Yep we do pack a punch for a small team.

Not sure why others are trying to bring the crap argument from R@H about the optimized clients in your recruiting thread here. Seems rude & out of place especially for a new team here and given past arguments about the topic.

No, no... I'm the one who brought it up. Sorry about that.

But what a dull forum it would be without a little thread-jacking now and again... :-o

There is a thread in the chatroom, but the experts weren't joining in.

Aren't everyone crunch this good project to help? Still I hope your dedication will help it to be developed more sooner than otherwise.

i think like you but dont feed them lets just show them what we have and what we can do we are in a quorum system and it doesnt matter if we have or we dont have cruncher files just lets take this project like we did in rosetta and show the world that we are the perfect crunching machine

It seems feelings are still a bit hot as well, but overall cooling. I took a week off from forums and everything in general. Work was a good excuse as we are in the middle of harvest up here in Alberta and 14-20 hour days on a tractor or a combine give you a lot of time to think about things.

Despite previous postings on our own forums I have decided to crunch WCG with my fellow team members (XS) for now. I do not like the credit system used but as can be seen in this thread the involvement by staff members is happening.. something that happens far, far too little. Such a small gesture can gravitate a great many things ;)

The question was asked of what the Cruncher "optimized" files are and I dont believe it has been answered well enough yet as this question has raised points on morality and proficiency.

Put simply enough, the optimizations present in Crunchers' files give the potential to inflate the floating point MIPS (Whetstone) and integer MIPS (Dhrystone) based on instruction sets used to run the BOINC based benchmark. These instruction sets include (but are not limited to) MMX(+), 3DNOW(+), SSE, SSE2, SSE3, SSE4 and etc.

Put forth in the question of proficiency (work unit produced per time unit) the cruncher optimized files do not produce more physical output work per time used because the projects utilizing the BOINC infrastructure are not (most appear not to be) optimized to use the many varied instruction (sub)sets. Subsequently the question of the validity of the methods used to benchmark rises. Due to heavy memory use or methods used to produce work units on the CPU core the L1/L2/FSB/HT may "thrash" causing a large slowdown in system performance. The Whetstone/Dhrystone benchmark system in and of itself is essentially flawed in this manner as it does not produce a benchmark number concurrent with actual system performance per work performed. (ie: one type of work is done to benchmark the system while another is done to perform actual work. Sort of like lifting a brick and carrying it 20 feet compared to pushing a wheel barrow full of bricks 10 miles).

This leads into the question on morality. Due to the flaw of the benchmark system in place, can you really cheat? Some believe the benchmark numbers produced do not reflect system performance adequately (myself included) and therefor choose to run an optimized client to produce numbers more in line with what they believe to be correct. This is not frowned upon by the project coordinators and thus the acceptance has become widespread among those who wish to compete with others in the stats for fun. In full this is a failure both on part of Berkely for using a flawed system, allowing the users to use optimizations, and in part the users for embracing the systems when they should have browbeat Berkley for showing open ineptitude..... (yes, this "small" screw-up does ruffle my feathers).

In the end this leaves everyone out in the cold. Berkely gives us a flawed system to use for performance matrices, we get called cheaters for attempting to remain competitive with the rest of the higher end stats junkies (love stats, they show me i'm doing something for the project!), and you guys lose us when we chose to move ahead looking for a better project because nobody is doing anything about the situation.

One thing I should point out however, is that there are past examples of how to do "benchmark" systems correctly. For instance Seti@home CLASSIC performed such that completing one work unit generated one point. D2OL and several other projects have also used this method which I believe to be one of the most fool proof and truthful methods available. Some of these same projects which succeeded for the user base also had a wide variety of clients available to utilize the feature sets of CPU's better. Seti@Home Classic probably had nearly 100 clients for the many OS's and CPU's and architectures. F@H is even coming out with one that works for GPU's!

Anyways, the point is that a system which is flawed and ignored from its inception is hard to measure how corrupt it was and how corrupt it has become. The simple answer is to drop it, not toss it around like a hot potato.

Thank you for your time. I hope i've answered more questions than will be raised