![]() |
|
|
#34 |
|
"Curtis"
Feb 2005
Riverside, CA
2×2,437 Posts |
My avatar also suddenly appeared with no explanation. I am told it is some flavor of pokemon. The admins do occasionally mess with settings-- a few weeks ago, all thread headings were in spanish, as well as the links at the top and greetings.
While BOINC would indeed get us more CPU power, it would also increase possible errors. You can find old threads here with erroneous LLR residues; the more individual computers submitting piecemeal LLR work, the greater the chance of owners of said machines not noticing their computers are not stable. I think most of us at RPS prefer to have people associated with work submitted, rather than getting the work done in the fastest manner. 6ms per iteration is still around 3 times what we would expect from your machine; I believe at 650k a timing near 1.0 to 1.2 ms/iteration is standard when not hyperthreading. Again, it's not important from a workrate perspective, so much as in indication that something in your system is not performing as you expect it to. Most likely, BOINC is somehow at a higher priority than LLR and takes 80-90% of the idle cycles. If you stop BOINC and measure LLR when it alone is running, you can discern whether your machine has a more serious problem/cycle hog. Perhaps you have a screensaver enabled, for instance. -Curtis |
|
|
|
|
|
#36 | ||
|
A Sunny Moo
Aug 2007
USA (GMT-5)
3·2,083 Posts |
Quote:
Actually, the figure for ms-per-iteration that I gave was a little off--it was really 5ms per iteration. The problem was that I measured the first time with this thread open in a browser at the same time--and I've noticed that when browsing this forum, the browser seems to use a bunch of CPU time (since it's kind of a 'heavy' page, you could say). Anyway, I looked at the measurements later on, and it was about 5 ms per iteration. Is that more like it? Also, you had mentioned about BOINC possibly introducing errors to the project; actually, you can set it to have a quorum of 2, so that two copies of each workunit are sent out, making for an automatic doublecheck. If the results don't match when they both come back, another one is sent out, and so on until the quorum of identical results is met. So, if you notice a particular host bringing back loads of bad results, you can then fire off an email to him informing him of the bad news. You could, if you want, also put a sieving app in for BOINC, in that case--using the example of Riesel Sieve again (I'm using them as an example because they work with the same type of numbers as RPS), they have both a sieve and LLR application for BOINC, with users having the ability to choose which one they want to work on. They use sr2sieve for sieving, which, I imagine, is what RPS probably uses mostly also. Thus, RPS could simply duplicate both apps (I'm sure they'd be glad to let you use their apps for your project; for example, the PrimeGrid BOINC project uses RS's LLR app for three of their subprojects), and since sieving doesn't require a perfect CPU like LLRing does, you could simply suggest that a user whose computer is returning bad LLR results switch to sieve. (Or, if he doesn't reply to email after a set period of time, switch him over.) Quote:
Last fiddled with by mdettweiler on 2007-09-06 at 01:49 Reason: I forgot to mention something originally. |
||
|
|
|
|
|
#37 | |
|
May 2007
Kansas; USA
242568 Posts |
Quote:
Gary |
|
|
|
|
|
|
#38 |
|
A Sunny Moo
Aug 2007
USA (GMT-5)
141518 Posts |
I kind of like mine too, so I think I'll leave it...but yeah, that is odd. I'm actually quite surprised that it's as widespread as it appears to be.
|
|
|
|
|
|
#39 |
|
A Sunny Moo
Aug 2007
USA (GMT-5)
624910 Posts |
Quite surprisingly, when I turned on my computer and LLR today, I noticed that I was getting iteration times of roughly 2-2.5 ms! Now that's more like what VBCurtis said I should be getting (doubled to account for hyperthreading, of course).
|
|
|
|
|
|
#41 |
|
Oct 2006
22×5×13 Posts |
When I got mine, it was a bookworm flipping through a, well a book...
I got that within a couple months of joining, just to see what it would be (I'd heard of other people getting theirs randomly too). I changed mine soon after to what it is now though... ![]() Roger |
|
|
|
|
|
#42 |
|
Nov 2003
1110001001102 Posts |
630-650 completed by B'maxx, no primes. Taking 655-665.
|
|
|
|
|
|
#43 |
|
A Sunny Moo
Aug 2007
USA (GMT-5)
11000011010012 Posts |
Status update: Currently at about 80% (counting the numbers tested, not the progress of an individual number). Estimated time to completion: some time this week, although I have been known to be embarrassingly wrong when I've tried to make completion estimates on other projects, so you can sort of take that with a grain of salt.
|
|
|
|
|
|
#44 |
|
A Sunny Moo
Aug 2007
USA (GMT-5)
11000011010012 Posts |
I've completed my range of 650,000-655,000. The lresults.txt file has been mailed to Kosmaj; no primes were found, unfortunately.
|
|
|
|