PS3 Folding@Home Begins with Impressive Numbers

hansamurai writes "As we've previously discussed, the Folding@Home client is now available on the PS3, and already some early results are in. The total number of teraflops generated by PS3s has already exceeded all other OS contributions combined and the entire project is heading towards one petaflop of distributed computing power. Stanford notes that their teraflops calculation is conservatively calculated so the total power could be under-appreciated. With the PS3 European release complete and the Folding client already available to them, the number of users will continue to grow for the time being, let's hope that the project does not run out of work units to pass out. Kotaku has some numbers that are a few hours old since the Stanford server is getting hit pretty hard with the renewed interest in the project."

Comments

Very old numbers

[cxreg]

Gizmodo has more current numbers (which are also a little behind). Currently they're showing 346 TFLOPS for PS3s.

Re:Very old numbers

[OddThinking]

I believe the numbers are being taken from this web site.

Re:Very old numbers

[Seumas]

The really impressive numbers will be how many burned out PS3s there are from running 24x7 and the electricity bills parents find themselves paying because their idiot kids are running folding for the pharmaceutical industry 24x7 (and don't kid yourself that it's anythign else, regardless of it being a Stanford project).

What about global warming?

[stratjakt]

Remember all that energy we aren't supposed to be wasting?

Last I heard, F@H was a feel-good novelty that is doubtful to ever produce any meaningful results.

Re:What about global warming?

[Anonymous Coward]

As opposed to SETI@Home?

Re:What about global warming?

[Sciros]

I predict that the average global temperature will go up by 3000+ degrees by the year 5,000,000,000 if we maintain the current rate of PS3 usage for Folding@Home.

Re:What about global warming?

[drinkypoo]

Last I heard, F@H was a feel-good novelty that is doubtful to ever produce any meaningful results.

Where did you hear that? I don't know any details, but it's easy to find a voice of dissent from your view:

""For the most part, it's not that we're looking for a needle in a haystack, but we're looking for broad properties that require good statistics," said Vijay Pande, associate professor of chemistry at Stanford University. As one of the scientists behind the project, Pande is proud to say that Folding@home has actually provided useful information to the scientific community. SETI@home, however, has yet to discover a single alien transmission."

""These successes are documented in peer review journals. Over 50 papers have resulted from Folding@home," said Pande. He and his students collaborated with developers from Sony Computer Entertainment of America to build a Folding@home client for the PlayStation 3, but that wasn't really Pande's idea."

(In-Depth: Sony, Stanford Experts Talk PS3 Folding@home)

"Now, for the first time, a distributed computing experiment has produced significant results that have been published in a scientific journal. Writing in the advanced online edition of Nature magazine, Stanford University scientists Christopher D. Snow and Vijay S. Pande describe how they with the help of 30,000 personal computers successfully simulated part of the complex folding process that a typical protein molecule undergoes to achieve its unique, three-dimensional shape. Their findings were confirmed in the laboratory of Houbi Nguyen and Martin Gruebele scientists from the University of Illinois at Urbana-Champaign who co-authored the Nature study."

(Folding@home Scientists Report First Distributed Computing Success)

Re:What about global warming?

[cxreg]

I don't pretend to understand all of the science, but I assume the people at Stanford know what they're doing. The results page has some things that sound interesting

Re:What about global warming?

[MindStalker]

I thought that was seti@home I believe folding at home does real protein folding work. This work is being done in bio Research all the time generally from paid for time on a super computer. Of course we have no idea WHAT we are folding. We could be helping create the next superbug for all we know.

Re:What about global warming?

[cxreg]

Of course we have no idea WHAT we are folding. We could be helping create the next superbug for all we know.

Heh. The client does tell you what protein you're looking at, if you care enough to investigate it. My understanding is that the most likely benefit to disease research would be finding how "bad folds" happen, which are responsible for things like Alzheimers.

Re:What about global warming?

[usrusr]

Folding on an idle PC that is used for CPU-easy things like typing /. comments and the like is one thing, the PC would be running anyways. But i don't (yet, likely to change with the increasing media-center role game consoles are supposed to take) see why a PS3 should be running at times when it is not running at full CPU demand (e.g. gaming). There should only be full throttle gaming or power off. Or can it still fold in the background while emulating PS2 games?

An exception is the scenario where somebody would be heating electrically anyways, then it's always a good idea to pipe the energy through some circuits before enjoying the inevitable temperature increase. If you shake up that enthalpy, shake it up in the most funky way you can. But then it's still much better to avoid electrical heating at all...

Re:What about global warming?

[GrievousMistake]

They have a Results page showing some of the problems that Folding@Home have been applied on. Not unbiased, but it seems like they're putting it to use.
As to useful results, it's just a distributed supercomputer. Why would its results be more feel-good and less meaningful than those of any other ~500 TFLOP computer? It's not like researchers can ever get enough processing power. Molecular folding is a processor intensive and parallelizable research problem with real applicability, and I'd rather see people spend power on this than on SETI@home.

Re:What about global warming?

[OddThinking]

I hate to repeat a post, but for additional information:

In case anyone is wondering about what the project has acheived so far, here is the link.

Concerning global warming, the processing statistics imply the PS3 is by far the most efficient. At 380 watts, using the statistics given (which are said to be conservative in the case of the PS3), that puts the PS3 at 63 teraFLOPS/megawatt, or 16.5 kilowatts/teraFLOPS. I'm not really familiar with this, but isn't that fairly good? It's definately better than using PCs. Blue Gene/L, which is supposed to be very efficient, will deliver 240,000 FLOPS/watt, or about 0.24 teraFLOPS/megawatt. My calculations may be off, but that suggests the PS3 is highly efficient and a better use of power than a supercomputer.

I'm sure I'm missing some important considerations, so can someone through a little knowledge at this?

Re:What about global warming?

[CronoCloud]

PS2's have temp sensors of some kind as I found out once. The thing started beeping and the red light started flashing, I was trying to figure out what was wrong, touched the thing noticed it was hot and then saw that the fan was blocked by an object behind it. (I can't remember what it was may have been a quilt).

If the PS2 had it, I bet the PS3 does too.

Re:What about global warming?

[qbwiz]

The numbers in that article can't be right. 240000 flops/watt*1.5 Megawatts = 360 gigaflops, not 360 teraflops. It looks like it should really be 240000000 flops/watt (or 240 teraflops/megawatt), which is relatively consistent with their explanation of 5.6 gigaflops per 12 watts per chip (that number is slightly higher, but it doesn't consider RAM and other components). This is therefore better than the PS3's 65 teraflops/megawatt.

Re:What about global warming?

[ScrewMaster]

True ... but "better" is relative for the supercomputer owner. Anyone with a conventional supercomputer pays for the power himself. In the PS3 Folding@Home project, not only the processing is highly distributed: the electric bill is as well.

Re:What about global warming?

[Curtman]

It's definately better than using PCs. Blue Gene/L, which is supposed to be very efficient

I wonder why there isn't a World Community Grid client for PS3 yet? It's IBM's project and IBM makes the processor for the PS3 don't they? I would think they would be using these numbers to sell more Cell's. Maybe that's what we need to get Team Slashdot back on top. :)

WCG has some very good work that needs done as well.

• FightAIDS@Home
• Genome Comparison
• Help Cure Muscular Dystrophy
• Help Defeat Cancer
• Human Proteome Folding 2

Re:What about global warming?

[drix]

No, doof. Whatever benevolent alien race we finally make contact with should be more than happy to bootstrap us with their limitless free energy generation technology. All part of the master plan.

Curing cancer... pfft. Like that's gonna help anyone.

Re:What about global warming?

[OddThinking]

Also, spreading the heat across a few thousand homes may decrease the power requirements for cooling (then again, it may increase it). A 380 watt heat source will cause the air conditioner to come on a bit more, but not much. In the winter, the 380 watts will nearly be offset by the reduction in heating costs. I have no personal experience, but I would guess the cooling systems for supercomputers run all year long due to the density of the heat sources.

Re:What about global warming?

[wheany]

Dumbass

That's all nice and good but

[Rosco P. Coltrane]

As we've previously discussed, the Folding@Home client is now available on the PS3, and already some early results are in [CC].

When will the SNES version finally be available?

Re:That's all nice and good but

[Anonymous Coward]

Where you been, man? I've been running it for three years now!
(I should complete my first work unit sometime in August.)

impressive

[mastershake_phd]

Impressive, but I wonder if this interest among PS3 owners will drop off. Especially when GTA IV comes out, or they get next months power bill.

The Real Question Now

[Shadow Wrought]

Is if you can write-off your PS3 as a charitable purpose since its spending the bulk of its time volunteering;-)

Re:The Real Question Now

[MindStalker]

If the folding@home organization is classified as a real charity than you can write off the extra power usage, but not its actual cost unless your still paying monthly debt payments on it then you MAY be able to write off a small percent of those payments. Of course I'd talk to my accountant first :)

Re:The Real Question Now

[Duncan3]

No. You can however donate to the Pande group at Stanford if you need a writeoff :)

http://folding.stanford.edu/donate/

.

Re:The Real Question Now

[steveo777]

You would also have to be able to itemize the PS3's time spent folding and gaming and present it as a percentage to the IRS. Same thing with work computers that have WoW installed on them.

Re:The Real Question Now

[MindStalker]

Same thing with work computers that have WoW installed on them... HAHA We're all screwed! :)

I do wonder if an actual business would ever do this (assuming its employees were allowed to play games during breaks.. douptful, but funny idea.)

For 64bit floats, the PS3 is a powerhouse

[maynard]

Cell is very optimized toward one data type for calculation: 64bit floats. If you want to efficiently use the PS3 in a cluster, just be aware that your code must:

a) use primarily 64bit floating point

b) either:
            - fit code and data segments within 256K for each SPU
            - crunch long enough between streamed data blocks such that DMA latency doesn't kill performance

c) have the entire calculation broken down into no more than six parts for streaming (one per SPU)

There are SPU userspace threading models that run cooperatively (similar to the old userspace pthreads, I guess), but the thread manager consumes valuable SPU RAM. Also, SPUs don't support a supervisor bit for memory protection... so... bad things happen when threaded code running on SPU goes tits up.

If you want to calculate 128bit floats, ints, or have lots of branch logic... buy a quad core2duo; cell don't do you any good.

BTW: Anyone here hacking GEANT or BLAST for Cell?

Re:For 64bit floats, the PS3 is a powerhouse

[Wesley Felter]

You mean Cell is optimized for 32-bit float. 64-bit double math is about 7x slower.

Re:For 64bit floats, the PS3 is a powerhouse

[maynard]

No. It's my understanding that Cell uses 64bits for single precision and 128bits for double precision.

Re:For 64bit floats, the PS3 is a powerhouse

[faragon]

PS3 Cells are optimized for single precission operations (32 bits), but future Cells, for other users than PS3s, will support double precission (64 bit) with no performance loss.

Re:For 64bit floats, the PS3 is a powerhouse

[maynard]

Single and double precision is entirely dependent on the hardware configuration. However, as noted in a prior post, It's absolutely true that Cell is currently designed to a subset of the IEEE floating point standard. So... 32bit single precision floats is right; I was wrong.

Re:For 64bit floats, the PS3 is a powerhouse

[faragon]

Yes, I {agree with | like} your the single/double precision definition, however, it is common to accept single as 32 bit ("float" data type), just like "int" is 32 bit (except on weird 16 bit compilers, where "int" were 16 bits and "long" 32 bits). Sometimes data types are, in my opinion, abused, like the Microsoft LLP64, keeping "long" data type for 32 bits instead the classical UNIX use of the "long" data type related to the microprocessor register width (when the microprocessor processor register >= 32 bits), just for making the application porting easier (usually related to lazy/ignorant programmers). What a wonderful world ;-)

Re:For 64bit floats, the PS3 is a powerhouse

[adisakp]

You obviously have no idea what you're talking about since you're wrong on every point you make. I'm a PS3 developer but all of the points I make below are well known, publicly released information.

Cell is very optimized toward one data type for calculation: 64bit floats

Wrong!!! Cell is optimized towards 4-float vectors of 32-bit floats. All the vector math operations on the SPU operate this way and it's capable of doing 8 32-bit operations per cycle (FMAC = 32-bit multiply + 32-bit add X 4 wide). On the other hand, 64-bit operations are scalar and non-pipelined. The take a minimum of 7 cycles for throughput and can have longer latency (13 clock cycles). The maximum 32-bit FP Single Precision (SP) rate is at least 28 times faster than the DP rate.

b) either:
- fit code and data segments within 256K for each SPU
- crunch long enough between streamed data blocks such that DMA latency doesn't kill performance

Wrong!!! Actually, a single code processing step and data should fit in considerably less than 256K. Preferably around 128K. Then you can double-buffer your DMA's for input and output. When you do this the DMA latency doesn't even matter unless your processing occurs faster than the DMA transfers since your DMA's are completely asynchonous to the SPU processing. This simple method of programming can hide nearly all DMA latency -- especially for code that repetitively iterates on multiple data blocks.

c) have the entire calculation broken down into no more than six parts for streaming (one per SPU)

Wrong!!! You can break the calculation into many more parts than six. As a matter of fact you could have 100 calculation parts on on chunk of data and simply swap in new code and work on old data. You can arbitrarily schedule more than a single task per SPU. Sony (and even IBM on non PS3 Cells) have libraries that allow you to share SPUs between many different tasks with only a very small minimal overhead incurred in switching between a task on the SPU.

Also, SPUs don't support a supervisor bit for memory protection

Wrong!!! The SPU's can only directly access their own local memory. All other accesses go through a protected external memory interface (the SPU DMA to main memory) and are controlled by memory protection. It is possibly to virtualize and lock-out SPU's from the rest of the system and run them in "safe mode". If the SPU's could run rampantly and access the entire memory there wouldn't be much point to Sony running a hypervisor to keep you out of their system space on the PS3 linux project and still give you access to the SPU. Also, it wouldn't make much sense not to have memory protection from IBM's point of view to use Cell's as CPUs for clustered supercomputers.

bad things happen when threaded code running on SPU goes tits up

Wrong!!! There is no reason on the CELL hardware why it shouldn't be possible to kill SPU threads / processes and the SPU rescheduled by the OS if necessary. This way a single task can no more take down an SPU than the PPSU. It is possible to even swap out an entire SPU programs pre-emptively on the fly and restore their state. This incurs a much higher swap cost (full SPU threading) than a more simple task manager because you have to save and restore the entire context of the SPU (including 128 16-byte registers and 256K memory region) but your implied limitation of the SPU's is definitely incorrect here.

If you want to calculate 128bit floats, ints, or have lots of branch logic... buy a quad core2duo

So quad-core2duo can do 128-bit floats ? If you're thinking SSE (4 X 32-bit floats) then the SPU's do the same thing. SPU's can run integer code albeit more slowly since most of the integer operations are scalar but still running integer code in parallel on an SPU can sometimes be faster than a Core2Duo - If you align scalar-processed integers to 16-bits for preferred slot l

Re:For 64bit floats, the PS3 is a powerhouse

[maynard]

I was already corrected on the issue of floats. I appear to have been misinformed. However, on many of these issues you appear to be splitting hairs. For example, you appear to argue that an SPU does not lack a supervisor bit - but instead go on a tangent about how SPUs don't have local memory access, but must instead perform a DMA transfer. Well, as it turns out, the SPU does NOT have a supervisor bit for memory protection.

Your other arguments boil down to: we don't program Cell like that. In particular, you're discussing how to double buffer and/or stream many code/data segments on Cell with low performance penalty. I'm sure you're right for your particular task (which I assume is game programming). However, that is not the best choice for coding up a Monte Carlo sim. Particularly since it might actually be possible to cram the algorithm and data segment straight into LS. And regarding your userspace threading argument: when a simulation that just spent CPU days (or weeks) in compute goes tits up, bad things do happen. Heh.

Re:For 64bit floats, the PS3 is a powerhouse

[nuzak]

> Wrong!!! Cell is optimized towards 4-float vectors of 32-bit floats.

> Wrong!!! Actually, a single code processing step and data should fit in considerably less than 256K.

> Wrong!!! You can break the calculation into many more parts than six.


A most informative post, but did anyone else have that SNL parody of the McLaughlin group going through their heads?

Re:For 64bit floats, the PS3 is a powerhouse

[default luser]

Your understanding is wrong.

SPU: 2 execution pipes, each 128-bits wide, for a total of 8 32-bit VECTORIZED (SIMD) operations per-clock.

8 * 3.2 GHz = 25.6 GFLOPS for each SPU. These are the same performance numbers being quoted everywhere for SPU SINGLE-PRECISION. This performance-level would not be possible if your math was accurate to 64-bits.

In fact, 64-bit (double) operations actually cut performance because the SPU has to re-use the single-precision SIMD hardware (it has to sacrifice the SIMD functionality). So the SPU can only output ONE double-precision float from each pipe per-clock, or 6.4 GFLOPS.

Just because the registers are 128-bit wide does not mean the math is accurate to 128-bits. See this article here for more details.

Re:For 64bit floats, the PS3 is a powerhouse

[Funk_dat69]

Maynard,

You are right in that the SPU local store doesnt have any protection. You can write straight off the end and screw up your SPU program. The SPUs can only access their local store directly, though, so they cannot ever trash main memory.

Re:For 64bit floats, the PS3 is a powerhouse

[adisakp]

SPU's do have local memory access. However, local memory access is not the same as unprotected memory access. All accesses to external memory (DMA) can be protected. DMA commands use the same type of translation and protection governed by the page and segment tables of the Power Architecture as the PPU (indeed there is MMU management for each Memory Flow Controller for each SPU). I suggest you read this paper to see the memory management on the SPU: http://www-128.ibm.com/developerworks/power/librar y/pa-celldmas/

The SPU's can run in a completely safe and sand-boxed environment with actual SPU threads and processes. In fact, you can set a secure "vault" mode in which the SPU is completely disconnected from external access and it's possible to ensure that SPU's are not capable of touching anything other than what is in their local memory (and nothing external can touch the SPU's local store in this mode).

If during normal SPU something goes wrong (i.e. access to a protected area), it's possible to kill the curren SPU process and restart the SPU with a new process -- neither that specific SPU nore the Cell processor need to be hung up by some bad SPU code. In this respect, it's possible to build a secure OS that allows SPU access. If something bad happens on an SPU, you can kill the offending process -- just like if something goes bad on a single core on a Core2Duo process, the linux kernel can kill the process. The machine won't go "tits up" -- just your process will get killed. But this is true of any bad program on any "secure" OS.

I don't know what you're claiming as a "supervisor" bit for memory protection. Perhaps you're referring to read/modify/execute (RWX) page settings on the page-table (and cached in the TLB) with virtualized memory access. The SPU's do not have a TLB for local store (although it is mapped into the PPU address space and that TLB). However, they can set up "privileged" areas in local memory (and local register address space - i.e. DMA regs) that can be used to accomplish memory protection for local store and access to non-allowed areas can cause an exception which the OS can then deal with appropriately. Additionally, there is an Local Store Limit Register (LSLR) to mask the available region of local store. Now witin the normally accessible area of local store, the SPU can store instructions and execute them allowing for self-modifying code. Before DEP on windows (and/or if your CPU is more than a year or two old and doesn't support execution page protection), you could make self-modfying code run in any region of writeable memory -- one reason why windows buffer overflows are so easy to exploit.

As far as the "this is how we program SPU's", sure you can program stuff to only run only 6 fixed SPU tasks with no swapping and use 256K data blocks (minus code overhead) so you have DMA latencies. However, you implied by stating those as limitations that was the *ONLY* way to program them which is wrong. All the papers out there from IBM and Sony suggest PREFERRED METHODS which are different than your implied limitation. They advocate cooperatively running multiple small task (dozens or even hundreds) rather than dedicating cores to fixed tasks and to double-buffer data (or even code) to mask DMA latencies. Singular task SPU usage can lead to idle SPU's very easily. Additionally, the double-buffer method can hide DMA latencies which can double the speed of your code if you're roughly equal on memory (DMA) and SPU compute time. We prefer to program SPU's in the manner in which they will be utilized much more fully and run up to 2X faster. I think anyone rewriting code for the SPU (since you have to retarget for the SPU anyway, there's no reason to write your code in a manner that will deliberately underperform).

Re:For 64bit floats, the PS3 is a powerhouse

[maynard]

Yeah. Two points:

DMA has latency, and requires a response from the PPU on an interrupt. I realize the DMA controller can move arbitrary pages mapped from the page table into LS. It's - what - broken up into 16KB blocks. Right? I think that's right. But within LS the SPU doesn't have real memory protection. Arguably, it doesn't need it.

WRT your last paragraph: you are absolutely right. I'm looking at this from the perspective of one who wants to solve a single problem, and who isn't a professional Cell dev. I have a problem I want to solve and am hoping that the PS3 will be a cheap alternative to PCs for my lab.

I just recently bought a PS3 and stuck linux on it. I've got the source to the tools I need. Now I need to figure out the best way to get the thing ported right. And... I'm doing it on my own off hours. Mostly because nobody at work really believes that a toy could be cost effective for scientific compute. I disagree.

BTW: your two posts are by far the most informative I've seen about Cell 'round here. Thank you very much for taking time to correct me. Would you be willing to drop me an email and give advice? My addy is public on my /. homepage. Feel free to send from a throw away address if you need to remain anonymous.

Re:For 64bit floats, the PS3 is a powerhouse

[adisakp]

DMA has latency, and requires a response from the PPU on an interrupt

FWIW, it is possible for DMA to be initiated and controlled completely from the SPU's without any PPU intervention - I'm not sure if this is exposed in PS3 linux but the CELL is certainly capable of completely SPU driven DMA from a hardware perspective. In this case, the only latency is in the actual time to fetch data from main memory to local store. This is the similar to the type of stall you get with a data cache miss on a general purpose CPU while moving data from memory to l2 to l1 cache (which can be hundreds of cycles for a single cache miss). The one advantage with programming the CELL is that properly coded algorithms can double-buffer and effectively hide memory access latency to the point where it's no longer an issue. Granted you can do the same on genreal purpose CPU's with Prefetch NTA instructions.

You are right that the maximum for a single DMA transfer is 16K. However, each SPU can queue up to 16 DMA transfers so effectively, an SPU can initiate loads to its entire addressible memory and then continue without interruption or latency in waiting for one DMA to complete before the next begins.

Another interesting note is that blocking data to fit into the L1 cache (which is only between 8K-to-64K on most current CPUs) is a data optimization that helps both general purpose code and makes the code easy to port to SPU. On our last game, I rewrote the particle system inner loop to operate on much smaller chunks that would fit entirely into L1 cache and this more than doubled the speed of the particle system update loop when running on a XBOX (which is similar to a 733MHz P3).

Re:For 64bit floats, the PS3 is a powerhouse

[CronoCloud]

Mostly because nobody at work really believes that a toy could be cost effective for scientific compute. I disagree.

As well you should. Don't your colleagues know about the NCSA PS2 cluster? They proved that scientific work was feasible on game machines, especially with optimized code.

Re:For 64bit floats, the PS3 is a powerhouse

[faragon]

Circa 1999 I used to develop under Digital UNIX, with 21164@600MHz CPUs, where longs where 64 bit (Digital C compiler, LP64). Not about elitism, but to broke my many years old 64 bit code snippets (!), which still work flawlessly under most UNIXes/Linux, but are broken for Windows x64. Click here for more info.

Interest might not drop off too much

[fistfullast33l]

I don't know about others, but I've turned on the auto-start option so that if the console sits idle it will run this in the foreground. It might take a while to complete a work-unit but at least it's still going. Plus it looks really nice. This way I don't have to conciously start it when I'm going to walk away from the PS3 for a few hours but know I'll play it later. Also, combined with the autodownload feature I can leave the console running to download a few movie trailers (Transformers, A Bee's Story and Shrek 3 hit last night) or game demos and have Folding run as well.

Re:Interest might not drop off too much

[CronoCloud]

Autodownload feature? I'd not heard of that, is it similar to the ability to set the PSP to download audio/video via RSS at a specific time?

Re:Interest might not drop off too much

[fistfullast33l]

Sorry, wrong word. I meant background downloading. I got it confused with the auto-start I guess.

Is this a bad thing?

[ArcadeNut]

let's hope that the project does not run out of work units to pass out.

If they are out of work units, doesn't that mean they are that much closer to their goal? To me, it seems that if they run out of work units, it means the work is being completed quicker then expected. Seems like a good problem to have.

Re:Is this a bad thing?

[Knara]

From an earlier comment about folding@home, it seems like they're not looking for anything specifically, but rather they have hunches about protein properties and need statistical backing to verify the hunches. I'd say that it's likely that the work unit creation is fairly automated and can pump out a large number of packets.

It's not like seti@home where work units represent some quanta of recorded astronomical data (though seti@home does apparently verify its work unit analysis results by having multiple clients analyze the same work units to ensure no one is messing around with the data).

Re:Is this a bad thing?

[still_sick]

They won't run out of work units any time soon.
 
There are a lot of potential individual F@H projects that got rejected in the past because even with the insane amount of computing power available, they would still take too many years to fully complete.
 
The "Worst Case Scenario" in this instance then is that the available TFlops skyrocket, all the current projects get "finished", and the bigger projects become realistic and feasible.

F@H ps3 vs 360

[Mike Bridge]

looking at the statistics for the recently released PS3 folding@home client, i begin to wonder how much better the xbox360 could do, since F@H already has an ATI GPU client, that according to the F@H statistics page, produces about 2.4 times the teraflops per processor as the PS3's CPU-only client. granted, a knowledgeable individual might be able to leverage the XNA dev kit to design a F@H client (if the dev kit even gives enough 'bare metal' access to the GPU), but that would see very limited distribution, i was wondering what could be done to have the xbox360 dashboard team design an integrated distributed processing client (a dedicated client, background client during dashboard/etc, or both) so that those idle CPU cycles (almost 11,000,000 CPU/GPU's worldwide, just 1% of them would contribute approximately 6.47 petaflops, over 10 times the current total teraflops in the project) could help cure cancer. as someone who lost his mother to cancer at a young age, i can't begin to express how significant this could be to help others avoid that same tragedy.

Re:F@H ps3 vs 360

[EGSonikku]

Yea, make it part of the 'console war', who folds@home faster!?! Who does more work units!!??! PS3 Vs. Xbox 360 !

No matter who wins, science wins.

Re:F@H ps3 vs 360

[JordanL]

As the PS3 project lead at Stanford noted, the GPU client is inefficient with its FLOPS. He was saying that for the GPU client a scatter is far more of a hit than a recalculation, so they end up using 2-3x the FLOPS to calculate the same data as a single FLOP might take on a PC or a PS3.

Re:F@H ps3 vs 360

[rustalot42684]

Well with the Wii, you can actually fold the proteins yourself using the innovative new motion-sensing controller!

Re:F@H ps3 vs 360

[nschubach]

Which makes sense since most video cards are not made for precision...right? I mean, how many people are going to be pissed that one pixel on the screen is .0006 pixels more to the right than it should be because the card didn't track enough decimal places in a floating point operation.

Re:F@H ps3 vs 360

[John Betonschaar]

Well, there's a GPU in the PS3 as well, so what you're proposing would be possible on the PS3 as well: thinkof a folding@home client that runs on the Cell *and* the GPU concurrently.

Also, as I understand it, the number of TFLOPS is not the only performance criterium for the folding@home project, and GPU's with the same # of FLOPs compared to different (CPU) architectures yield less useful results.

Re:F@H ps3 vs 360

[smallfries]

That's not what he means by scatter. GPUs perform stream processing, so imagine a stream of data coming into a GPU program and a stream of data coming out. The basic "shape" is something like this:


for(int i=0; i<size; i++)
      out[i] = f( in[i] );


Where f is some arbitrary function. Note that the positions in the array cannot be changed - these corrospond to the xy positions of the target pixels. For a scatter operation (very useful in simulations, sieving etc):


for(int i=0; i<size; i++)
      out[ g(i) ] = f( in[i] );


Now the results of each computation can be directed to where they need to go. Without this operation you either need to recode your algorithm to work inplace, where the performance hit comes, or resort your data afterwards. On a GPU sorting is very expensive so most people recode. This problem should disappear on DX10 hardware because the new geometry shaders can be used to do real scatter operations. At the moment people try and fake this with multiple pixel->vertex>pixel->vertex passes but the point cloud performance sucks on a device optimised for rasterisation.

Results

[OddThinking]

In case anyone is wondering about what the project has acheived so far, here is the link.

Concerning global warming, the processing statistics imply the PS3 is by far the most efficient. At 380 watts (at least this is what I've heard), using the statistics given (which are said to be conservative in the case of the PS3), that puts the PS3 at 63 teraFLOPS/megawatt, or 16.5 kilowatts/teraFLOPS. I'm not really familiar with this, but isn't that fairly good? It's definately better than using PCs. Blue Gene/L, which is supposed to be very efficient, will deliver 240,000 FLOPS/watt, or about 0.24 teraFLOPS/megawatt. My calculations may be off, but that suggests the PS3 is highly efficient and a better use of power than a supercomputer.

I'm sure I'm missing some important considerations, so can someone through a little knowledge at this?

What's so impressive here...

[MarcoAtWork]

... besides the number of PS3 owners that are running this? The PS3 seems to be significantly slower than the GPU client for example

GPU: 41tflop 697cpus
PLAYSTATION®3 346tflop 14138cpus

so basically the GPUs are 2.4x as powerful as the PS3s.

Re:What's so impressive here...

[ucblockhead]

The combination of the numbers and the power. Folding@home just tripled its power. If you think Folding@Home is important, that's important.

If, on the other hand, you're only worried about arguments about what machine is "better", then yeah, you're right.

But not everything in life is about CPU dick-waving contests.

Re:What's so impressive here...

[MarcoAtWork]

I like that folding got a significant speedup, don't misunderstand me, I just don't particularly care for the fanboyism that's happening with people going on on how much more powerful the PS3 is than a PC and yadda yadda yadda. I really appreciate that the PS3 crowd is contributing so much, although I think this might die off a bit when the novelty factor won't be there anymore.

Although the folding folks were smart to put on an animated 3d screensaver with the client which will make it that much more likely that folks will keep this running, I wonder about burn-in of the white-on-dark text line statistics that I've seen on the various videos: I wonder if those could be disabled (or if they orbit around the screen).

Re:What's so impressive here...

[EGSonikku]

After being idle a few minutes the stats go away and it's just the (moving) protein and the (moving) globe in the background.

Conversely, you could just turn off your TV.

Re:What's so impressive here...

[MarcoAtWork]

if I had a ps3 and an lcd tv I'd probably leave it on, it looks quite cool in my opinion... of course people with plasmas and dlp would prefer turning it off instead...

Re:What's so impressive here...

[Phil Wilkins]

This is correct, although to achieve that, the GPU client requires twice as many FLOPs to process the same amount of work.

power bill

[DogDude]

Somehow, I doubt that people buying a $600 game system will care if their power bill goes up $1 (or $10 or $20) a month. Power is one of those things that most people ignore and simply pay unless it's completely out of whack. My commercial power bill fluctuates by sometimes as much as a hundred bucks a month, but even that's not enough to make it worth my time to figure out what might be causing it.

Re:power bill

[EjectButton]

yeah most probably wouldn't notice
according to this
http://www.hardcoreware.net/reviews/review-356-2.h tm
the ps3 uses about 200watts maximum
and if you look at the cost per kwh around the US http://www.eia.doe.gov/neic/brochure/electricity/e lectricity.html
and round up for the sake of argument, so say you run it for 24 hours a day, you never play any games on it, and you are paying $0.10 /kwh, that's $14.60/month

more realistically say you pay $0.10/kwh and only run f@h when you are asleep, so 8 hours a day, less than $5 a month more than you would have paid otherwise.

Re:power bill

[ucblockhead]

380 watts == 9.12 kWh per day. Here in California, I pay about $0.22 per kWh for "above baseline" power, which means it'd cost me almost $2/day, or $60/month to run.

I wonder if that can be written off as a charitable expense.

if you're looking at power efficiency

[MarcoAtWork]

a GPU would be even better, since you could get 2.4x the TFLOPS for half of the watts easily (100-130W for the GPU, 30-40W for the rest of the system idling). Heck, if I was a supercomputer manufacturer I'd just put a laptop-friendly slow CPU on an SLI motherboard with two GPUs and with a solid state read-only flash for booting, removing even more power-related overhead and failure points.

Re:if you're looking at power efficiency

[nschubach]

According to the Folding Team, you divide that in half since the results for the GPU need to be calculated at least twice to verify since the precision of the GPU is much less.

The really cool thing is

[WillAffleckUW]

the PS3 uses a lot less energy to do this, and most of the time people aren't using it anyway.

Re:What's more interesting...

[WorldDominationOrBus]

Sony's Howard Stringer (some bobo at the top) has claimed that current PS3 games only use 20-25% of the PS3's total cell-computing power, so that should leave plenty available to fold, even while gaming.

Re:From much less CPU's too

[CronoCloud]

Shit no, Sony builds them to take a lot of punishment (at least CPU wise) Let's see how long is midgar's (my ps2 Linux kit's) uptime

3:34pm up 11 days, 23:20, 2 users, (I've telnetted in from the windows box, I was running Second Life) load average: 1.00, 1.00, 1.00

Would be longer if it wasn't for the damn power outages, thing runs 24/7 otherwise.

You're absolutely right: It's IEEE floats

[maynard]

I'm wrong. SPUs support a subset of IEEE Floating point standard, which means 32bit single precision. See here:

http://www.research.ibm.com/cell/SPU.html

Single precision floating point computation is geared for throughput of media and 3D graphics objects. In this vein, the decision to support only a subset of IEEE floating point arithmetic and sacrifice full IEEE compliance was driven by the target applications. Thus, multiple rounding modes and IEEE-compliant exceptions are typically unimportant for these workloads, and are not supported. This design decision is based the real time nature of game workloads and other media applications: most often, saturation is mathematically the right solution. Also, occasional small display glitches caused by saturation in a display frame is tolerable. On the other hand, incomplete rendering of a display frame, missing objects or tearing video due to long exception handling is objectionable.

Different work units?

[pavon]

When the GPU client first came out, it was pointed out that it was actually using different work units than the normal PC version and so the numbers weren't directly comaparable. I don't know what the situation is for the PS3, but it may not be using the same work units as either the GPU version or the PC version, and thus not directly comparable to either.

/. team number

[4g1vn]

Slashdot team# 11326 Now go get your PS3 and start crunching numbers.

Personal Results

[Panzergheist]

Well, since I'm currently running this I thought I would chime in. The PS3 is more than likely using different WUs. The average WU size for the PS3 that I have seen so far is between 400k and 500k frames. The usual range for PC WUs that I have seen is between 5k to 20k frames.

My PC, which is only an Athlon 64 3500+ (benchmarked at 7190 in F@H) can crunch through a frame every 1 minute and 7 seconds.
My PS3 is going through a frame every 0.067 seconds.

Frame performance doesn't mean much in F@H because different WUs can have different frame sizes. Nevertheless, the PS3 client is performing a lot of work in a very short period of time. I would be curious to see benchmark results released from Stanford showing the relative performance difference between a high-end Core 2 Duo and the Cell in the PS3.

GPU performance

[Panzergheist]

Also, since I haven't seen anyone mention this yet, the GPU client on the F@H site are all ATI X1900s. The work units performed by GPU clients and Cell clients are of a different type than those performed by general purpose CPUs. Check the F@H FAQs for more information.

that loud noise you hear

[A_Non_Moose]

Gizmodo has more current numbers (which are also a little behind). Currently they're showing 346 TFLOPS for PS3s.

Is hundreds of thousands of PC users shouting "AIMBOT" and "CHEATER".

Seriously tho, I've been shocked at the GPU/PS3 #'s...still doesn't make me want a PS3...ATI/AMD AGP 19XX
Radeon, maybe. (do they exist? I thought they did, but nothing on the 'egg...yet).

Re:that loud noise you hear

[yoyhed]

Nope, they don't exist, and I wouldn't count on ever seeing one. I think the best AGP ATI card is the X1600-something (although if I was going ATI on AGP I'd probably just get an X850 XT PE).

If you want the best possible AGP card, get BFG's Geforce 7800 GS OC (it's overclocked out of the box and has nice cooling to cover it). I got mine at Best Buy for $185 about a month ago (which I wouldn't normally do, but the same thing was $210 on the 'egg). I'm able to run Oblivion smoothly at 1280x1024 with Ultra High Quality.

Re:that loud noise you hear

[dragoneye1589]

The ATI radeon X1950pro comes in AGP. http://www.newegg.com/Product/Product.aspx?Item=N8 2E16814131035

Re:that loud noise you hear

[A_Non_Moose]

The ATI radeon X1950pro comes in AGP

Ah, thanks, I missed that one.

If it did not have that hoover vacuum cleaner/blower attached to it, I'd consider it.

Sticks out way too far for my case, judging by the measurements...double height I could live with, but the
width is a bit too much.

Will someone please downmod my parent post?

[maynard]

Yes. I was absolutely wrong. I've been corrected numerous times, but because that post is improperly modded up it's getting undue attention. Thanks for the detailed correction.

Re:From much less CPU's too

[bunbuntheminilop]

Is that a slim ps2 or a old ps2, and what's the longest you've had it up before?

Re:From much less CPU's too

[gardyloo]

Is that a slim ps2 or a old ps2, and what's the longest you've had it up before? Ooh. Slashdot come-on lines.

Re:Comparision of Ps3 vs. PC flawed.

[Anonymous Coward]

I still run folding at home, I support it, but my Bittorrents, my video tools, my firefox will all take away the precious cycles that Folding is after Even when you're not using your computer? I don't know about you, but when I'm not using my computer, the CPU idle time is %99 or so. F@H on PS3 only runs when nothing else is.

Re:From much less CPU's too

[CronoCloud]

It's a fat PS2 of course, with an HD in it. I've had uptimes of over a month easy. IIRC I hit 100 days once.

Early results?

[NeMon'ess]

So what has all this computing power actually accomplished for science thus far? What new protein folding processes have been discovered that are going to help treat and cure diseases?

Re:Early results?

[NeMon'ess]

Nevermind, here's the answer.

Re:Comparision of Ps3 vs. PC flawed.

[robaal]

Bit Torrent doesn't need a user present to seed, and it tends to make the firewall more active because of the high traffic...

I wonder how a new $500-$600 PC would fair against the PS3 when folding (I'm guesstimating that one could purchase a bare PC with some cheaper A64 X2 and an X1950XT at that price).

Re:Comparision of Ps3 vs. PC flawed.

[dch24]

I kind of disagree with what you said here:

believe me, the processor has stuff that's directly against good programming design for video games, the size of the memory available to each process is a big problem

It is true that an effective memory size of 128Kb per processor (SPU, needs to be split in half for DMA double-buffering) limits what you can do in one chunk of code. But I think adisakp said it best:

All the papers out there from IBM and Sony suggest PREFERRED METHODS which are different than your implied limitation. They advocate cooperatively running multiple small task (dozens or even hundreds) rather than dedicating cores to fixed tasks and to double-buffer data (or even code) to mask DMA latencies.

Now I understand what you're saying about limited memory. I haven't tried to fit data+code in 128Kb since DOS, and it could mean a return to assembly coding for some of what I'm writing. But the Cell isn't going away, and the memory size on the SPUs will increase next year. In the meantime, evaluate the cost/benefit, talk it over with somebody, etc., etc... and then cut some code.

Or don't.

Re:What's more interesting...

[Burpmaster]

If only the thing would multitask...

Re:Comparision of Ps3 vs. PC flawed.

[HappySqurriel]

So the PS3 looks amazing, up to 20 times faster then an average computer?

Well it could be 20 times as powerful as the average computer on folding at home ...

I guess the question is how many PCs on this system are Pentium 2, Pentium 3, Pentium 4, Core Duo, and Core 2 Duo based PCs?

I could be wrong but I would suspect that the bulk of these PCs are University lab computers and people's obsolete home PCs (that is, not a gaming PC) so I wouldn't be surprised if most of the PCs are Pentium 2, Pentium 3 and slow Pentium 4 PCs.

I guess what I'm trying to say is the PS3 could be 20 times as powerful as the average computer on folding at home but is certainly not 20 times as powerful as the average new computer you can buy today ...

Re:What's more interesting...

[KDR_11k]

Cell, maybe. RAM, no. Devs have been complaining about the RAM already, there's no way in hell they're going to hand more of that precious stuff to another program.

Re:Comparision of Ps3 vs. PC flawed.

[kinglink]

The response to adisakp's post hits it on the head. The DMA which is "easy" has latency on it. So you need to avoid DMA as much as possible. Assume you have a 129 Megs of data that constantly interacts with each other, Take a large scale Havok model as a for instance. If you don't know Havok is slow and needs quite a bit of power to do work, throwing in DMA latency every couple frames will just cripple the system.

The cell process is only here because Sony forced it onto game programmers. It's a brilliant piece of crunching hardware as I said but for game programmers it's forcing the programmers to do unnecessary work that no one should need. Keep watching the 360 vs. Ps3 battle, 360 games will keep having minor improvements, (the PS3 has 40 megs of Ram missing still) and more functions for the user. The code for networking that the 360 produces is why almost every 360 game has a online component. On the other hand the Ps3 has the less memory, worse hardware for most games, especially open world styled stuff (the variable information just plays with the memory in such ways that what's easy on the 360 becomes a chore to even correctly load into memory and if some data that needs to be worked on can't fit on one of those 128 meg blocks (let's say you have 10 blocks of 51 megs of memory, you can arrange 8 of them so they will work efficently, but the last 2 will slow the system down, on the 360 it avoids the DMA, while it might be slower, it's easier to program for in this type of problem)

In the end the 360 is the easier hardware to work with and produces similar if not better results so far. I've yet to see a Ps3 game that made me go "wow", and my company has every major game (I just tried motor storm, ok game, nothing impressive though). The bigger problem is moving a game to the PS3 can cause issues, moving a game to the 360 is a much easier transition.

Note: when I say "simple" I don't mean weak, I mean it has tools and is setup in a way that the programmer has the freedom to do what he wants with out having to constantly battle the hardware.

GEEEK FIIIGHT!!

[xtracto]

Wait!

Le'me get the popcorn!

Re:Comparision of Ps3 vs. PC flawed.

[dch24]

Well, it's true that some programmers are having to do extra work. I'll be the first to admit, Microsoft has a decade more experience building world-class development environments, and they've got Sony beat. No problem.

But the Cell is being pushed by IBM, not just Sony. What I hope will happen (see, I'm pretty anti-Microsoft) is that the Havok engine, along with other libraries, will be developed for the Cell. However, havok.com is not likely to release source code. What's missing, in my opinion, is the forum for PS3 game devs to get together and share in the effort of producing really fast Cell code. I'm sure the chip is capable of the same level of performance as the 360.

So, yeah, I agree with you. Microsoft has not only about a year lead in the market, but an easier platform to develop for. I just don't like the company.

Re:Comparision of Ps3 vs. PC flawed.

[kinglink]

I'm not saying Microsoft is the best thing for gaming, but Sony is sure as hell not the best thing. Ibm did invent the Cell, however Sony forced it into the Ps3 even after developers told them it's not the best for gaming. They then have forced blu-ray onto the public as well and then claimed a flawless backwards compatibility and gave a less than such.

Currently, I think both companies are pretty bad, the 360 is the best platform but relying on Microsoft to continue to make good hardware and not screw over the fans is like asking my dog not to eat the treat I left in the middle of the floor, it might work for a couple minutes, but eventually that treat is going to go.

These numbers imply a fallacy

[DavidBeoulve]

According to Folding@Home, GPUs account for the greatest amount of calculations (while CPUs are necessary for others). But the only GPUs supported are flagship tier ATI 1900XT & 1900XTX (presumably 1950XT as well), not Nvidia nor 1800XT's. This makes the PlayStation 3 greatly weighted against the PC, because not only do they have seven Cell processors (GPUs, in Folding@Home's eyes), but most PC's don't have the required videocard to even try to take on the PS3. This makes the PS3 seem to dominate computers, when the reality is that it's not an even playing field.