NO, NO, NO I am NOT suggesting that in any way, shape or form!!!! What I am saying is there is a threshold point where there is too much data coming in to be analyzed in a timely manner, it is simple math. Whether WCG can EVER get to that point is something I would have no clue about, but hope it NEVER happens. I do know it did happen over at Seti though, a while back they had so much data being returned they just put them on tape and on the shelf for later when they had more time. Seti has had MANY growing pains over the years, while WCG does not seem to be affected as much, probably thru better planning and resources! I was just saying that the simple act of returning tons and tons of units has to be analyzed before you open the tap and let them flow. MANY projects over the years have increased the amount of crunch time for each unit so we users don't flood the Servers.

agree

i'd give away all my badges just if this would bring the project results faster to the scientific community

First, I have to say that life isn't fair, and not everything is black and white.

If an investment vehicle gives 10% return, and a millionaire invest 10 million while you invest 10 thousand, should you tell the bank to give the millionaire only one thousand, because you too only get one thousand?

If your CPU can process 24 work unit per day - and you get a day - why a person that process 240 work unit per day also get a day?

Why don't YOU go out and get yourself a GPU?

Again, like I said, life isn't fair and not everything is black and white. In golf we have handicap, we have different league and all that. But pulling everything down to your level does not solve everything. I will give no comment to this different badge or not thingy.

It just that on the basic level, it should be based on the amount of work unit they process. They produce the same thing, isn't?

Time for me to express my own personal opinion on this badge issue...

Let's start with what we have now.

What is a CPU? It's not (necessarily) a chip; but some machines do have multiple sockets and so multiple chips. Then again, some chips have multiple CPUs on them. And some CPUs have dual instruction pipelines for a single execution unit ("hyperthreading"). At then end of the day the CPU time we get credit for is the time it took for each WU to process, irrespective of what the execution device architecture was.

When we come to GPU processing, what counts as a single execution device? Do be perfectly honest I don't really know because I haven't looked into this in any detail -- but it's fun to speculate. I would guess, and I think it actually turns out to be a very fair way of thinking, that the execution unit is a "card". Whether you have a card with 16 processors or 512, I think that all those processors will work together to to complete one WU. If you have a card with more processors then each WU will finish quicker, but you'll still do one card's worth of processing in a day. (This is just like the difference between an old P4 or a more modern 2600K -- each thread still does a day's work in a day, it's just that the 2600K gets through a lot more WUs). If you want to get more time (=badge) credit, then you'll need to get more cards.

If my speculation is correct, then nothing really changes. A bronze badge represents 15 days of WU processing time, irrespective of the architecture used to process it. If you have a zillion GPUs on a card then your WU-completed count will rocket through the roof, but your processing time will still clock up a day a day (if you don't do any CPU processing as well). You might think that that's harsh if you've invested heavily in a top-of-the-line GPU card, but it seems reasonable to me. WCG is mostly about using the facilities that people already have but don't fully use to do work that benefits humanity. It's great when people invest specially to help humanity, but I see no reason to change the reward system. Helping humanity is a reward in itself.

A card if run 24/7 will register 24 ''elapsed'' hours. That's what's being used at GPU projects to compute the amount of calculations performed aka, the credit/points base **. If somehow someone is able to subdivide a GPU so it can run multiple HCC GPU tasks ***, then as with subdividing a HT capable CPU into e.g. 8 threads reporting 24 hours each instead of 4 times [the physical processors], see no problem to have a card report time in multiple for concurrently processed HCC-GPU assignments. Of course, whether that will fit into the memory banks of a GPU card is the secondary question. We'll going to learn when the extended Beta arrives, in early 2012.

** Other projects don't seem to maintain any form of time contributed statistics, AFAIHH.

*** which I understand will have multiple targets enclosed, similar to what DDDT1 did, to get them to run long enough and not inundate the schedulers with avalanches of task request and reports.

--//--