I don't get the "prove it" posts. Coming with arguments that you can't expect people to pay $20 or $100 more for a GPU.

The people asking about GPU projects already have GPU available.

As for "prove it"...well, I have a CPU and a GPU. With WCG I can use 100% of my CPU....and that's it. If a GPU could only pump out 10% more work, that's still 10% more work.

But, do you think we're so stupid to think a GPU with dozens of cores cannot crunch through a whole boatload of numbers in parallel, if given the opportunity? So, 100%CPU + X will _always_ be bigger than just 100% CPU.

Then, there's my home system with CPU _ two GPU from ATI and Nvidia. I expect (but cannot "prove") that this crunches more work than just the CPU alone.

Try to ignore the points, but I picked up 2.5 million in the last few months by projects that will use my GPU capacity. Without those projects, my CPUs would keep humming along, but not doing as much work.

That's with one laptop, one desktop. 2 CPUs and 3 GPUs.

The argument that WCG is somehow more deserving and better for the world is moot. The resources are there, and they're not being tapped for financial, political, or other reasons.

For those who wish to keep their old computers and not spend a cent except for personal entertainment...that's your choice. Some of us need new hardware for work, for education, or just to make better use of our time. There's plenty of horsepower leftover.

Revving up the GPU does not take 300W 24/7. I _can_ prove that by telling you my desktop PSU is on 300W max rating, and my laptop is 90W. If you have even a marginal education about PSUs, you'll know that unless you pile on the hardware, a 300W PSU does not consume 300W.

Most of us ask the question for an honest reason. We want to contribute. The small number of work units performed gives pause. If WCG really want to become the biggest and best...then show it.

Stop whining about how your old P4 is just fine. Retire that dog, consume less power, and crunch numbers all night and day with a faster more efficient machine.

I'm crunching with CPU and GPU near 100% right now, while surfing the web, checking email, even doing some Photoshop work.

For progress...one needs to move forward. Not hold on tightly to decades-old thinking.

Curmudgeons may think their witty. But in the computing world, they're slow and counter-productive.

Hi bono vox,

In case it's useful information, I have a GT 430, which I bought for work purposes but which crunches when I'm not doing development work on it. The GT 430 is the lowest end of the Fermi line (has only 96 cores), but it is inexpensive ($79 when I bought it in Jan.), it doesn't require its own power supply, and it draws only 49W when going full tilt. It also runs cool, so I expect to be able to keep it crunching all summer.

I am looking forward to the new CUDA 4.0 apps at GPUgrid, because right now the GT 430 is slow on GPUgrid WUs. (It can reliably crunch even the extra long-running ones in plenty of time, just not as fast as the big guys.) I also crunch Milky Way on it once in a while since it can do double precision.

Kate

Yes, quit telling us to upgrade hardware.

It's "new thinking" that doesn't consider anything beyond the limited experience of those who are too green to realise that they don't know everything.

And curmudgeons would know the difference between"'their" and "they're"





Well, here's one.

Unfortunately, when I was young, I studied and invested my income rather than spending it all on wine, women and song. Consequently, I have qualifications and assets.

Since I'm now over 40, and hence a curmudgeon, I have been unable to locate employment for nearly 10 years. Because of the assets, which I cannot liquidate, I am excluded from Social Security payments and am living on my dwindling savings. Since I don't get Social Security payments, recruitment agents don't make as much money from me as they would placing green "new grads" so they're not interested - and the people who want IT work done insist on using agents, not recruiting direct.

Consequently, I don't actually have the money to upgrade my system. My old core2Duo does what it can, and even my GT210 manages to process a dozen or so SETI units each day, despite those who believe I should get a newer GPU for THEIR convenience.

I haven't touched C since I got my degree in 1985. I haven't got involved in BOINC development because frankly, the documentation is gobbledegook. There's no map of HOW the system works - a block diagram, for instance - and no apparent guide to where particular modules which handle particular aspects may be located. It all appears to be an exercise in clever use of terminology.

My particular interest would be to have ALL of the returns used - not have to recrunch any positions at all. Quite which magic modules to choose I'd have no idea since there appears to be no guide. Quite which C compiler to use isn't specified. Where is the free compiler-download link? I certainly can't afford to buy a compiler solely for this purpose...

And now there is another BOINC project for GPU's supporting distributed.net called MOO! Wrapper. http://moowrap.net/ It does both nVidia and ATI. As far as I can tell, it is GPU only.

BOINC-client-support for OpenCL will be included in v6.14.xx. Some of the new code is already in the source, but alpha-testing of v6.14.xx haven't started yet so no idea how long it will take before ready for release.