folding is currently working at petaflop speeds OMFG!!!!!!!! holy @#%%. i've been watching the stats for a while been waiting for this to happen!!!!

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats

i hope the petaflop boundrie for folding gets broken in my lifetime!!!.

Looks like about 75% of the credit (both literally and figuratively) belongs to Sony's PS/3 (hint, hint, hint)...

Linky

I just purchased a quadcore, and am thinking about a PS/3 next.

Would be nice to split between F@H and R@H... (hint, hint, hint)

Very cool for them, though (from what I know of it) I can't say I'm as excited by the science done by folding as I am by the science done by Rosetta.

---

Very cool for them, though (from what I know of it) I can't say I'm as excited by the science done by folding as I am by the science done by Rosetta.

iam not really excited by the science done by folding or rossetta iam just intersted the numbers of flops it can achive. i've paticipated in either of the projects but after i see this goes to 1 petaflop iam joining folding and asking a few of my friends to do so as well i would like to see a exaflop in my life time :)

also i might be thinking of getting a ps3 to push that number even further this is so cool distrubuted computing aching a petaflop you know how much it would have cost to acivhe this if its done by supercomputers

I have a PS3. I bought it for Folding@Home, and it did that for a few weeks. Then I learned it could work just as hard via BOINC at PS3GRID (full-atom molecular dynamics simulations):

http://www.ps3grid.net/PS3GRID/

3500 credits/day, if that helps... =;^)

And the *real* bonus is that, since this is BOINC on linux, it can run *other* BOINC projects too. SIMAP and SETI have apps that will run on the PS3 too, although they do not take advantage of the SPEs. Applications need to be redesigned to do so. However, when projects realize the potential (16x a 2ghz opteron)...

Rosetta, I know you know this. But reminders may help! =;^)

---

Reno, NV
Team: SETI.USA

I didn't realise that there were BOINC projects on PS3... yes, if the PS3 achieves the ubiquity of the PS2, there is a massive 'market' there, although you'd probably want a very easy setup (admittedly things have been improving at BOINC, it's now pretty easy already).

3.5k/day is nice, about the same as a q6600@~3ghz at SETI (on optimized client) I think.

---

Crunching Rosetta as a member of the Whirlpool BOINC Teams

As noted, PS3GRID have a Boinc client, so all that would be needed is to port the Rosetta client to cell linux for PS3 and we'd be a go. Yeah, I know. That's *INCREDIBLY* easy to say.

Doing that in such a way that it uses the SPE's effectively will not be easy, largely due to the limited memory capacity of the SPE's.

To the best of my knowledge, the basic idea used by the PS3 F@H team was to simply subdivide the work and brute force it. I believe that on a PC the main "atom to atom" computation is only done if a first pass determines that it's worth while. On the PS3, they took a rather more brutal approach: just try everything, and discard insignificant results after the fact.

Even doing this, you can see the speedup they got.

Which raises the question for the Rosetta team .....

Does the Rosetta algorithm work on the basis of attempting to compute a force/distance/whatever between every pair of atoms for a given comformation, or at least every pair of atoms where it makes sense?

In the more general case, at a conceptual level at least, what exactly does the "inner loop" do? While I'm at it, is there a "cliff notes" overview of the algorithm, as in something that outlines its operation in five minutes or less of reading?

---

As noted, PS3GRID have a Boinc client, so all that would be needed is to port the Rosetta client to cell linux for PS3 and we'd be a go. Yeah, I know. That's *INCREDIBLY* easy to say.

Doing that in such a way that it uses the SPE's effectively will not be easy, largely due to the limited memory capacity of the SPE's.

A fastest and easiest way is to compile the Rosetta application as native Power/PPC Linux application. The PS3 has IBM Power CPUs, so the Linux PPC/Power binary would work at the PS3. But, the it only use the PPE.

---

[quote]The PS3 has IBM Power CPUs, so the Linux PPC/Power binary would work at the PS3. But, the it only use the PPE.

A little bit of sneaking around discovered this...

"Power Processor Element
The PPE is the Power Architecture based, two-way multithreaded core acting as the controller for the eight SPEs, which handle most of the computational workload. The PPE will work with conventional operating systems due to its similarity to other 64-bit PowerPC processors, while the SPEs are designed for vectorized floating point code execution. The PPE contains a 32 KiB instruction and a 32 KiB data Level 1 cache and a 512 KiB Level 2 cache. Additionally, IBM has included a AltiVec unit which is fully pipelined for double precision floating point and each PPU can complete two double precision operations per clock cycle, which translates to 6.4 GFLOPS at 3.2 GHz; or eight single precision operations per clock cycle, which translates to 25.6 GFLOPS at 3.2 GHz." [courtesy of wikipedia]

Then, if you take the number of t-flops for PS3's in F@H and sivide it by the number of b]active[/b] PS3 CPU's in F@H (where it calculates it's t-flops from), then you get (975/39302 = 0.0248) or 24.8 G-flops. Which is pretty near the theoretical peak for the PPE in the CELL BE, meaning only one processor is being used?

If you take a look at the CELL BE's potential it appears it is far more powerful than this when it's 8 SPE's are used...

"IBM's new planned supercomputer, IBM Roadrunner, will be a hybrid of General Purpose CISC as well as Cell processors. It is reported that this combination will produce the first computer to run at petaflop speeds. It will use an updated version of the Cell processor, manufactured using 65 nm technology and enhanced SPUs that can handle double precision calculations in the 128 bit registers, reaching double precision 100 GFLOPs."

Meaning that F@H's performance could yet go through the roof and i have too much time on my hands :0

---

F@H USES the SPEs. So the theoretical peak for the whole PS3 is around 200 gigaflops. Right now, F@H is getting 35 gigaflops (if you're crunching 24/7, it's 35, not the average of 25); they think they can realistically get 70 gigaflops out of the PS3.

Please don't port stuff to only use the PPE. If you're gonna go PS3 or GPU, then use projects that utilize the whole system.

F@H's GPU beta is getting 90 gigaflops per ATI x1950xt which is still superior than the PS3, but far more user-intensive. You can even do AGP (instead of PCI express) versions of the x1950xt...

---

Please don't port stuff to only use the PPE. If you're gonna go PS3 or GPU, then use projects that utilize the whole system.

Why not ? - It is a quick "first" step to get the PS3 running. Further steps as a SPE version or GPU can follow. If there would more BOINC projects support the PS3, I think more better is it for the whole BOINC framework.
Also, the development for the PS3 to use the SPEs is a lot of work and need a lot of code rewriting...

---

Why not? Because Folding@Home is a legit project and why use 1/8th of a PS3 on a project you like a little more when you can use the full power for Folding@Home? Folding has already done a lot to improve efficiency of their crunching and I'm fairly certain they have shared these ideas/results.

If I could donate 35 gigaflops to Folding or 5-7 gigaflops to Rosetta, I would donate to Folding, no question in my mind.

Crunching on a PS3 in a suboptimal way would not be progress. Just look at Windows 3.1 through Windows XP... It'd be best for the Rosetta project to design/program something from the ground up.

---

Finally broke down.

Was considering a second Q6600 quadcore or a PS/3.

The PS/3 won.

A little cheaper at $499 (vs ~ $650 + tax), and 7 "free" blu-ray movies, no tax, and free shipping.

I'll be away until about 10/11, so it will have to wait until then to go "live", but I'll be throwing it a F@H, for exactly the reasons you state.

Now, if Rosie ever makes it over to the PS/3, we'll talk.....

Why not? Because Folding@Home is a legit project and why use 1/8th of a PS3 on a project you like a little more when you can use the full power for Folding@Home?

Cool. For folks willing to deal with the hassle and maintenance of keeping it running, the GPU folder (x1950pro or x1950xt) is twice as powerful per total dollar spent (including power costs).

---

twice as powerful (per total dollar spent) as cpu or ps/3?

may be i am waiting in vain, but hoping that when dx10 hits its stride, these x1950 dx9's will come down in price. been watching them on newegg and fleabay.

f@h says they want 512 on these babys, not the 256 version.

maybe around black friday?

that's also about when the warranty on my Compaq A64 3800+ expires, and i'll be looking for a cheap a64 x2 to throw in as well.

guess i'll also have to consider the power supply thing if i want to put a x1950 into a Compaq...

Cool. For folks willing to deal with the hassle and maintenance of keeping it running, the GPU folder (x1950pro or x1950xt) is twice as powerful per total dollar spent (including power costs).

t

guess i'll also have to consider the power supply thing if i want to put a x1950 into a Compaq...

Im running a x1950xt with a 420 w power supply, important is, that your power supply has atleast 28 A on 12 V. (specifications should be written on the power supply)

---

http://www.MIAteam.eu

Cool. For folks willing to deal with the hassle and maintenance of keeping it running, the GPU folder (x1950pro or x1950xt) is twice as powerful per total dollar spent (including power costs).

Paydirt
What do you mean by 'deal with the hassle and maintenance of keeping it running, the GPU folder'?

---

Jmarks

Running Folding@Home on PS3 is mostly "fire and forget" like BOINC. Doing PS3GRID on PS3 is much more hassle (you have to install linux, etc).

GPU folding is still in its beta stage and it is difficult to set up on a computer. You have to use specific drivers (you can't always use the latest drivers). There's a limit to how much you overclock, but with x1950xt cards you have to clock your card because the Folding application does not trigger the card to run in standard 3d-gaming clock speeds. So you have to use a secondary program to increase core GPU clock, give it enough voltage at those clocks and run the fan faster than standard (60% flat fan speed is good). It's also good to have a GPU card that is a "2-slot" card that spits the heat directly out of the PC and have an intake vent on your computer case that is directly over the GPU. (Not so simple as buying a PS3, downloading the app, and then running the app) :)

I do think GPU folding is worth it because it has twice the performance of PS3 folding PER dollar spent over 3 years (including 3 years of power costs).

---

With the new Ps3 price drop the cost per flop will need redoing.

'Well folks, it looks like the funky, PS3 price-drop happenings you've been hearing about lately may very well be dropping all over the place. Apparently, hot on the heels of a new filing with the FCC, and various rumors concerning an approaching $399 console, Sony (which has heretofore denied the chatter) has announced a "Big Bang" press conference to take place on October 12th, where many believe they will intro a new, lower priced 40GB console.'
http://www.engadget.com/

---

Jmarks