I just noticed that there is a new version of BOINC (6.4.5) and it is now advertising that these applications can support CUDA now if they so desire to. I for one would love to see Rosetta support CUDA.

I just noticed that there is a new version of BOINC (6.4.5) and it is now advertising that these applications can support CUDA now if they so desire to. I for one would love to see Rosetta support CUDA.

Personally I think most projects will have some kind of support for video cards at some point in time. The problem right now is that there are only a few cards supported. Currently only 11 cards are in green and 5 cards are in grey, those are ALL the cards currently supported!
Here is a link {url]http://www.gpugrid.net/forum_thread.php?id=316[/url]

---

I just noticed that there is a new version of BOINC (6.4.5) and it is now advertising that these applications can support CUDA now if they so desire to. I for one would love to see Rosetta support CUDA.

Personally I think most projects will have some kind of support for video cards at some point in time. The problem right now is that there are only a few cards supported. Currently only 11 cards are in green and 5 cards are in grey, those are ALL the cards currently supported!
Here is a link {url]http://www.gpugrid.net/forum_thread.php?id=316[/url]

gpugrid is not boinc. They do not support all the cuda cards, but boinc does.

Boinc is middleware.

gpugrid is not boinc. They do not support all the cuda cards, but boinc does.

Boinc is middleware.

gpugrid is not boinc. They do not support all the cuda cards, but boinc does.

Can somebody tell in brief, what are the projects, supporting cuda and supported by BOINC. In other words, using BOINC middleware, what projects take significant advantage from GPU. I'm just assembling 2xGTX system for my fun and searching projects to participate. Currently I know only one life sciences project: GPUgrid.

thanks for Your tips

Jaak Ennuste

SETI@HOME is another. You need the latest BOINC and the latest NVidia drivers. The early versions of the 8800 GTX/GTS do not work with cuda. Be warned though, after reading the messages over on the SETI board it seems to have a lot of bugs. Some people have also had problems with overheating. I'll wait til they come out with version 2.0 and work out the bugs.

SETI@HOME is another. You need the latest BOINC and the latest NVidia drivers. The early versions of the 8800 GTX/GTS do not work with cuda. Be warned though, after reading the messages over on the SETI board it seems to have a lot of bugs. Some people have also had problems with overheating. I'll wait til they come out with version 2.0 and work out the bugs.

Thanks for Your info. I'm looking forward to donate to life sciences. My new steel has 2 x NVIDIA GTX 280 cards running on Phenom host with total 1,8 TFLOPS.

SETI@HOME is another. You need the latest BOINC and the latest NVidia drivers. The early versions of the 8800 GTX/GTS do not work with cuda. Be warned though, after reading the messages over on the SETI board it seems to have a lot of bugs. Some people have also had problems with overheating. I'll wait til they come out with version 2.0 and work out the bugs.

Thanks for Your info. I'm looking forward to donate to life sciences. My new steel has 2 x NVIDIA GTX 280 cards running on Phenom host with total 1,8 TFLOPS.

It's important to remember that CUDA is an API that an application can use. How well it is used, and how "buggy" the code is dependent on the programmer, not CUDA.

The SETI CUDA application seems to be having major teething pains.

The GPUGRID CUDA app is a little farther down that path, and it is running much more reliably, when work is available on that project.

It's my understanding that several other projects are considering supplying CUDA-based apps.

Mark

---

SETI@HOME is another. You need the latest BOINC and the latest NVidia drivers. The early versions of the 8800 GTX/GTS do not work with cuda. Be warned though, after reading the messages over on the SETI board it seems to have a lot of bugs. Some people have also had problems with overheating. I'll wait til they come out with version 2.0 and work out the bugs.

Thanks for Your info. I'm looking forward to donate to life sciences. My new steel has 2 x NVIDIA GTX 280 cards running on Phenom host with total 1,8 TFLOPS.

It's important to remember that CUDA is an API that an application can use. How well it is used, and how "buggy" the code is dependent on the programmer, not CUDA.

The SETI CUDA application seems to be having major teething pains.

The GPUGRID CUDA app is a little farther down that path, and it is running much more reliably, when work is available on that project.

It's my understanding that several other projects are considering supplying CUDA-based apps.

Mark

wouldn't they be better off using opencl now?

---

Boinc is middleware.

gpugrid is not boinc. They do not support all the cuda cards, but boinc does.

Can somebody tell in brief, what are the projects, supporting cuda and supported by BOINC. In other words, using BOINC middleware, what projects take significant advantage from GPU. I'm just assembling 2xGTX system for my fun and searching projects to participate. Currently I know only one life sciences project: GPUgrid.

thanks for Your tips Jaak Ennuste

I can tell you one Project that I am now contributing to that uses my new Nvidia video card...folding@home. It was easy, credit comes thru quickly and the only problem was finding the number of my Team so I could join it. I had to go to my Team webpages and ask them for the number, I could not find a way to search for it on the folding page. Folding even supports ATI video cards, although I have not tried that yet. Here is a link to the page about the video cards
http://folding.stanford.edu/English/WinGPUGuide
It runs in the tray but does not run under Boinc.

---

Boinc is middleware.

gpugrid is not boinc. They do not support all the cuda cards, but boinc does.

Can somebody tell in brief, what are the projects, supporting cuda and supported by BOINC. In other words, using BOINC middleware, what projects take significant advantage from GPU. I'm just assembling 2xGTX system for my fun and searching projects to participate. Currently I know only one life sciences project: GPUgrid.

thanks for Your tips

Jaak Ennuste

There are only two BOINC projects supporting CUDA ... SaH and GPU Grid. Older slower cards will not be that suitable for GPU Grid with the one week deadlines. I have an 8500 that is CUDA capable and it would not finish one task in that time (as best as I could tell)... my 9800 takes about 12-17 hours per task and the GTX280 about 4-6 ... you can mouse about over there and look at the work done by the two computers and get a feel for it if you like ...

SaH's application has a number of difficulties the most damaging one is that sometimes the app will crash the card and corrupt the display requiring a reboot and all tasks run after the crash for either project will fail until the reboot.

GPU Grid current window app uses more CPU than it should and a new version is promised "soon" ...

Running the GPU with as little impact on the CPU set requires some extra fiddling or the use of 6.5.0 ...

All of that aside there are serious issues with the handling of the GPU in BOINC Manager in that the current versions, including 6.5.0 don't manage the GPU as it should ... see the BOINC Dev mailing list for the discussions ...

Rumors abound about Einstein having a GPU version in the works, but as far as I know they have not acknowledged having this version at this time ...

---

OpenMM workshop -- @ Stanford Feb 12, 13

I've talked about OpenMM before -- it's the library we're releasing to allow other scientists to take advantage of the great speed up found in our GPU code. We're now having a workshop to describe it. Here is the announcement that went out (see below). Check out this link for more details.


Want to learn how to quickly set up and run molecular dynamics (MD) simulations? Or are you a veteran to MD who wants to know how to obtain 100X speedups of MD code on GPUs?

Whatever your MD background, the free upcoming two-day OpenMM workshop promises to be an exciting, knowledge-filled opportunity, featuring experts from both molecular dynamics and the GPU industry.

Day 1 of the workshop is geared towards MD developers and will discuss how MD code can be sped up on GPUs and in particular, with the new OpenMM 1.0 library. Day 2 of the workshop is for researchers interested in using MD in their research, introducing MD theory and the new OpenMM Zephyr application, which makes it easy to run and visualize MD simulations. You can sign up for one or both days.

Read more about OpenMM, OpenMM Zephyr, and molecular dynamics (see attached). Or visit http://simbios.stanford.edu/EventsOfInterest/OpenMMWorkshop.htm for more details.

OpenMM workshop -- @ Stanford Feb 12, 13

I've talked about OpenMM before -- it's the library we're releasing to allow other scientists to take advantage of the great speed up found in our GPU code. We're now having a workshop to describe it. Here is the announcement that went out (see below). Check out this link for more details.


Want to learn how to quickly set up and run molecular dynamics (MD) simulations? Or are you a veteran to MD who wants to know how to obtain 100X speedups of MD code on GPUs?

Whatever your MD background, the free upcoming two-day OpenMM workshop promises to be an exciting, knowledge-filled opportunity, featuring experts from both molecular dynamics and the GPU industry.

Day 1 of the workshop is geared towards MD developers and will discuss how MD code can be sped up on GPUs and in particular, with the new OpenMM 1.0 library. Day 2 of the workshop is for researchers interested in using MD in their research, introducing MD theory and the new OpenMM Zephyr application, which makes it easy to run and visualize MD simulations. You can sign up for one or both days.

Read more about OpenMM, OpenMM Zephyr, and molecular dynamics (see attached). Or visit http://simbios.stanford.edu/EventsOfInterest/OpenMMWorkshop.htm for more details.

Anyone involved with mini rosetta code gonna be attending this?

---

In seti someone with i7 3ghz and 9xxxGT is doing 8500 rac with single desktop with cuda computing. Maybe it is time for Rosetta to jump in and earn the computing source. Can anyone close to Rosetta please to inform us when might Rosetta Cuda start?

I hope many rosetta members want to push rosetta admins to start cuda, it is reality that better computing technology is already out there, so the project very likely need to evolve into it.

---

In seti someone with i7 3ghz and 9xxxGT is doing 8500 rac with single desktop with cuda computing. Maybe it is time for Rosetta to jump in and earn the computing source. Can anyone close to Rosetta please to inform us when might Rosetta Cuda start?

I hope many rosetta members want to push rosetta admins to start cuda, it is reality that better computing technology is already out there, so the project very likely need to evolve into it.

The project is aware of the issue. But the question is if the application is suitable to be recoded as a CUDA application.

If you want to do Bio and CUDA the option right now is GPU Grid ... though the 8500 is low end for their application ...

Someone is working on an ATI accelerated version of Milky Way and we can hope if that gets working that it is ported to the Nvidia cards so that we all can jump into the pool ... :)

I am sure that in the next few months that a number of other projects are either going to come into play or that existing projects will port their application over to CUDA.

Be aware that there are still issues with CUDA and video drivers are one of the keys to making it work ... some, particularly those with 64-bit systems have been having some trouble getting stable at GPU Grid.

Yet there are not few people in seti are successfully using single pc and doing 6000-9000 rac with their cuda, which means this is realistic advancing solution for dc projects in the near future. Well let's hope someone should announce something in the front page of this site about cuda maybe this year.

---

Yet there are not few people in seti are successfully using single pc and doing 6000-9000 rac with their cuda, which means this is realistic advancing solution for dc projects in the near future. Well let's hope someone should announce something in the front page of this site about cuda maybe this year.

It is not that CUDA is not a viable means of doing models. Or even doing models relating to the problem space of protean folding. The question is if the approach THIS project is taking is amenable to the way that the GPU cards process programs.

If the cards prove to be suitable, then an entirely new application has to be developed.

And I have watched the CUDA kerfuffle on SaH and there are issues with the CUDA application though they seem to be working them out pretty quickly...

Then you have the current issue that with ATI and Nvidia having different APIs and programming models to satisfy the universe of participants the project would have to support three different applications... a pretty tall order ...

Still, it is early days and were I running the project I would be waiting a bit to see if the generic programming model gets support so that I would only have to develop one application to one API that would run on both types of GPU ...

Hello With all.
My graphics card is ATI RADEON Version 2008.0923.2139.36956, and I n'do not have, (on my Desktop machine Intel Core 2 CPU 6300 Windows XP), of chart NVIDIA of GetForce.
My question is: Then I to download on my Computeur the thecnology of CUDA?
Is ess that possible?, and bond of remote loading exists 1 however there?. In this case thank you to say to me which file I must download and install.
Thank you very much d'advances for your answers and your councils concerning my question....

---

Hello With all.
My graphics card is ATI RADEON Version 2008.0923.2139.36956, and I n'do not have, (on my Desktop machine Intel Core 2 CPU 6300 Windows XP), of chart NVIDIA of GetForce.
My question is: Then I to download on my Computeur the thecnology of CUDA?
Is ess that possible?, and bond of remote loading exists 1 however there?. In this case thank you to say to me which file I must download and install.
Thank you very much d'advances for your answers and your councils concerning my question....

I will start...right now Rosetta does not support video card crunching, only cpu crunching. Seti is one project that does support video card crunching, so does GPUgrid. Both, Seti and GPUgrid, require Boinc version 6.4.5 or higher to work with your video card.

---

Hello With all.
My graphics card is ATI RADEON Version 2008.0923.2139.36956, and I n'do not have, (on my Desktop machine Intel Core 2 CPU 6300 Windows XP), of chart NVIDIA of GetForce.
My question is: Then I to download on my Computeur the thecnology of CUDA?
Is ess that possible?, and bond of remote loading exists 1 however there?. In this case thank you to say to me which file I must download and install.
Thank you very much d'advances for your answers and your councils concerning my question....

I will start...right now Rosetta does not support video card crunching, only cpu crunching. Seti is one project that does support video card crunching, so does GPUgrid. Both, Seti and GPUgrid, require Boinc version 6.4.5 or higher to work with your video card.

Thank you for your answer.
Just a precision, the BOINC on my Computeur is with version 6.4.5 since December 9, 2008.

---