Pathetic amount of credits for 24 hours crunching?
26 Jun 2012 7:09:44 UTC
Topic 196421
(moderation:
Hi,
I just compared the credits from POEM@HOME & Einstein@HOME and they are dreadfully low.
POEM@HOME (37% GPU Usage) turns around 120,000 credits a day.
Einstein@HOME (82% GPU Usage) turns around < 30,000 credits a day.
Why such the pathetic credit rate for Einstein@HOME? I'm loosing the will to participate on the project as it's lowering my RAC ranks, the only reason I have left is to may be find something and get certificate for it.
Pathetic amount of credits for 24 hours crunching?
)
We roughly try to align our credit to that given out by SETI@HOME, the biggest BOINC project. This is easier for projects that offer CPU and GPU apps for the same workunits. Of course there is nothing that keeps other projects from giving any amount of credit they want, and certainly it would take us only 1 minute to change the configuration on our servers to give 10 time, 100 times or 1 zillion times the credit we give now. But we don't believe in this kind of "race" between projects :-).
Cheers
HB
LHC@home gives even less
)
LHC@home gives even less credits/hour than Einstein@home but I am running it because of its science value.
Tullio
I want an increase of 2% of
)
I want an increase of 2% of the credits to keep up with the inflation...
It is impossible to compare the credits from different projects if they can set up the credits per task. It is nice to have a goal like the discovery of a pulsar. But I won't try to compare it with other science projects and the subjective value of the results.
Is a project worth more if it gives more credits/day?
Credits are like money. The
)
Credits are like money. The more there is the less it is worth.
Tullio
I always thought credits were
)
I always thought credits were defined by the processing power you contribute, so the certificate you can print from BOINC is inaccurate? As credits seem to apply to the amount of calculations done? or is this an error / bug.
When I run POEM@HOME it increases by around 0.12 Quintilian so surely if more processing power is used by Einstein@HOME then it should be greater but's actually a 1/8th of when POEM@HOME is running.
I can understand not giving out as many credits, as it's not a race and I'm in the project to find a Pulsar (A certificate on my wall, is my dream). But I would definitively like an accurate read-out of how much processing power I've contributed.
Personally I'd of thought Einstein@HOME would of been using around 0.25 Quintilian or more a day.
Specs:
AMD FX-8150 3.6 GHz - 4.2 GHz (Turbo Core)
AMD HD 6990 O/C (880 MHz | 1350 MHz)
Gigabyte 990FXA-UD7
OCZ Professional 1.2 kW PSU
RE: I always thought
)
The issue is not what project uses more or less computer power, the difference is how many points or credits each project gives you for the work you do. Also, in BOINC it doesn matter how much load the project apply to the hardware, it just counts how many time is using it (at whatever level of utilization).
The original system of credits was intended to give 200 credits per day of work on a host like a Pentium II computer... (or something in that line). Faster computers will get more points and slower gets less.
There was a lot of changes after that, the credit system changed several times then was added support to GPUs, and nowadays there are projects that give up to 10 times that while others gives much less...
Eisntein credits are between the normal range of credits the projects should give, in fact it's giving a bit more. On a certain host one whole day of work for Einstein gives around a 25% more credits than the same host if it were used for SETI.
We name SETI, because SETI was the project for wich the BOINC plataform was released and it's supossed that any project should give the same credits than SETI. But that is just a recomendation and is not easy to achieve specially for certain projects where the processing time of a task can vary significatively depending on the specific data being analized in that task, and that there are a lot of different combinations of hardware and software that can affect the time to process certain tasks.