Overclocking and corrupt data????

[13oots2]

Have been busy OCing my new hardware the past couple of days. Until yesterday, out of 202 WU's, I have NEVER returned an interesting signal.
Then yesterday out of 4WU's I had 2 interesting signals returned. Luckily my Seti folder was backed up before I stared playing (Just incase of the worst). So I lowered the bus by 1MHz, and no interesting signals were returned. Luckily I had not returned the results to Berkley.
The reason I am posting this is a thread was started about 3-6 months ago, asking how many interesting signals had Team Members returned. Some people had perhaps 10% of their results showing interesting spikes ect. I, as mentioned before have NEVER found an interesting signal in 202 WU's, with a moderate OC. Yet return two spikes at my CPU's limit, everything else seems to run fine.
Could this be a cause for concern, returning false data to the Seti server?

[JigPu]

Well, sending bad data to SETI isn't going to do them anygood. You should run SETI at stock speed (or even underclocked if you really want the right results), and then process the same WU at each level of your OC. Use a file compare program to check to make sure the two result files are completely the same, and then up your OC. Once they are diffrent, you have run SETI beyond what will retun the same results. Back it down to your last setting.

Thankfully, I don't OC this computer, so I know (or at least hoping....) all my results are good! Untill I play a MIDI or two.... :mad:

JigPu

[Thelemac]

You can also test with the Team Lambchop benchmark doo-hickey.

[TC]

It wouldn't hurt to check it out - that is see if you're getting corrupt data at that overclocked speed. I'll be the first to admit that I don't check all of my overclocked systems, but I do have a gut feeling of how far I'm pushing a particular system based on experience and comparisons. If I'm not pushing a system to the extreme I don't check, but if I plan to put a questionable overclock into service I will. Use the TLC bench file or one of your choice. Compare the results from stock versus whatever overclocked setup you're worried about. They should be exactly the same, otherwise you've got a problem.

[Mictlan]

As I have two systems I have done some other stuff to test the amount of noise my oc is introducing in the WU. I crunch a WU in my work PC (this has done over 150 WU, none returned as intersting by SETISpy). Then I backup one WU and crunch it at home (over 600 WU, one interesting spike) and compare them. To this date I have seen no so many differences.

Besides, I have gotten some high gaussians and SETISpy don't record them as intersting. I really don't understand the criteria it uses to classify them. :p

[Sir-Epix]

If your system is stable in Prime95, then there is no need to be worried. Just check it out first for stablility. If the system is not 100% stable, then it can return corrupt work units to seti.

[JigPu]

Originally posted by Mictlan
As I have two systems I have done some other stuff to test the amount of noise my oc is introducing in the WU. I crunch a WU in my work PC (this has done over 150 WU, none returned as intersting by SETISpy). Then I backup one WU and crunch it at home (over 600 WU, one interesting spike) and compare them. To this date I have seen no so many differences.

Besides, I have gotten some high gaussians and SETISpy don't record them as intersting. I really don't understand the criteria it uses to classify them. :p
Interesting gaussian....

"Returned when Fit < 2+1.6*Power"

Since SETI hasn't found a gaussian in my WU yet, I don't know where the power can be found, but I know that most of my fits are around 8....

JigPu