Hi gang,
I may have missed an already posted question regarding this topic and if I did I apologize.

Are there any plans in the near future to have a gpu client like seti for rosetta and the boinc project?

This question has been asked a lot in the past, but I don't remember seeing an answer to it from the project in the context of the new GPGPUs that should be easiest to code for and that should have more high-precision capabilities for scientific math.

---

Hi gang,
I may have missed an already posted question regarding this topic and if I did I apologize.

Are there any plans in the near future to have a gpu client like seti for rosetta and the boinc project?

There is a quote from one of the guys at the BL - the problem with GPGPU for Rosetta is that for Rosetta the project is the code and not the results - the results are the test of the code (although there have been some very interesting/useful results produced) because accurate protein modelling/design software doesn't currently exist. That means that the code is constantly being modified, and not just by the Baker Lab...

Having said that, I would think if someone would offer to port a stable part of the code to another architecture (OpenGL?) then i'm sure it'd be gladly accepted.

---

Damn!
I wonder how big the Tflop spike for SETI is going to be.

Sucks I only have ATI video cards.

---

Damn!
I wonder how big the Tflop spike for SETI is going to be.

Sucks I only have ATI video cards.

Folding uses ATI cards

---

Damn!
I wonder how big the Tflop spike for SETI is going to be.

Sucks I only have ATI video cards.

Folding uses ATI cards

I don't really like Folding@Home. :)

---

I did folding at home when it first started, when the work units were alot smaller too. I still like folding but right now I like rosetta because it is quick and easy to setup with boinc and my work computers have no trouble connecting.

With f@h I couldn't get the clients to connect on the work pc's. Boinc connects fine. I just recently updated my two home boxes to quad cores but my old athlon xp 1800 wouldn't cut the mustard for folding anymore.

I would love to see the cuda client put to use on the rosetta project.

Damn!
I wonder how big the Tflop spike for SETI is going to be.

Sucks I only have ATI video cards.

Folding uses ATI cards

I don't really like Folding@Home. :)

I am not even sure what Folding is looking for, other than that they are "folding" proteins. BUT they do let my new video cards crunch, and I will NOT do Seti again, so there is little choice right now. Yes if other projects let us use the video cards my cards would move, but for right now it is the only option, for me.

---

I did folding at home when it first started, when the work units were alot smaller too. I still like folding but right now I like rosetta because it is quick and easy to setup with boinc and my work computers have no trouble connecting.

With f@h I couldn't get the clients to connect on the work pc's. Boinc connects fine. I just recently updated my two home boxes to quad cores but my old athlon xp 1800 wouldn't cut the mustard for folding anymore.

I would love to see the cuda client put to use on the rosetta project.

I do not do folding with the cpu just the video card, although I am not sure there is any difference in the end. The dual core cpu is actually doing ABC@home while the video card crunches for Folding. Yes if Rosetta came up with a CUDA version, I would be crunching for it instead.

---

Having said that, I would think if someone would offer to port a stable part of the code to another architecture (OpenGL?) then i'm sure it'd be gladly accepted.

I meant OpenCL ;)

---

GPU Grid is doing folding of some form or another, and a pair of other things ... only works on Nvida cards and then it really only works well on cards at about the 9800GT level or better ... though some have been doing work with slower cards.

The only bad thing is that it also puts a pretty heavy hit on the CPU to run each GPU core (a GTX295 card has two "cores" and so you can have 2 tasks in flight at the same time) ...

More interesting the i7 seems to tolerate the GPU Grid load better with only a 7% load (costing me 21% to support 3 cores) while on a straight 4 core Q9300 I lost 22% CPU for one GPU ...

A lower "cost" application is promised ... still waiting ...

Something to my mind that is almost better is what is looking like a more stable Mini-Rosetta application that I have been running in Ralph ... hopefully coming soon to a computer near you ...

---

The only bad thing is that it also puts a pretty heavy hit on the CPU to run each GPU core (a GTX295 card has two "cores" and so you can have 2 tasks in flight at the same time) ...

I tried folding@home through my ATI card and I found that too about the the CPU hit. It was so bad I gave up before the first WU was completed. Impractical for the machine I had at the time and I'm not keen to revisit it even with my new system with an uprated card as well as processor.

---

I tried doing folding@home on my 9800gt while my q9550 did 4 wu of rosetta but it never failed that one or the other would fail. Now I took folding completely off my machine, added another hard drive for raid setup and reinstalled the os and I am still getting compute errors. I have no idea what is going on but hopefully I can get it worked out soon.

my q6600 has 4 cores for Boinc, and there is very little cpu overhead used by my nVidia 9600 gso crunching for Folding@Home. They play VERY nice together...

my q6600 has 4 cores for Boinc, and there is very little cpu overhead used by my nVidia 9600 gso crunching for Folding@Home. They play VERY nice together...

But can it run crysis?

---

dude, i don't know if even SkyNet could run crysis, lol...

dude, i don't know if even SkyNet could run crysis, lol...

Once I get the stability issues check out and everything working right I may have another go with running folding on the gpu again while running rosetta on the four cores. My temps sensors are stuck on two of the cores at 41c, they work if it gets above there. I may rma the cpu back to newegg.

apologies in advance if this is somewhat off-topic, but i also think it is somewhat on-topic, as the "limitations" of the PS3's Cell BE architecture may be somewhat analogous to those of gpu's:

PlayStation 3 tackles black hole vibrations - Consoles capable of running simulations that rival supercomputer

The system is tailored to simulations that require massive amounts of computations, but relatively little RAM memory.

apologies in advance if this is somewhat off-topic, but i also think it is somewhat on-topic, as the "limitations" of the PS3's Cell BE architecture may be somewhat analogous to those of gpu's:

PlayStation 3 tackles black hole vibrations - Consoles capable of running simulations that rival supercomputer

The system is tailored to simulations that require massive amounts of computations, but relatively little RAM memory.

Depending on the card and GPU architecture the types and accuracy of floating point can also be an issue.

---

my limited understanding is that one of the primary reasons there may be difficulties in porting Rosie to PS3 (and possibly, gpu?) is the limited amount of available ram...

but i could also see where fp concerns could arise.