PDA

View Full Version : Huge number of circuits evaluted?



spectra
01-07-2004, 09:34 PM
How is that on occasion some clients evaluate what seems to be a huge number more of circuits over other clients?

Example for January 7th 2004:

adreamG4b 828255
srr2 36945
Omnipotus 35910
Feral_Boy
Proto_Clown 34950

adreamG4b did over 22 times more circuits in 1 day. How does that happen?

Thanks
Spectra

adream
01-07-2004, 09:59 PM
i was wondering this too, as i run adreamg4b here and it did puzzle me...
im running adreamg4 on one processor and adreamg4b on the other processor of a dual g4 running os x 10.3.2

cheers

adream

safemode
01-08-2004, 10:56 AM
this could be simply a result of different architectures. Perhaps the calculations and such that dhe does lend itself to the G4's abilities. Some features are slower than x86, some are faster, being dual cpu probably cant hurt.

spectra
01-08-2004, 09:57 PM
If a G4 were superior in architecture, wouldn't that machine be at the top of the list frequently? I have no idea what any island is being processed on, but today we see another client has done over a million circuits. While the rest of the crowd picks up in the mid 30,000 range.

I was wondering if it has something to do with a generation change on a particular island.

safemode
01-09-2004, 09:55 AM
every now and then you can get a good number of circuits that only take about 15 seconds to process on a P4 1.7Ghz and about 20 seconds on a athlon tbird 1.13Ghz. I wonder if anyone is clustering their island instead of setting up a team for their local computers.

HitchHiker
01-16-2004, 11:37 AM
I myself use a G4 dual 1,42 and on average over the past 85 days it evaluated some 26000 circuits/day.
Just an hour ago it went into a frenzy it seems. It's evaluating a circuit every 1.5 second instead of every 1 or 2 minutes! Is this just because all the islands were reordered or what?

safemode
01-16-2004, 11:55 AM
new target. my athlon 1.13Ghz is doing an island every second and my P4 1.7Ghz is doing an island every 600 or so milliseconds.

HitchHiker
01-17-2004, 06:16 PM
My G3/266 is doing a generation every 8 seconds now too. I almost can't imagine that a new target would make so much difference. Isn't it Miguel optimising his software?
If only he would tell us something every now and a while.

safemode
01-18-2004, 03:23 AM
there are a lot of simple generations that fail really fast in a new target. I've seen this happen 3 times across targets and it's always when it switches over. It's not until later generations that they last longer as they begin to pass more and more tests. The logic there is pretty clear, it's just a bunch of dead on arrival generations. Also, these last two targets have had much more of a restriction on what type of circuits even get evaluated in the first place.

HitchHiker
01-18-2004, 09:58 AM
This could be the case indeed as in the beginning figures were not very good esp po & p1 close to 0.5. But at the moment, figures are very much more better: fitness close to 1 (still) p0 & p1 both at 1 en p2 at 0.5. These aren't disposable generations I should think. :D