Assimilator1 - here's where I'm coming from... I have volunteered my machines to help in the search for some indication of a radio frequency signal (more likely a directed one, although perhaps a stray one might not be out of the question), that might be extra-terrestrial in origin. However, I do not get hung up on the <quote> science <end quote> of this project because when it comes to real scientific pursuits and research, data collection and analysis is not done in the manner that this project is doing it... And even with it being done on such a large scale, some controls could have and should have been put in place a long time ago.
I consider what we're doing sortof using a big sieve, to essentially gross screen the data... So when people argue about the various clients and how not running X version would somehow "miss some science" (and narzy - really.... although you had the opposing point of view from networkman, I'm not jumping on you but am speaking of the general consensus that I've seen posted around the net), then I'm gonna laugh in your face... That is just my opinion as a real live scientist, and any other real live scientist must agree.
This is a cool project and unprecedented in its scope, but the "science" bit must be put into perspective.
And regarding the variability of WU times with 3.x. Assimilator1 - in order to illustrate what I've written above, absorb the following, based on.... hmmmm.... I don't know, maybe thousands of WUs that a number of us have sifted through, put into spreadsheets, and charted:
Why is it that irrespective of how many gaussians, peaks, or spikes are in a WU, the ONLY processing time variation (run on the same platform/machine) that you will see with 3.x, is based on angle range ONLY.
Now think about that for a minute.
With 2.x, EVERYONE knew when you had a WU with a gaussian and/or alot of spikes because they always took longer. NOW... with 3.x, it doesn't matter what's in the WU. If any of you have Setiqueue or Setilog or Setispy or something that logs your results along with angle range data (and I guess this is where BK's proggie would come in handy...😉), try putting that in a spreadsheet and sorting by angle range. Sort it by machine as well so that you don't try to compare cross-platform. Take note how the variability between supposedly disparate WUs don't vary by more than a fraction of a % to maybe a few percent at the same angle range.
So what does this say about "the science"?
Now I understand that they reportedly are using a "more optimized fft calculation" supposedly meaning "faster" gaussian calculations... But based on our data... all of that is irrelevant.
Believe me, when a bunch of us starting digging out and sorting through our 3.0 results by angle range, and then posting our findings in a few Ars threads - that's when I realized that... well... I won't even say it... 🙁
I'm not going to be the one to start any conspiracy theories, but all you need to do is check for yourself. And as a note, the comparisons were made on dedicated seti crunchers. You will see more variation between WUs with the same angle range if the result came from a machine that was used for other things...
That's why I ask - what "more science" is my machine doing with 3.x if it doesn't even appear to be doing any "science" at all? Ie., if I have a million gaussians in a particular WU, that's irrelevant because the WU time will come out almost identical, or perhaps +/- 1-2% different, from a same angle range WU run on the same machine with no gaussians... This is why when Ars selected a new benchmark WU, there was a wink wink about not even really needing to use the actual archived unit - that you could do the bench on any 0.417 angle range WU and still have a valid benchmark - despite what was in your WU.
You all be the judge. What Beyond, Roelof, I and a few others did was to put our money where our mouths were and take the time to sort through what we had... And Roelof posted some of the charts he made on his Setispy website. We gathered data about the VLAR WU problem with the 3.x CLI and were apparently ignored....
So.... maybe all of this means that our WUs are "blanks" in essence, and that an ET signal at a certain AR might trigger a real long WU time as compared to others... Who knows?
Sigh... 🙁
I will keep crunching - pretty much on my fastest machines only, I'll fold on the slow ones, and try to start shunting some more over to RC5, perhaps to the Feddies.... 😉 Unless someone is interested in making me a Borg Queen... 😛 (although my # of blocks is way too pitiful to justify that, but it would be a good P.R. stunt though... 😉).
