Hello,

Many of you have been talking about Ps3 crunching in Rosetta. Have a look at this: http://folding.stanford.edu/abeta_2spu_test02.avi

As you can see it's possible and it's already working. I hope in the future rosetta may run in PS3 to.

---

Hello,

Many of you have been talking about Ps3 crunching in Rosetta. Have a look at this: http://folding.stanford.edu/abeta_2spu_test02.avi

As you can see it's possible and it's already working. I hope in the future rosetta may run in PS3 to.

We have already talked about this :-) and we know it's possible :-).

Going to take BOINC (and Rom) to set the groundwork first though.

---

Team mauisun.org

Hello,

Many of you have been talking about Ps3 crunching in Rosetta. Have a look at this: http://folding.stanford.edu/abeta_2spu_test02.avi

As you can see it's possible and it's already working. I hope in the future rosetta may run in PS3 to.

We have already talked about this :-) and we know it's possible :-).

Going to take BOINC (and Rom) to set the groundwork first though.

I don't think BOINC will be the major problem. It's a) open source, so it's easy to compile to any OS and b) it's only middleware, no real work done, so no need for any optimisation to the fancy possibilities of the cell professor. The hard part is to get the real stuff working, the scientific application, and to let it use those new possibilities.

Looks like F@H will get a real boost from the PS3!

From my limited understanding, I get the feeling it might be easier to make use of the processing capabilities of the xbox 360 than the PS3 due to the Cell's very small cache.

That'd bring about a whole side to the console war!

Anyone like to take a guess to correct/verify?

---

Just from a price:performance perspective...

I understand that the manufacturer's sell the gaming consoles at a loss, and hope to profit on the sale of the game cartridges.

Given this, and that an individual with $500 to spend desires only to crunch (no gaming), how is this $500 best spent?

Purchasing a subsidized PS/3?
Building a cruncher with a (unsubsidized) Cell processor?
Building a cruncher with a AMD/Intel processor?

Looks like F@H will get a real boost from the PS3!

From my limited understanding, I get the feeling it might be easier to make use of the processing capabilities of the xbox 360 than the PS3 due to the Cell's very small cache.

That'd bring about a whole side to the console war!

Anyone like to take a guess to correct/verify?

Just from a price:performance perspective...

I understand that the manufacturer's sell the gaming consoles at a loss, and hope to profit on the sale of the game cartridges.

Given this, and that an individual with $500 to spend desires only to crunch (no gaming), how is this $500 best spent?

Purchasing a subsidized PS/3?
Building a cruncher with a (unsubsidized) Cell processor?
Building a cruncher with a AMD/Intel processor?

Looks like F@H will get a real boost from the PS3!

From my limited understanding, I get the feeling it might be easier to make use of the processing capabilities of the xbox 360 than the PS3 due to the Cell's very small cache.

That'd bring about a whole side to the console war!

Anyone like to take a guess to correct/verify?

I would presume the PS3 would be the way to go for Folding@home, unless Sony put some caveat on the crunching ability such as it must run a game for a couple of hours a week...

I don't think you can buy a cell co-processor off the shelf - I know there's a card you can get with a number of Cells on there but I think that's a pretty hefty price and I don't know if Folding would make use of it without alteration to the code...

For Rosetta, I hope Billy G comes through - I only need a small excuse to buy a 360 and that would be it!

---

Given that the 100Gigflops rating of the optimized F@H client is somewhere between 20 and 100 times faster than a typical AMD/Intel system, the $500 is definately better spent on the PS3 vs spending it on an AMD/Intel system. At least for projects with an optimized PS3 client.

It'll be interesting to see what kind of speed increase the cell processor add in boards bring to an AMD/Intel system - or if we'll have to wait for the new Opteron boards with options for plugging in a non Opteron cpu such as the Cell. AMD just announced that it is licensing the socket so there will be more cpu choices for the newer Opteron boards.

---

F@H is also working to implement GPU folding though, at which point a top of the line ATI card (I believe it only works with them right now, it's still in alpha) may perform even faster than the PS3.

---

And, cheaper too I would assume!

All this begs the question, cost aside, for Rosetta what is the "best" processor architecture that is currently available? Available near term? Available long term?

Does someone buid a new system? Buy a PS3/xBox? Buy a high-end graphics card? Buy a specialized math coprocessor? A physics coprocessor?

These are exciting times technology-wise, just trying to figure out what it all means for Rosetta...

F@H is also working to implement GPU folding though, at which point a top of the line ATI card (I believe it only works with them right now, it's still in alpha) may perform even faster than the PS3.

All this begs the question, cost aside, for Rosetta what is the "best" processor architecture that is currently available? Available near term? Available long term?

Does someone buid a new system? Buy a PS3/xBox? Buy a high-end graphics card? Buy a specialized math coprocessor? A physics coprocessor?

These are exciting times technology-wise, just trying to figure out what it all means for Rosetta...

For Rosetta, we can do an accurate comparison of the different computer setups we have available if we can get an xml containing the WU Name, No Decoys, and Computer ID as I posted here (and here!).

I think it'd be useful to know which CPU gives most bang per buck, what difference RAM makes etc.

Devs - can we get an xml with this info in pls?

Where an xbox or ATI GPU would fit into the listing we won't know until BOINC gets ported to the consoles and then maybe someone will manage to port Rosetta too. Or maybe Billy G will come through put Rosetta on the xbox360 - that'd make the console war more interesting!

cheers
Danny

---

AMD outlined its Torrenza initiative to us before and this week anounced some mates that have signed up to toy with developing co-processors for the Opteron super socket.

Sun, Cray, Fujitsu Siemens, HP, Dell and IBM said they thought collaboration with the Torrenza Initiative interesting enough to say so publicly. It remains to be seen if anyone comes up with anything but an ATI-sponsored Phyiscs co-processor won't be long in the making, we suggest.

From The Inquirer

As a result, customers could start seeing some odd but fantastic motherboards with, for example, a mix of Opteron and Cell chips or even an UltraSPARC T1/Opteron combo.

From The Register

And in perhaps 5 years.. if the PS3 and AMD boards that allow Cell cpus to plug in alongside an Opteron, we may have options like this:

Quad-core? Pah! Intel has produced an 80-core chip, the world's first programmable microprocessor with teraflop performance capabilities, the chip giant claimed today. It's not compatible with the x86 instruction set - it's a proof of concept part designed to show how a production processor might operate.

Intel's demo 80 core chip

Is it possible to track down an IBM rep and a Sony PS3 rep to see if between the two they'd be willing to donate a coder to port the Rosetta code to run on the Cell processor and specifically on the PS3? I once tried contacting the 64 bit coder team at AMD for another project - and the bums in charge were always off on vacation or out of the office when I called. After two months, I gave up. :)

---

Is it possible to track down an IBM rep and a Sony PS3 rep to see if between the two they'd be willing to donate a coder to port the Rosetta code to run on the Cell processor and specifically on the PS3? I once tried contacting the 64 bit coder team at AMD for another project - and the bums in charge were always off on vacation or out of the office when I called. After two months, I gave up. :)

MS and the xbox 360 might be a better bet - Bill Gates has donated to the lab so that must hold some sway, and I'm no expert but I've read a fair bit about the massive throughput on very limited datasets from the cell. I would expect the xbox CPU (based on similar IBM broadband tech I believe) is a better target with its 3x 3.2GHz CPUs with a decent amount of cache each(?)

Either way, I'd buy the console that ran it, so I'm fairly sure they'd directly make more money from this than they'd loose, plus all the advertising it'd generate...

---

I see on Yahoo! Technology News that about an hour or so ago, Intel announced the planned November release of its Quad-Cores, and:

"Intel officials already have indicated that chips with dozens of cores might be possible by the end of this decade. The company hinted that, 10 years down the line, chips with hundreds of cores might be possible."

Intel's Quad-Core Chips Coming Soon

Quad-core? Pah! Intel has produced an 80-core chip, the world's first programmable microprocessor with teraflop performance capabilities, the chip giant claimed today. It's not compatible with the x86 instruction set - it's a proof of concept part designed to show how a production processor might operate.
Intel's demo 80 core chip

I see on Yahoo! Technology News that about an hour or so ago, Intel announced the planned November release of its Quad-Cores.

Intel's Quad-Core Chips Coming Soon

Quad-core? Pah! Intel has produced an 80-core chip, the world's first programmable microprocessor with teraflop performance capabilities, the chip giant claimed today. It's not compatible with the x86 instruction set - it's a proof of concept part designed to show how a production processor might operate.
Intel's demo 80 core chip

njkid was at XS was running Rosetta on one of these Kentsfields a month or so ago - they're quick, but two xeons (or a mac pro?) will probably be cheaper atm...

---

Erm..
Now I see that the ATI GPU client for F@H is a subset of the Gromacs core; and while it does what it does at 20-40 times faster than the normal client - it's got limitations. And it's not getting 20-40 times the credit than a normal cpu client gets. (440 ppd for default clocked cards)

The same will probably hold true for the PS3 client. It's no doubt a subset of the Gromacs core - but it can do what it does at 100 gigaflops. And it won't be given 100 times the default computers' ppd...

If Boinc and Rosetta are ported over to a future GPU client or to a client that runs on the PS3, we may run into the same issues. It may be an incredibly fast client that only does part of the work our current Rosetta clients are doing.

---

Is it possible to track down an IBM rep and a Sony PS3 rep to see if between the two they'd be willing to donate a coder to port the Rosetta code to run on the Cell processor and specifically on the PS3? I once tried contacting the 64 bit coder team at AMD for another project - and the bums in charge were always off on vacation or out of the office when I called. After two months, I gave up. :)

MS and the xbox 360 might be a better bet - Bill Gates has donated to the lab so that must hold some sway, and I'm no expert but I've read a fair bit about the massive throughput on very limited datasets from the cell. I would expect the xbox CPU (based on similar IBM broadband tech I believe) is a better target with its 3x 3.2GHz CPUs with a decent amount of cache each(?)

Either way, I'd buy the console that ran it, so I'm fairly sure they'd directly make more money from this than they'd loose, plus all the advertising it'd generate...

we are definitely pursuing this angle!! in fact we are having ongoing discussions with local GPU experts on which aspects of the calculations could be most sped up using GPUs. we are also discussing ways of allowing users to actually guide the searches via an interactive video game like interface, but this is still pie in the sky at this point.

---

From the folding at home website:

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

167 X1900 GPU's produce 12 TFLOPS. That's 72 GFLOPS each. Pretty impressive.

---

list of my results

At an ATI demonstration of Stream Computing, there was this quote:

Peakstream is a company that makes a middleware layer and tools to facilitate Stream Computing. Michael Mullaney, its VP of marketing, showed off how its tools would allow you to use GPUs for everything from stock trading to oil and gas exploration. The biggest advance it has is a profiler that can peer into the code and point out thrashing and bottlenecks. It provides the API, you use it for whatever code you want.

more here.

So there are those out there trying to help port apps to run on GPUs.

---

I think if you're going to purchase a dedicated cruncher, the best thing to do for now is to save the money and either invest it in stocks and bonds or save it in the bank. It's unknown whether PS3 or XBOX360 will be the better cruncher or which company will even allow crunching on their machines, and Intel will be coming out with 1 teraflop chips in 5 years and supposedly they will be available to the masses. If so, then they would be 10 times as effective as PS3 for at most twice the cost of a single PS3/XBOX.

So, I think the right call is to wait and see before plunking dollars down. Maybe the whole ATI thing will rock and people will be rockin' on dual ATI SLI setups. Even then, it still may pay to wait.

---

paydirt

You will find historically that old computer technology is cheaper then new technology. And there are pros and cons either way. This will be true in the future as well... so 5 years from now, you'll be able to buy an "old" .3TFlop processor for about half the cost of a new 1.0TFloper, so again you will find yourself "waiting for the prices to come down a little more".

By helping with what we've got now, hopefully when the computing power is readily available, the software is ready too. And these new computers can be used to learn about the thousands of proteins in the body, rather then to learn about how to study the proteins in the future.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/