I started running seti@home on my router box, and remembered seeing a post in the old forum about an Applefritter team.
So I joined it!
Since there's no info about it in the new forum, I thought I'd post a note to current and new 'Fritter users that the team exists.
Team stats are here:
http://setiathome2.ssl.berkeley.edu/stats/team/team_138701.html
I still 0wn the average-time-per-workunit stat. ;^>
--Peace
(And am in thrid place!)
PS-Eudi, with what machine are you achieving that? :o
My guess is that Eudi is using more than 1 machine to get that speedy rate.
I'm slightly under 9 hours per unit, but I'm only using one computer - an Athlon XP 2500+ running XP Pro with 1GB RAM.
Actually, running multiple simultanious sessions doesn't help you. It averages execution time per work unit on an individual basis. (So if you have four computers which can each do a work unit in four hours running at the same time your average will still be four hours, even though you're returning a work unit an hour aggregate.)
My secret is the only thing I ever use SETI for is for stress-testing servers. The slowest machine which contributed to those scores was a 1.33Ghz Athlon... which can do a work unit in about six hours. I have no idea why your Athlon XP is so slow. (You're running it in the background? I use the Linux, FreeBSD, or other UNIXoid commandline client.)
The fastest thing I've used is a secret. It can do one in about two hours. I should run it some more and make my average even more frightening. ;^>
Just to note, the Applefritter server takes about 3:40 per work unit.
--Peace
That's terribly funny (and impressive!) as my router box hobbles along at 36 hr 13 min 15.0 sec average per unit on its 233 MHz PI.
If this were a race, I'd be the turtle.:)
Strange, I've got a 366 Celeron with 128Mb of RAM running XP that takes over 100 hrs. to do a unit...
[In any case, I'm now up to second place! (I know that it's not a race, but I'm a hopeless dork who is too useless to compete in anything else, so this makes me feel kind of good
)]
I joined.
I have no idea what I am doing, i just left everything at defaults. Should I install this on all my computers using the same account? (i have about 7 that are on 24/7).
i think ill join sometime. I Got a 400mhz celeron, a 1.8ghz Duron, and a PIII 933 that could run.
A real performance killer on low end computers is leaving the screen saver animation on, that it to not go to "black Screen" in the seti options. On a PM6100/60 the screendrawing took 80% of the CPU power or something like that :o
Yep! Use the same account, that way all completed units credit one account.
One reason this machine is slower is that it's used for development during the day. It's also running the screensaver version, but is set to analyze all of the time. Also after the Athlon 1400 was released, AMD switched to the 'model number' naming convention. The Athlon XP+ 2500 actually is clocked at 1830mhz.
Depending on what I'm doing during the day, development can bring this machine to a crawl.....
The screen saver animation was still on, so I just disabled that. Currently it's lowering the average time slightly (closer to 8hrs than 9hrs). I'll leave it running and see how it goes.
Update/Note:
To give you an idea of what I run during the day:
- Visual Basic 6 Professional (sometimes 2 or 3 instances)
- Access 97
- ACT
- Virtual PC 2004 emulating Windows XP Professional and/or Windows 2000 or both
- Mozilla
Most of these are run at the same time.
Joined!
Is there a log file to see how long it has taken to complete past units? It's hard to see how long the last unit took if the machine is at the office and I'm not.
I'm 9th! woot!
mobilepet has you beat by two minutes...ya better step it up!
I joined with one of my systems. I've got a few more that might make it into the lineup, one of them however doesn't have a network connection (anyone got a 802.11x card or an Ethernet card that'd work in a 1400?) Hopefully we'll spur some interest from others as well...
- Macinjosh
It's tempting, but... I hate wasting electricity on it when I don't have a good reason.
Maybe I'll get a chance to test it for a few days on one of the 64 bit Xeon servers we're just starting to get at work...
--Peace
I'm getting tempted to install it on the 15 G5 machines in the journalism computer lab at work...
...which leads to a question. Those lab computers, when not being used, sit at the login screen. Is there any way I can have the SETI software run while at the login screen, but pause while someone is logged in?
I'm still a bit confused, what exactly is meant by these:
Results received \ Total CPU time \ Average CPU time per work unit
What i mean is what kind of scores are best for each of those values?
I'm assuming that the more results the better, but what about total cpu time vs average cpu time per unit, whats better? Higher or lower amounts?
Seti downloads work units of data. Each unit is a portion of a radio telescope scan. Once complete, seti uploads that unit back, and records it as a result received for your account.
So...the greater number of results received, the more units you've analyzed. The higher average CPU time, the faster you've analyzed each unit. Total CPU time is the length of time seti has been running.
I've been signed up for a little bit now... the only two machines running it constantly are my Compaq (667 Pentium III, 128 MB RAM) and one of my blue G3s (300 Mhz, 384 MB RAM). Sometimes I have another G3 (almost same specs) running it.
Part of the reason that your 366 is taking so long is that 128MB is the minimum for Windows XP. 256MB is recommended. It uses swap space a lot with 128MB. I would recommend going into System Settings, then Advanced, and change the Performance Settings to Best Performance to help with the sluggishness.
While I have managed to install XP to a P2-233, I definitely would not like to use it on a daily basis.
Those running seti on non-Classic OS systems might want to look at the command-line only version of Seti@home.
In a completely unscientific study by yours truly, it took about three hours to hit 19% completed on a Win98 pII/400 stripped to the bare OS using the GUI version. With the command-line version it took just over an hour to reach about 19% (checked with Seti-Spy).
I just added a 7300/7600/8500 hybrid to the SETI cause. It's still using its original 180mhz 604e processor. Should take a few days.
I was going to try to set up a 6100/66 with a 240mhz NuPowr G3 upgrade, but forgot to put the hard disk in it this morning.
Update: Hard disk is in, MacOS 8.1 is installed, and it trounces the hybrid. Guestimated time is 26 hours per unit.
I've finally got my primary machine at the office crunching data in the background. Running the graphical client for XP it's managing 1 unit in about 2 hours - far faster than the 20 some hours it takes on my G4. So now to work out how many work units I need to do before I smash Eudi's average unit record...
Snicker.
I gave into temptation and fired up my "secret weapon" to churn out some more units and lower my average, it being the fastest machine I have at home. Unfortunately, it's been about six months or so, so the state of the art has sort of caught up with it. ;^b
(There's still a few tricks I could try to optimize things, depending on how lazy I am. If I really wanted to go "faster" I'd only run one instance of SETI at once, since the OS I'm using seems to has substantial SMP overhead.)
Nothing like wasting a few thousand BTUs in the name of vanity, I guess. ;^> Feh.
--Peace
I'm still confused as to the stats that you guys are comparing and who is winning but I'm gonna fireup my boss's new server that he got for our lab hopefully later today.
its a dual p4 xeon system with 2 gigs of ram, a whole bunch of drives in a raid array, blah blah blah. Its brand new but i can't remember the speed of the cpu's.
In regards to the stats that you guys are discussing, is it actually better to not put the seti@home on all your computers but instead only put it on your fastest machines?
currently i have it on 4 machines at home, the slowest being a dual pentium 3 600/650 (can't remeber which) and the fastest which is a p4 2.8 ghz but is that dragging my average time thingy?
The stat we've been yakking about is the "average time per workunit" one. And that one is all about how fast your fastest computer is. Running on multiple machines will drag it down to the average speed of all your machines combined.
I.E., if you have five machines, capable of doing WUs in 2, 4, 8, 12, and 14 hours respectively working at once, your average is going to be eight hours. You'll be returning... let's see... about 8.2 units per eight hours, which means your *agregate thoroughput* of about .95H/wu is much better then what your fastest machine can do by itself, but the average time per workunit stat isn't going to reflect that.
Just to note, it's fairly futile to try to keep the top "average time" slot it if you've been a member of SETI for any length of time. A machine which could pound out a WU in six hours was pretty impressive three years ago, but isn't very hot now. But all the units done on that machine will be averaged against your current ones. Which gives someone who's just started now a real advantage, since they don't have to "work off" the (by today's standards) slow units.
The stat SETI wants you to care about is "total work units", as that measures the total work you've done for the project. And that's the one that people run SETI on whole herds of machines for.
--Peace
The contest seems to be who can get the fastest per unit time. I know I won't win
Running it on slower computers will increase the average time per unit rating, so it you're going to make it your goal to keep that score low as possible, only run it on your fastest cpu.
Think of all your computers as a kids sports team. All the top kids can do their best, but that one awkward kid can bring down the whole team's ability to win (OOPS...eeun's metaphor quickly undoes 30 years of inclusive feel-good sportsmanship)
Me, I'm just going for maximum units completed over the long haul because I've no hope of crunching units at the same speeds as Eudi and Mobilepet. I've got it installed on the router, running 24/7; on my Athlon, running when I don't need the cpu cycles; and sneakily installed on a 'dead' PII somewhere that isn't at my home...
Does anyone here know how to get seti at home running on windows xp pro boxes that are joined to a domain and are idle? By that I mean that I want seti at home to run when the students have logged off the lab computers without effecting them when the students or faculty logon to the computers.
Also at this point, i really think that the lowest average time per unit score is a waste of time and goes against the spirit of seti.
By that I mean that seti wants as many units completed. They don't want the lowest number of time to complete a work unit, they want as many work units done as possible. So if someone is gunning for the average lowest time, they will not install it on their other slower computers at the sake of total work units. That to me doesn't sound right. I am installing it on as many comptuers as possible to help the seti team, not go for some sort of drag race which looks good to the few but doesn't necessarly help the team as much as if you were to install seti on that fast comptuer and on other slower computers that you might have at your disposal.
No offense intended, thats just how i see it.
If I am wrong, please correct me.
[quote=Eudi]I gave into temptation and fired up my “secret weapon
I don't see the 'drag race' and the goal of seti to be divergent at all. Those on the 'fritter team who are crunching units faster are also the ones who are quickly racking up the most units completed. Also, as was pointed out somewhere above, not all of us have access to labs and offices to do multiple installs, or wish to take on the expense of running multiple home machines just for seti. *any* units completed benefit seti, so if it takes some competition to get the units in, good for competition!
Think of the time per unit contest as a goal within a goal; a chance for some of us to engage in a little friendly competition in the name of a good cause.
My $0.02, anyway.
BTW, I see some names signed up on the Applefritter list who have yet to complete a single unit.
Getting back to that disruptive sports metaphor I made above,
Don't make me talk like an obnoxious coach to you, ladies...;)
I'd be careful and ask those in command before proceeding too far. Some organizations have specific fiats against running things like SETI.
(At my company we'd murder someone if they fired it up and left it running on the engineering machines, for instance.)
It does cost a non-trivial amount of power, if nothing else. (An Intel Pentium IV only draws a few watts executing an idle loop, but can suck over sixty running full out. That turns a nightlight into a reading lamp, essentially.)
You know, in theory I agreee with you? I just have to take my jollies where I can find them. ;^>
(Power bills are prohibitive where I live, so I'm sort of opposed to running computers *just* to process SETI. I do have sporadic opportunities to stress-test fast machines, however. So essentially it comes down to the "time-per-work unit" being one area where I can legitimately "compete". Yes, it isn't a race, but... we have to have a little fun, don't we?)
It's loud, it's ugly, and it doesn't execute x86 code. I hate to be more specific then that. ;^>
--Peace
Well in regards to the lab machines, i am the admin for these labs and they run 24/7 and they are at my descretion. The big thing is that I don't want to effect the students who logon. I want them to walk over, move the mouse (turning off the screensaver) and then login and do their thing.
So i just want to run seti while it is at the login screensaver. currently all that is running is that default windwos xp screensaver, the one that is nothing more than the winxp logo moving around the screen.
How long did a unit take on the 6100/60?
I also have a 'secret weapon,' but I need to get it operating again. Of course it has dual cpus, all of that other good stuff. I can't wait to see how it does. I'll repost on that in a few days. Even tho it's not an apple, it's fun to have it on the team.
I'ma gonna guess that those "secret weapons" run multiples of either Sparc or MIPS CPUs. That being the case, I'm guessing that they're either Suns (or clones) or SGIs. Which models? I dunno. Maybe some Alphas? There were some hefty SMP Alphas around, right?
Yeah, I set up my dual Opteron 246 to crank out some units. It has 6gb of ram. Later..
Actually, the RISC CPUs in my WMD (Weapon of Mass Deduction) have a sparkling minty Blue flavor.
Anyway. It appears I'm going to have to concede the speed contest to Mobilepet, alas. (Using the standard SETI client for my hardware I'm averaging around 2.6 hours per work unit, which is fast, but not quite Opteron fast.)
Feh. I still win Miss Congeniality for fastest non-x86 in the running. ;^>
--Peace
I should really get my Xbox up and doing this, at least for a little while.
A P3 733 with 64MB RAM wouldn't crank out too bad an average, especially running the command line version under 'nix, and anything is better than my current score.
Besides, while all of you have your duallies and quads and workstations and servers and impressive expensive machines of all shapes and sizes and weights, I'd be able to say that I do SETI on a console. A cheap, inexpensive, ghetto(compared to the duallies and the quads and the opterons and the RISCs) debian-running Grand Theft Auto playing console hooked up to a TV along with a DVD player and an amp. Who can say that, huh? WHO?!
I will be impressed if they make a client for the atari 2600.
Could you imagine just how long it would take to crank out one unit on an atari 2600?? wow.
(seriously, i like the idea of running seti on the xbox)
If it were, I would run it on my quad-processor Pentium Pro Digital server. Since it doesn't, there's not much point to leave the machine running - it's rather loud when the fans are going top speed.
The Xbox does sound like a good idea - it would beat 2 of the 3 machines I'm using!
Wait, you are telling me that having duallys or quads is NOT going to allow seti to run faster?
So how are all these monster machines really cranking out low times?
They have multiple freakishly fast CPUs. ;^>
(The extra CPUs let you run multiple processes, of course. Which is good for accumulating total WUs, but would slow you down a bit going for the time-per-WU title.)
--Peace
As Eudimorphodon says, you can run a SETI process on each processor in the computer, but multiple CPUs can not work on the same unit. The other ones mentioned on this topic are much faster than my quad - mine are only 200mhz processors. Seeing how well my 6100 G3/240 and the hybrid are doing (about 30 hours and 50+), I'd rather not try it on the server for the amount of noise it makes.
Posting from the Xbox-- I've got 'er primed and ready for crunching.
Edit: apt-get is being naughty, and I'm not in the mood to compile anything right now. But mark my words, this Xbox WILL be crunching numbers for the good of mankind (and my personal bragging rights) by the end of the week!
Mmm, Big Blue, eh? I'd guess some nice iSeries or tasty pSeries, hmm?
Is there something wrong with the fritter group? I click on the link and it brings up only like 3 lines of data with no black backgr4ound or anything.
The script that creates the page plays up every so often. The stats should be there next time the script runs in about 3 hours.
well its been around 20-24 hours now.
Err.. I guess i will just be patient and wait.
...and woo, Mr. bobotech, last time I looked you were under 50 units, and now suddenly you're over 100!
There goes my dream of getting out of 12th on the list. :cry:
Pages