Hi. I am building a new pc that will have 3 nVidia GTX 295\'s running in SLI. Each board has 480 cores in parallel and almost 2Gb of dedicated RAM. Each is capable of 1.788 Teraflops of processing power. Altogether, by running 3 in SLI, this will give my machine 1440 cores and a whopping 5.364 Teraflops of power! Yes, it\'s like having a supercomputer sitting on your desk. But, unless your project is able to run CUDA (which the latest versions of BOINC support now), you can\'t make use of any of that power. It would represent a massive increase in processing power for work units. Most of the newer nVidia graphics cards support CUDA. Any plans to add CUDA support for your project? or will I have to keep using my CUDA power for other projects? Thanks!

If you so an \'Advanced search \' for processing on graphics cards, you will find the topic has been addressed, over and over again, over the years.

The short answer is \'no\'. \'Fast\' gets nothing if the instruction set isn\'t sufficiently robust to support a million-line Fortran program -- and graphics cards don\'t fill the bill. Further, rewrites, testing, and certification of the output would probably take years, even if the cards were capable. One more \'small\' matter, there isn\'t available staff or funds to undertake such an endeavor/endeavour.

It\'s a nice idea whose time is unlikely to come in the current generation of climate models.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I\'ve read this same message in the AQUA, Einstein and SETI forums. On Einstein I received a mail in my private mailbox from someone advertising Tesla products. Sounds like a marketing effort to me. I don\'t have any graphic board and I am not planning to buy one.
Tullio

---

As with tullio, I\'ve seen this spam message by pharrg popping up across multiple projects. He just posts the same message over and over again without any regard to whether Cuda is feasible for that project or not. I wouldn\'t pay him any heed.

---

~It only takes one bottle cap moving at 23,000 mph to ruin your whole day~

Thanks Mary and Tullio for pointing that out. Appreciated.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I\'ve no objection to a cruncher posting the same message on the forums of various projects as long as it\'s relevant to all of them. When this is done it helps if the original poster returns and acknowledges the responses to their post.

Private messages shouldn\'t ever be used for recruitment or advertising on any BOINC project. We\'ve said this in the CPDN News threads and I think the moderators and administrators of just about every BOINC project would agree that this constitutes spam.

Tullio, have you reported the spam private message to a mod or admin at Einstein?

Misuse of the PM system on any BOINC forum makes it very difficult for us to ask our admins in Oxford to enable default email notification of PMs on CPDN. We really do need this so that members are more likely to find and read important messages sent to them by moderators.

---

Cpdn news

Tullio, have you reported the spam private message to a mod or admin at Einstein?

Misuse of the PM system on any BOINC forum makes it very difficult for us to ask our admins in Oxford to enable default email notification of PMs on CPDN. We really do need this so that members are more likely to find and read important messages sent to them by moderators.

Yes, thanks.
Tullio

---

I must say -- it shouldn't matter if 'pharrg' is spamming -- the matter remains -- CUDA is here to stay (at least, for nVidia customers), and can provide a performance boost, which would be very useful, given the size of the Climate Apps dataunits.

My computer's been working on the same units for nearly 1000 work hours, and they're now a month overdue. I installed the newer version of BOINC, to take advantage of CUDA, and now, the workunits' time-to-completion has _increased_.

I respect and understand that there are technical (engineering) challenges to make this all work, and that funding may not be there for a long time (if ever).

I just want to add a vote that CUDA-capable computing, for this project, would be a benefit.

---

Hi HobbesPDX, welcome to the forum.

Yes, we know that CUDA processes stuff really fast. There's another thread about CUDA here.

Now about your models.

A BOINC upgrade to a CUDA-capable version shouldn't make your HadAM models' time to completion increase. I see you have 4 of them, 3 more advanced than the other. Only one model has trickled since 11 April. Has your computer been processing them since then, or has it been turned off over Easter?

Don't worry about going past the models' deadline; the CPDN servers all ignore it and accept late returns.

However, I'm rather surprised to see that your HadAM models eg this one have a report deadline just 6 months after download. I thought all CPDN models were supposed to have longer deadlines than this. Have I completely misunderstood some aspect of Hobbes' models, or are these deadlines a mistake that the programmers need to correct?

---

Cpdn news

I've just downloaded a couple of HadAMs to see what their deadlines are. 22 Oct 2009 ie exactly 6 months from now. But my current HadAM3Ps which are much shorter models have deadlines in April 2010. This doesn't seem sensible or logical. I will report this anomaly to the programmers.

---

Cpdn news

wow. I had no idea that it was an error.

My computer is [almost] never turned off -- PLUS -- I have BOINC running all the time, albeit at low priority.

Thank you VERY much for the information about late work. [that it's accepted]
It would be great if you'd post that somewhere prominently, so others can know that, too. (I almost deleted the workunits, thinking I'd not get credit)

Thanks for the response!!

---

The acceptance of past-deadline results is prominent in the CPDN README collection about running the models, and we quite often post useful reminders in the forum News threads. Both can be reached through the links in my sig.

---

Cpdn news