Some WU's completing faster than usual
7 Jun 2006 23:53:28 UTC
Topic 191343
(moderation:
I joined over the weekend, and the time it is taking to complete a work unit on my Pentium D machine has been pretty consistent at 43-45 minutes per WU. But today, I had and still have a number of work units completing in 12 minutes. As a result, the granted credit has been very low ~ 2.
Is this normal or are their some tainted WU floating around?
Greg
"It's not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change."
Charles Darwin
Some WU's completing faster than usual
)
This is normal. For details, look here.
BM
BM
RE: This is normal. For
)
thank you, very good explanation.
Greg
"It's not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change."
Charles Darwin
RE: I joined over the
)
Welcome to Einstein. We hope you enjoy your stay here.
Obviously you have discovered the magic of Akosf and his marvellous optimised science apps. The standard app takes a lot longer than that :).
Initially you had the more common "long" results to crunch. Now you are likely to get a whole batch of these "shorts". Please read the link Bernd provided as the difference is fully explained.
My purpose in adding to what Bernd said is simply to alert you to the fact that if you do get large numbers of these 12 min results, you will quickly reach your daily allowance of 64 for your two virtual CPUs (32 per cpu). To avoid the possibility of running out of work, you can have backup projects or you can investigate the use of modified BOINC clients which have the ability to fake the number of CPUs in your machine to increase your daily download capability to a maximum of 128 (4 CPUs x 32/CPU). Check out the thread on BoincStudio in Cruncher's Corner. The early information in that thread is quite dated now as there has been quite rapid development. However it is useful to read the whole thread to get a better idea of potential problems by seeing what went wrong for the early adopters. There is supposed to be documentation in English available RSN.
One final point. You noted your low credit claim of ~2 for the short results. Your results list also shows very low claims for your long results. You will no doubt have seen that you are being awarded around 40 for your longs. When they validate, your shorts will give you around 12. In both cases this is about a factor of six times more than what you are claiming. This is entirely due to the factor of six (or more) speedup achieved by the Akosf optimised app. Most people running EAH seem to adopt the view (correctly in my opinion) that the proper amount to be granted is what the standard app normally claims. That is why the use of "calibrating clients" seems to be the correct thing to do, according to many who post on these boards. Check out the sticky thread about this in Cruncher's Corner.
Once again, we hope you enjoy your stay here.
Cheers,
Cheers,
Gary.