Message boards :
Number crunching :
-9 still validating
Message board moderation
Author | Message |
---|---|
Bill G Send message Joined: 1 Jun 01 Posts: 1282 Credit: 187,688,550 RAC: 182 |
I think mine is good but the two that validated are wrong. Ongoing problem. http://setiathome.berkeley.edu/workunit.php?wuid=1802396926 SETI@home classic workunits 4,019 SETI@home classic CPU time 34,348 hours |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
|
Rasputin42 Send message Joined: 25 Jul 08 Posts: 412 Credit: 5,834,661 RAC: 0 |
I just produced a few -9 results. I just changed the number of tasks per GPU from 2 to 1, and in no time flat,produced about 40 faulty results. The problem is, that it goes very fast and instead of being marked as "invalid", they become "validation inconclusive" and causing more trouble. I was using cuda32 from lunatics(ver0.41). No change to driver. (i think, the -9 results should be marked "INVALID", straight away! and, as these are computed very quickly, the loss of computing time is minimal) |
Darth Beaver Send message Joined: 20 Aug 99 Posts: 6728 Credit: 21,443,075 RAC: 3 |
Rasputin42 look at the other thread http://setiathome.berkeley.edu/forum_thread.php?id=77395#1686856 wow that must have been very slow you only have 512k ram on that card mate i would never try and run 2 units on a card like that take to long my 9400 gt had 1 gig of ram and was able to do 2 but very slow .She must be a old Lappy your using there . |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I just produced a few -9 results. There is nothing inherently invalid about an overflow result. SETI@Home Informational message -9 result_overflow This can happen if a satellite happen to cross over the path of the reviver or something similar occurred. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Rasputin42 Send message Joined: 25 Jul 08 Posts: 412 Credit: 5,834,661 RAC: 0 |
But there must be a way to identify these tasks. This all seems to be a fairly recent phenomenon, so what changed and when? Yes, she is an old lappi, but still going with no cd drive and no keyboard. I know about the other thread, thanks. |
Darth Beaver Send message Joined: 20 Aug 99 Posts: 6728 Credit: 21,443,075 RAC: 3 |
Yes, she is an old lappi, but still going with no cd drive and no keyboard. wow no keyboard or cd rom . I thought i've used some crappy setups but good on you for getting it to work and still crunching |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
But there must be a way to identify these tasks. It isn't recent & has been ongoing for MANY years. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
I'd guess there's a lot of info hiding in the numbers, with 'Genuine' overflows meant to account for about 5% of multibeam results. Without having been present for such decisions, I believe the signal thresholds are set to yield around that, so that statistically 95% of the runs poke above the system noise floor, that noise floor including all the messy radar noise, Zombie or broken crunchers, cranky immature GPU applications and all manner of legitimate noise. Reflecting that, I seem to see about a 1-2% inconclusive/pending ratio on a healthy system, with no invalids, no errors other than the occasional ghost task marked as abandoned since resend lost tasks has been off. I think that rate of inconclusive is about 5 x better (used to be around 10% for v6 multibeam, less mature applications involved) than it used to be on a task by task basis... But then the GPU is definitely greater than 5x the throughput of my old 9600GSO, so probably the volume of inconclusives is greater. Multiplied across all users that could easily look like an expanding problem, rather than simple growth. I'm guessing the lines are a bit blurrier than that of course. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Rasputin42 Send message Joined: 25 Jul 08 Posts: 412 Credit: 5,834,661 RAC: 0 |
That is even worse, that nobody has found a fix in years. If there is so much interference to cause a -9 overflow, the result is useless anyway, so why not discard it? I very much doubt, that these are analyzed further, when there are thousands more promising signals to deal with. |
rob smith Send message Joined: 7 Mar 03 Posts: 22199 Credit: 416,307,556 RAC: 380 |
The trouble is interference comes in various forms, sometimes it is a very narrow range that only affects a few work units in a "tape", this sort is hard to detect until we are doing our filtering, others can affect a whole "tape", and that is much more obvious at the outset. Actually interference affects only a small fraction of the number of work units being processed, if my crunchers are anything to go with it is well below 1% Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
That is even worse, that nobody has found a fix in years. Well it isn't known until the data is analyzed if it is just basically noise. That is out task to preform. They set a threshold probably for several reasons. I would guess. To keep the returned results from becoming to large. Also anything over that threshold is likely not going to be the kind of signal they are looking to find. Larger results may also take longer to run in the validator. Any solution to "fix" the problem of run away GPUs & CPUs causing false overflows. By adding more checks on the server side would cause more complexity to an already complex system & take time by the project admins. Time is always in short supply for them. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
But there must be a way to identify these tasks. There is a way to do it: If a *single* result seems to have run through without -9 overflow, than every *other* result showing -9 overflow is invalid! Simple as that. This way false positive -9 overflows can only slip through, if they both are the first responses. Aloha, Uli |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.