While I find it a little diaspointing that my Linux box and Windohs box of the same spec get different scores, I'd be happy to lose the points completely if we could get optimised science applications that get more work done. The work is why we're here, not some masturbatory exercise in statistics.

Sekerob,
The claim made in that thread is false. The claim was that optimsing the Linux kernel allows WUs to crunch faster. You might get a decrease in crunch time in the order of less than 1% but that's all. Rick Alther's response in the same thread explains why there can be no significant performance gain (the Linux science apps are static compiles that do not depend on kernel routines).

Dark Angel,
I would gladly throw out the credit system too, get rid of it entirely and forever. However, underestand that the credit system is not what delays/prevents optimised apps. It's because optimised apps give slightly different results and the results may not validate in a quorum. That problem could be solved fairly easily, I think, but most projects have decided the best way to crunch more WUs is to recruit more crunchers and wait for faster CPUs to hit the market. If that works for them then that's what they should do, maybe. I say optimise the apps AND recruit more crunchers but they're the experts, not me.

My reason for wanting all projects to abandon the credit system completely and stop publishing member stats is that some crunchers are proposing that projects should use a quorum of 3, even when 1 or 2 is sufficient, just to reduce cheating. I have never heard of anything so childish and stoopid and I fear that projects may eventually cave in to the requests of those morons.

As has been pointed out, optimization of the agent has virtually nothing to do with the speed that a work unit is processed.

As has not been pointed out, the differences in Windows and Linux in terms of # of work units that can be processed is also marginal (presuming that they are both running BOINC). Big differences almost assuredly are just the differences in the individual work units that were run.

The reasons for having a quorum is not only cheating but more importantly for differences in processors, undetected overclocking errors, transmission errors etc. etc. Actually the cheating is probably the least important of the reasons.

It is critical to the science that the results of the calculations are correct.
At the same time it is the very nature of distributed computing projects that not all participants have the necessary equipment to guarantee absolute accuracy; hence a quorum system.

Think of it this way: Would you trust a therapy that must be correct to save your life to a calculation that was run on a computer built by a newby hobbyist with dubious lowest cost used parts, flaky power and transmitted over rural telephone lines?

I know we all want to get as much as we can done to fight the disease but we must do so with deliberate speed.

Kind regards to all. You are each important to the cause and the rest of us thank you.

I had a classic example yesterday of the need for a quorum....one FAAH work unit needed 6 results submitted before validation was possible. Thankfully I was no.6.

BTW a quorum means four, I'm not sure what three would be, a tri....something or other.

Cheers

I just love to learn new stuff...

I assumed, wrongly as usual, that the english derivatives would be quad, quarter etc., My scant background in Social Club structure, which always required, perhaps by chance, that a quorum of 4 be required to hold a meeting or official assembly has led me down the proverbial garden path yet again.

Oh! Yes I'll put my hand in the air on the earlier discussion too. I'm a points hound, and proud of it!!! The points system, if properly managed and promoted, can be one of the best incentives available to these projects. Everybody joins for the science and the milk of human kindness factor but the dropout rate is about five in six, or so I'm told.

Points may be used by teams to retain interest through friendly competition, sub team and inter team rivalry and "races" to predetermined goals etc., You watch your stats improve and as they slow you work out how to make them better. That means more productivity ultimately and the science can't help but reap the rewards.

IMO most of these projects fail to recognise the true value of their points system or take full advantage of it. Understandable but lamentable too.
It generally falls to the "go getters" in the teams to get the ball rolling and keep it going, no easy task but very rewarding in the long run for all involved.

I'm Aussie if you hadn't guessed by now and that leads me to one more question. Why would the team that I joined be listed as "International"? I did not form the team but am the only active member atm. Team name is OverClockers Australia and is represented strongly in many DC projects around the globe.(I'm here because I lost confidence in the previous project) Is there a way to have BOINC recognise the team nationality?

Cheers ozylynx