So astrolabe, are you suggesting that all crunchers be held to some standard of work each week, and if that cruncher is too small or not up enough to produce a valid large WUs in a given week, that WCG should send them an e-mail letting them know they should 'buzz off' - in a polite PC sort of way while thanking them for the handfull of results they did do over the years?

I do agree that WSG probably doesn't have enough programmers sitting on the sidelines to rewrite code, but I still suspect there could be a few adjustments made (like more than 10 days allowed to complete WUs), that can be done with little or no code changes made.

As a side note, I suspect these same sort of discussions are occuring on financial blogs discussing the Wall Street need for small investors anymore versus HFT computers and whether small investors should buzz off too - same rational can be used in both forums....

Don't strain yourself, just read what I wrote. Besides, who is truly more important to WCG: someone who finished 150 days today or someone who gave us a total of 4 minutes of time today. Who would you devote your energy to?I expect that about 7% come from the 40% of crunchers that contribute the leastWhich means you want the Techs to decide what to do and leave it up to them. Apparently that is how the system works right nowI completely agree and it is too bad that doesn't happen for all.

If there's an easy way to generate multi-sized wu's, and this should be the fact for atleast some of WCG's projects since fairly resently many of the projects got longer expected cpu-times, handling-out small wu's to slow computers and larger wu's to fast computers doesn't need much programming...

... Since, some BOINC-projects already does this, by sending larger wu's to the fastest GPU's, and another project denies Nvidia to get some wu's based on the wu-parameters. Using similar options for cpu isn't a problem.

Basically what's needed are:
1: A method to generate multiple-sized wu's.
2: A method to indicate the size to BOINC. This can be easy "hacks" like including a special string as part of the wu-name.
3: Configuring one or more plan-classes to send/not send work depending on CPU, wu-name, active_frac and so on.

It is safe to say that that "big contributors" are those with a great many CPU cores under their userid. The number of active computers or cores is not published, or even the number of hours per week per computer for the "big contributors", so we (the average cruncher) just don't know. I.E. we are arguing without the needed facts. If we are going to compare, then we need to compare the results per computer, perhaps even per core, not per user id. I think that enlarging job sizes could affect a "big contributors" in a negative way as well, depending on the computer types and how often their computers run WCG. They could be set to only run at night.
A "small contributor" may have the newest, most powerful computer, but only has one. If it only runs 5 hours a week, then that's all it's going to get done. Most of the posters on these forums don't fall into this category, so we are looking at the problem through an inherant bias. If we don't have all the needed facts, then we shouldn't discuss untill we do?
[satire]
If we are contemplating abandoning the 5 hrs/week cruncher, then we need to change the welcome message on the home page.
[/satire]

Cheers

Look on the official project stats page or on dashboard http://bit.ly/WCGPI1 for mean run times + history. If that does not make choice easy, I don't know what.