That's exciting news! I need to start checking in on the BOINC site more often and keeping up with these developments. Would love to help with beta testing of this!

Quoting myself

"depending how fast-tracked the 6.14 development is."

;>)

Yes, you have no information, indeed!

;P

edit May 23: 100% reliable sources explained that WCG wont need in the first instance an OpenCL detecting client, though it's definitely an OpenCL running app. The hurdles to take to get passed issues such as homogeneous redundancy are substantial!

The best arguments are "their" v. "they're" and ... shut up unless you're going to code the software yourself. Empty.

Why does the fake "old school" think 1990s thinking is better? I started out on 8086 and Apple II and Tektronix green screens for APL class. My HP programmable calculator was better than feeding mag strips like my dad's. My first red LED watch was replaced by an LCD. And I was happy to move off the manual typewriter to a Selectric.

I appreciate that some people don't want to spend money on buying anything. Don't spend it. But don't scold the rest of the world who keep up with current technology.

The purpose of WCG is supposed to be "...to create the largest public computing grid benefiting humanity..."

Real-world: Size is not as important as effectiveness. Who wants a herd of slow bleating oxen when you could have a fleet of GPS-guided combines? If it's only for publicity, then the marketing/PR guys can tout how many thousands of people are "participating".

But I find it hard to believe that you still deny that targeted data crunching on GPUs will not be faster than CPUs.

The NRW folk are reporting 260-500x faster speeds for their post-processing when they tested GPU against CPU. Is it worth it?

Their conclusion:

"GPUs with the power of small supercomputers are becoming ubiquitous in consumer computing devices. The test system that we used is a modest one and is exceeded by many home computers used for gaming or for high definition video."

and

"Effective GPU adaptations of routines for commonly used calculations such as the optimal superposition/RMSD are important if we are to fully utilize the enormous power being made available through the generosity of participants in projects such as NRW."


http://www.worldcommunitygrid.org/about_us/viewNewsArticle.do?articleId=160

1. Do good. 2. Do more good. Keep your trusty Pentium 4 foot warmers and feel cozy. I'll vote for "more" and hope fewer and fewer researches feel the way you do.

Short-sightedness keeps applications tied to CPUs. Invest some of that research into efficiency...and it will pay off. That's critical thinking. And, that's aiming for the future.

If the world wants to pass you by, just wave and move over one or two lanes. You may find that sooner, rather, you won't be needed. You may be appreciated for PR...you know, counting heads of cattle and all. But, statistically speaking, you won't be making a significant contribution.

There are many hours, days, years of SETI crunching on PowerMacs and 386s that seemed high-enough-tech at the time. I can crunch more data in a week. I don't pine for the good old days. I embrace that we have cheap, raw power available for those who can use it.

Maybe something will pop up in beta WCG soon.

If you"ve been reading all the GPU threads for the past 4 years, like I have, you"ll realise, like I have, that WCG will not adopt a GPU application until it can be run without impacting the user.

My 2 cents.

Base setup principle at WCG has been and always will be: Completely eliminate "user impact" or at the very least minimize AMAP [The CEP2 "one per client default after opt-in"as example]. If volunteers are OK with some or more influence on their everyday use they can select one of the 4 pre-defined profiles [MAXIMUM OUTPUT e.g.] and then customize to their hearts content.

For GPU crunching the default is to NOT run such a science during user input for it would be noticeable [and there will be those coming along who have other experiences, but that is not WCG's departure point of view when it starts offering sciences to the wider volunteer community]. In that, options are available to for instance do CPU crunching during use, most don't notice, but not GPU crunching. If you do while using, you might for instance see your mouse make moves you'd never seen before.

Nothing so bad as having brand new Distributed Computing volunteers arriving and departing within the weeks or month. BOINC world wide retention remains to be less than 20%, WCGs is higher, and certainly no plan to further negatively impact that value.

--//--