Probably makes sense to finish the CPU tasks ready for processing, together with the GPU tasks...

How about those workplaces that crunch on PCs/laptops which don't have graphics cards.

Crunching is crunching irrespective of what you use.
Sure, I wouldn't power up a P3 24/7 specifically for WCG if I had access to something a lot better.
However, isn't one of the aims of distributed computing to use unused processor cycles? Every little bit helps.

HCC is our first GPU project. We have been working hard since March getting the new BOINC 7 server code running and working on website and background support.

So now we can discuss how we would like to run GPU projects for the future. How do we handle the widow's mite?

I am never short of ideas, good or bad. Personally I would like to see a 'No Automatic Opt-in' policy for GPUs. In other words, though a member may choose to automatically select new CPU projects, any new GPU projects should be chosen manually. This is prudent since GPUs can be tricky and troublesome. We should develop standard methods of publicizing specific software needs for various GPU boards, The interesting questions arise when we have a project with CPU and GPU versions. My purely personal opinion is that we should have a CPU project check box and a GPU project check box. Both should count time for the same time badge. If ever a GPU board can run multiple GPU jobs, then we should count each separate job for time, just as we count each separate job for credits. This is what we are doing now. Everybody is credited equally for their time contributed to HCC even as the GPU contributors also earn vastly more credits. The job that runs for 2 hours on a CPU but in 1 minute on a GPU should be worth the identical amount of credits.

But I think we should hedge our bets. Internally, we should retain a breakdown into CPU and GPU credits in case member demands result in a public display of this breakdown. It might make sense to retain the same breakdown for CPU and GPU time. If we ever change the badges to add a 'G' onto the GPU project badges, then we should be prepared to double the badges for projects that run on both CPU and GPU.

Ending on an aside, I normally run all projects without fiddling with selections. That way I let the WCG staff decide what projects should run in what proportions. But hearing that the flood of GPU HCC work units is placing a potential burden on the server software, I am switching my CPU cores to all non-GPU projects, omitting HCC. This is not a judgment against HCC. It is just pragmatic recognition that HCC will run faster even with no CPUs than it did before it had a GPU program. And maybe it will help the overworked server.

Lawrence

What about people (like me) who's graphic card isn't accepted by the HCC GPU? I'm trying to make it to emerald right now.

CJSL

Crunching for a better world...

[edit]: added quote

Well, having invested in 50+ CPU cores, those of us that have all that money tied into CPU crunching are feeling more than a little bit left out. My honest opinion is that the GPU points are somewhat too generous and risk discouraging the CPU crunchers all together. I'm a little on the fence about this though, I mean with new technology comes more possibilities. GPU crunching is one of these things... Lower power, more results in less time, and much lower cost than CPU crunching. Maybe it's time to cut these rigs loose and just invest into all GPU? Or maybe I just shut them down and run a couple rigs and save on all that electricity. Decisions, decisions...