Average Credit
30 Aug 2012 13:22:19 UTC
Topic 196502
(moderation:
I am new to boinc and am curious about the average credit rankings. what is this the average of? average hour/day/week/month?
I am sorry if this has been answered before. I did a few searches with zero results.
Thank you.
Average Credit
)
From the BOINC wiki (Computation credit):
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
The calculation method is of
)
The calculation method is of the exponentially weighted moving average type--which has some interesting characteristics which distract some people and entrance others. If you make a sudden change (such as entering or leaving a project), it takes weeks to make most of the journey from the previous operating point to the new one. But it is not a simple rolling-window average, but weights recent activity much stronger than older activity, in a continuous rather than discontinuous fashion.
And it's been my subjective
)
And it's been my subjective experience that it takes 2-3 weeks to stabilize.
For a project like this that is much more concerned with throughput than performance it's a pretty good metric.
RE: And it's been my
)
At Einstein it takes a few, 2 maybe 3 weeks to stabelize, true. Now it can run
on both ATI & NVidia GPUs, I've added 2 HD5870 GPUs driven by an i7-2600 CPU*.
CPU use for the ATI GPUs is much lower compaired to NVidias GTX470 & 480.
That is 2* cores. The NVidia GPU have none full CPU core. Each ATI GPU has
1 thread of the i7, HT=on.
The 5870 is about as fast as the GTX480, both doing 2, they take ~2,884.93 to
3200, CPU=271.69, Binary Radio Pulsar Search (Arecibo) v1.28 (opencl-ati) and 3,475.66, CPU=1,132.06 seconds for the GTX480.Binary Radio Pulsar Search (Arecibo) v1.25 (BRP4cuda32)
No throughput without performence and vice versa ;-)
RE: No throughput without
)
Of course that's true Fred but I think of them on different time scales
Hi Malcolm, welcome to the
)
Hi Malcolm, welcome to the Einstein project. We hope you enjoy your stay here.
The thing that caught my eye when first reading your message was that you joined only 3 days ago and yet you already have a total credit of ~65K and a RAC of ~5.5K. Those are some pretty serious numbers for such a short time so you obviously have added some decent crunching power to the project.
As others have intimated, your RAC will continue climbing for quite some time and will eventually plateau at a very much higher level than the current value. By the time you read this, the numbers will most likely be higher than the ones I've listed, since they are progressively updated and not locked to what they were at the time of posting.
You've joined the project at a pretty exciting time, with all the new radio pulsar discoveries that have been made of late. I expect another flood of new discoveries in the future when the results of the gamma ray pulsar search are eventually published. This search has been going on for over 12 months now and there are broad hints that there will be some exiting news about it in due course.
Please browse the various boards to see what's been happening and please feel free to ask any questions you have.
Once again, enjoy your stay and thanks for deciding to join.
Cheers,
Gary.
Thanks everyone. you have
)
Thanks everyone. you have all been very helpful.
Gary Roberts wrote:The thing
)
Malcolm, let me add my welcome. Those are some pretty intriguing hosts. It appears that while four of your hosts are listed as including 3x Tesla M2090 GPUs, that none of your work listed in the tasks lists is GPU work.
To Malcolm (out of curiosity): are you not enabling GPU work for a reason? (it is on the Einstein@Home preferences page in your account, with a separate entry for each location--aka venue), and titled "Use NVIDIA GPU".
To the local experts: is there a reason the M2090 cards would not work for, e.g. Einstein BRP calculation. His Sandy Bridge CPUs are pretty capable, but adding even a small fraction of the potential computational output of twelves Teslas would be a substantial addition.
hmmm... i just checked my
)
hmmm... i just checked my preferences it is listed to use the GPU's. I was curious myself if they were kicking on, but after i my machine's output started to climb, i just assumed they were working.
Any suggestions about enabling them would be helpful.
I work on a team that helps configure and assemble server level hardware. Sometimes we don't get as much experience on the application end of things.
RE: hmmm... i just checked
)
Each host has a single BRP4cuda32 Wu, they also look to have plenty of Cuda work for Cosmology@Home, and also a little Cuda work for setiathome,
Claggy