shocked, 33min wu's? [Archive] - Overclockers Forums

View Full Version : shocked, 33min wu's?

LuKE

09-09-01, 12:45 AM

http://setiathome.ssl.berkeley.edu/10000.html

that guy, yuminsu, 33min wu's? how in the hell?

09-09-01, 12:58 AM

Interesting. Seems like it'd have to be a supercomputer, but I can't think of any that can run any version of the seti client. Maybe a supercooled P4 or Athlon could muster that, but still seems a stretch. I had my P4 supercooled to 2.2GHz once and I tried to run the TLC bench file. It crashed several times but I kept restarting it. It finally finished the unit in 2hrs 49min, but it turned out to be a corrupted result. Not much to go on, but just guessing it would have to be an awfully fast machine.

09-09-01, 10:17 AM

Wow, I'll be an old man before I can hit the big 10K. (oops, to late)

Cy;)

Thelemac

09-09-01, 12:07 PM

Rather reminds me of Ice Tea....

09-09-01, 04:47 PM

Originally posted by Thelemac
Rather reminds me of Ice Tea....

Yeah me too, but at least this guy is taking longer than a few seconds. Seems like they would have some means of monitoring people like this.

Thelemac

09-09-01, 04:57 PM

I agree.

derek

09-09-01, 05:12 PM

Even better, how about the person at #953 overall.

Shifty, 8 minutes 54 secs. 10,342 WUs in 1535 hours.

I want what ever they are running. I have been looking for a deal on an EV67, but the Alphas on the CPU page aren't even close to that.

Derek

chaosdriven

09-09-01, 05:33 PM

Originally posted by LuKE
http://setiathome.ssl.berkeley.edu/10000.html

that guy, yuminsu, 33min wu's? how in the hell?

Perhaps he was using the "speed cheat".

http://www.overclockers.com/

See article entitled "Benchmark Faking", under benchmarking.
I'm not making accusations, just suggesting a possible explanation, as I believe this cheat does fool the seti client.

09-09-01, 09:03 PM

Originally posted by chaosdriven

Perhaps he was using the "speed cheat".

http://www.overclockers.com/

See article entitled "Benchmark Faking", under benchmarking.
I'm not making accusations, just suggesting a possible explanation, as I believe this cheat does fool the seti client.

I don't get it. Since that thing doesn't really make your system faster, how can it make the seti client work faster?

Morpheus

09-09-01, 09:14 PM

Yes, Chaosdriven... it does, but it does not fool the science server @ Berkley...

Buy the time I had emailed Eric Korpela about IceTea (on our team) they already had him flagged... these folks will be producing WUs for accounts that will suddenly become "locked" (Berekely's tactic).. in other words the accounts will appear accept WUs but stats will not change & new WUs will not download... hence the cheater losses what they obviously value most.. STATS...

In IceTea's case, he was dropped from our team--and any credit for the WUs he processed was subtracted from our team totals... His IP & email was logged and he is not welcome back on our team under any name, legit or not...

IMHO, patches & cheats are just not worth being booted from the project & team...

chaosdriven

09-09-01, 09:36 PM

Morpheus, I certainly agree, cheating is counterproductive.

Well, maybe the original fellow in question is running an old Cray supercomputer!!!:)

Any other possible explanations for such work unit times?

Could someone with appropriate expertise "fragment" the unit and run it on multiple machines????

Would that then be distributed-distributed processing?:p

Morpheus

09-10-01, 11:25 AM

I believe a beowulf cluster could break a unit down into parallel processes (with the right software)... so, perhaps, others might as well...

the only conforting fact is that those fast (under, say, 2 hours) WUs REALLY draw the attention of the SETI@home folks if they are at 100% complete...

derek

09-10-01, 04:32 PM

I don't think breaking up WUs to run on a cluster is allowed. I think in the faq or somewhere else on the SETI site that question is addressed. IIRC, the answer is that to do it you would have to hack the code to split up the WU into pieces that could be distributed to the nodes and that isn't allowed.

As we have been seeing the server is having trouble keeping up with the speed at which regular machines are doing the WUs, they last thing they would want is someone one figuring out a way to do them in <1 hour.

At the rate cpu speeds are increasing I imagine it is only a matter of time before Berkeley releases another version of the Seti@Home client which has additional code to slow the WUs down. Much like the shift to version 3.

Sincerely,

Derek