SIMD ALUs) relative to contemporary x86 chips. The fact they were packaged in relatively inexpensive consoles with Ethernet ports and had Linux available made them attractive machines for compute clusters. They certainly were no worse than a cluster of cheap Dell or HP desktops.
They fell out of favor because commodity PCs got much faster and had things like gigabit Ethernet and way more memory. A 2008 era Dell desktop would be as much or less than a PS3, have better computing specs all around save for graphics, and have more robust Linux support. After 2010 the PS3 wasn't at all competitive and GPU computing was a useful thing and made the PS3 vastly outclassed.
So the PlayStations were interesting when they were new and then unattractive when they ceased to be more powerful than the competition. This is so shocking.
Re:Wait, what? (Score:5, Informative)
The PS2 and PS3 had powerful CPUs (well SIMD ALUs) relative to contemporary x86 chips.
Yes, but it required deep wizardry to get the full performance out of the hardware, so only a small handful of titles managed it. And they weren't vastly more powerful, only a bit more powerful. So it turned out to be a dumb way to make a game console. Requiring weird tricks to utilize the hardware is most of what killed the Saturn, while the Playstation 1 was dramatically easier to program. Then Sony ignored that factor of their success with the PS1 when they made the PS2, in essence making the Xbox possible. It lured in developers with the promise of easy development, since it ran a familiar OS and had a straightforward architecture. Then they ignored Microsoft's success with their PC-based console for another whole generation, creating another goofy-assed console that everyone hated programming.
Microsoft's entertainment division might be a smoking hole right now if Sony had just done the obvious thing with the PS3, and made it PC-based then. Granted, Microsoft didn't do that either, but they did use an essentially PC-like architecture for the 360. Microsoft probably would have gone x86 for that console too, except they were still smarting from the ease with which hackers turned the original Xbox into a general-purpose computing platform, and they didn't yet have their lockdown hardware ready for prime time.
Re: (Score:2)
Picking a chipset may be a bit more complicated than "smarted from being hacked".
Nintendo, Sony and Microsoft have all used a PowerPC processor in their lineup.
All 3, IIRC, used IBM as their fab too. Sony use Toshiba for the PS2. ARM was non-existant as a competitor and Intel might have told them to piss off.
Re: (Score:2)
Re: (Score:2)
Re:Wait, what? (Score:4, Informative)
The Emotion Engine and Cell were vastly more powerful than their contemporary x86 counterparts (and PowerPC) in one important area: floating point SIMD. The EE's VPUs and the Cell's SPEs could tear through matrix calculations (or scalar FP calculations performed as vectors). They also had good caching and DMA for those ALUs so the CPU core just sort of set up the data streams and let the vector ALUs go nuts. We didn't see that sort of throughput in x86 until the introduction of AVX.
While both the PS2 and PS3 were hard to program for...that did not seem to affect the popularity of either console. It's hard to imagine Sony selling more PS2s through the console's lifetime. The original XBox was possible because Microsoft was fine throwing tons of money at the project including buying Bungie since Halo became a system-seller. While the PC-like architecture was certainly nice for onboarding developers the top XBox games weren't Windows ports but originally XBox titles. If they saw a PC release it was after the XBox release or they were multi-platform games to begin with.
The Xbox 360 was almost as goofy as a machine as the PS3, the design of the Xenon was definitely influenced by the FP performance of the Emotion Engine. It relied heavily on streaming data through vector ALUs for a lot of graphics tasks. While not quite as bad as the Cell it wasn't exactly a bog standard CPU design. I doubt the XBMC project had anything to do with Microsoft going the Xenon route with the 360. IBM wanted to make high power (processing and electrical) PowerPC chips and Sony and Microsoft were willing buyers.
But back to my statement: researchers with certain classes of problems like cheap hardware to do the work. The PS2 and PS3 were attractive to people needing to do a lot of embarrassingly parallel matrix/scalar FP math with little need for shared memory. The CPUs of the PS2 and PS3 were ideal for this sort of processing and the fact they had Linux and Ethernet available made for relatively easy clustering. The departments building these clusters also got some free advertising from "OMG doing real work with video games" fluff articles. There were even official distributed processing projects supported by Sony on the PS3, there was a Folding@Home client you could set to run when the system was idle.
Both fell out of use because x86/x86-64 CPUs improved and later GPU compute became a thing. The PS2 lost out to cheap PCs with Pentium 4s and Athlons for Beowulf-style compute clusters. The PS3 lost out to cheap PC compute nodes with mid-range GPUs with several times the compute power of a single Cell chip.
Re: (Score:2)
> it required deep wizardry to get the full performance out of the hardware, so only a small handful of titles managed it
I'm not sure how relevant that was to research uses though - I seem to recall reading that the architecture was actually extremely well suited to common research software languages and platforms, which had already long been developed to run well on a variety of complex parallel microprocessor architectures.
Or am I misremembering?
Re:Wait, what? (Score:5, Interesting)
I'm not sure how relevant that was to research uses though
I sincerely doubt it was at all. The Cell SPUs weren't difficult to program at all. Far easier than trying to make shaders do GPGPU tasks- which was the entire point of using the Cell.
The real reason it was "difficult" to program for, is it required a programming model that simply wasn't used in games at the time- parting out task kernels to many cores and combining the work into a frame. These days, that's not so weird, since GPGPU took off.
Gaming Systems (Score:2)
I assembled a PS3 cluster back in the time (Score:3)
I configured a 10-box arrangement at the time (the infamous "beowulf cluster") for the local university, and I can confirm that even with software that did not make use of the specialized cores in the powerpc CPUs, the boxes were cheaper, more reliable, and twice as performant than white-boxes x86 systems available at the time.
(I did maintain a PC-based cluster on the same university, and the headache was enormous)
Example SPU Code (Score:2)
For anyone who wants to see what SPU code looked like, here is a an old article of mine from IBM's DeveloperWorks on the subject:
https://www.ibm.com/developerw... [ibm.com]