I'd lik to see what averae WU time people are getting on this list who use G4/400's. Since reinstalling my system and upgrading to Seti 3.03 my time went from 13-16 to just under 30hours per unit. Is Seti downloading a bigger chunk of data or is it my machine.

Kevin

[gumby5647]

it is neither my friend. the culprit is 3.0.3
This version deliberatly takes long to analyis units. This all is traced back to the problem of limited bandwith of the SETI servers. I guess they hope that they will lighten up the bandwith by making computers take longer to crunch the unit........
it sure does suck though......
my little DV SE 400 went from right about 9-11 hours to 17-18 hours.......
grrrrrrrrrr

GUMBY

I have 9 older Macs and a couple of vintage Pentiums crunching numbers (all on version 3.0). Recently I switched one of these older Macs from 3.0 to 3.0.3 and the average on that machine went from 21 hours to over 50. I average 32 hours per block now on all these vintage machines and I'll be damned if I am gunna sit around and babysit them while they take 70-100 hours per block. When the folks at SETI force us to go to 3.0.3, I will switch all these machines to RC-5 or some other "search".

[Scotttheking]

Hey, anyone reading this thread who doesn't use SETI, sign up for it and join our team.

Oh, my G3-400 takes about 18 hours per unit using 3.03.
Maybe it's the ram?
I gawt 384MB of the stuff.

SETI 3.03 does more scientific calculations on the data than previous versions did. All platforms- Mac, PC and Linux- complete each work unit slower than before. It takes longer to do the additional science.
One posting said he would switch to a different computing project, like the RC-5 project, because SETI was slower. Why??!! Because SETI does more science per work-unit?? Because you can't watch your totals rise as fast as before?? Well neither can anyone else. So don't be a baby...
If you want to run SETI because you want to find ET, than it shouldn't matter how long each work-unit takes- everyone's is now "slower" and Macs are still faster than PCs. If you just want to watch your numbers go up, than write a program to count from 1 to 1000 each day and count to your hearts content.
My blue & white G3/450 completes work units as fast as my Pentium-III 1GHz!

[greenerx]

Actualy I believe the added is time is built-in the new software, not to slowdown bandwith growth but added search algorythms. Check it out: http://setiathome.ssl.berkeley.edu/version303.html

------------------
"Ca changera pas l'etat
de l'univers."

I noticed slow-downs, too, but recommend two simple workarounds that should dramatically speed up your processing: 1.) set your screensaver to go to a blank screen after 1 minute, 2.) when running S@H in the background, minimize/collapse the display window (i.e. windowshade it). Apparently the graphics eat up a lot of processing time, and disabling them to the extent you can drastically improves your CPU time. Before taking these steps, it was taking my G4/400 20 hours or more; since, I'm down to the low-to-mid-teens! (I've got 256 MB RAM, FYI.)

It has little to do with "watching numbers go up". It has to do with a client version that will cease to function. If Apple told me that 7.6.1 or 8.1 would simply just stop working so I couldn't use any of these older Macs or if MS told me that Win 3.11 would just stop working and I would HAVE TO upgrade the OS, I would likely look for alternatives and take these PCs to whatever OS was available. Why should I make an exception for SETI?

We still have millions of blocks to analyze. At the current rate using the current clients, it will take us several more decades to complete. At that time, we could do the extra and complete analysis from 10-50Hz/sec and from -10 to -50Hz/sec.

This client change has to do with bandwidth, and that fact that we are completing blocks faster than the fine folks at SETI can collect/record new ones for analysis.

Originally posted by ComicShop:
I have 9 older Macs and a couple of vintage Pentiums crunching numbers (all on version 3.0).

How about shutting down the machines if they're not being used? It's one thing to run SAH in the background if the computers have to be on, but it's unbelievabe that people run their computers 24/7 just to crunch SAH blocks.

Before 3.03 I was getting 11 hrs per packet on avaerage with my wife's iMac 350mhz/190mb. Now I get 20 hrs per packet with 3.03. I believe those of you with G4s should be getting about 15 hrs or less. I run my setting in screensaver mode and blank screen after 1 minute.

I'd be interested in knowing what a P3 1ghz Win2k box does with 3.03. Maybe an equivalent Athlon. Also, anyone with one of those dual processor Macs running OS X PB please post your average times.

I notice SETI AT HOME breaks down the Windows OS, but not the Mac OS.

BTW, this iMac does a packet in OS X PB on average 22 hrs, but the latest version is 3.0.

Guys & Gals,

I work closely with the SETI@home team (wrote much of the material on the site...) and I think you may have missed the point of the SETI@home project. It is not a race to see who can complete the most work-units, it is a scientific search for signals from extraterrestrial sources. The latest version of SETI@home does quite a bit more calculation in order to refine the search to a higher degree, and in general will take almost twice the time per work-unit. The SETI@home project can do this because all of you have been so generous with your computer time. Many more people have signed up to participate than was ever dreamed of. This has allowed SETI@home to do an even deeper analysis than would have been possible with fewer participants. It also utilizes the resources provided by you in a better fashion. Work units can only be generated at a fixed rate. With version 3.0 and earlier, the huge SETI@home user base was crunching units faster than they could be generated. Rather than send out a work unit multiple times to multiple users, 3.03 spends more time analysing each work unit, reducing duplicate analyses. As a side benefit, yes, it does reduce the bandwidth that the project uses. (BTW, SETI@home is now using 30% of UC Berkeley's bandwidth! That's very expensive!) The reduction of bandwidth was not, however, the main purpose of the 3.03 update. The purpose is to do better science and hopefully increase our chance of detecting life out there. That's what we all want, isn't it?

Ron Hipschman

[TylerR]

It has little to do with "watching numbers go up". It has to do with a client version that will cease to function. If Apple told me that 7.6.1 or 8.1 would simply just stop working so I couldn't use any of these older Macs or if MS told me that Win 3.11 would just stop working and I would HAVE TO upgrade the OS, I would likely look for alternatives and take these PCs to whatever OS was available. Why should I make an exception for SETI?

Because it's not an OS! This is an analysis software of data that YOU VOLUNTEERED TO DO! It is (suppose) to be a noble aid in the search for a possibly alien-intelligence generated radio signal from space. What data they wish to collect is completely up to them. They had their deadline extended and are currently looking into starting to crunch radio packets from the other hemisphere. This is not an ego driver to see who has the most units crunched with the lowest per unit time, it's a scientific search for extra-terrestrial intelligent life. This is becoming fairly ironic considering the lifeforms we have doing the searching, if this 'logic' is any indication...

This client change has to do with bandwidth, and that fact that we are completing blocks faster than the fine folks at SETI can collect/record new ones for analysis.

If your attitude is not one of trying to help the scientific community and even the whole of humanity, but of whining how you can't upload data units as quickly as you'd like because they want more information, please, LEAVE THE PROGRAM! Anyone who bitches about the increase in time is selfish and uncaring about the true goal of SETI.

I'm glad my units take 50% longet than they use to because it means the project is getting 50% more information from me about that packet. They are still getting the same amount of data crunching from me. Maybe all this proves is the intelligent life SETI is looking for is ONLY extra-terrestrial...

I'm starting to get really disgusted with humanity as a whole...

Tyler.

[This message has been edited by TylerR (edited 01-20-2001).]

Let me disgust you some more Tyler.
:-)

I'm doing it for the money.
As soon as popular power came out for the mac I pulled SETI off of my computer and installed Popular Power. http://www.popularpower.com
Hopefully they start paying soon. :-)

For MaxOS X PB users, you can get the apple-darwin-1.2 version running in back ground, it's a text version (thus faster). The graphic version of OS X was taking 24/25h to complete on my iBook 366, and I could'nt do anything else with it. With the text version it takes the same time, but I'm playing arround with the computer (I use about 60% of the processor).

Think about it.. it's a valuable option... and you have it running 24/7 no matter what

[druber]

seti does do scientific work, but seti users don't. i leave the computer on. that's my scientific effort (for the seti project, anyway). it should come as no surprise that people become more interested in the number of units they can crunch than in a scientific process they don't personally work towards and haven't seen any positive results from. don't tell the competitors that they're attitudes need adjusting. smile and nod, and chuckle that they're helping your search.

Wow, didn't mean to start a debate. Just came back to MacNN and saw "Seti slow down" threads link and saw that it was my post and an avalanche afterward.

I was just trying to find out what the average completion time on G4/400's was for seti WU's. And find out of the WU sizes have changed since I do not have anything to compare to on my new squeaky clean hard drive. I was starting to wonder about my systems processor or system bus. At just under 500MB Ram I knew that was not the problem. I alot about 30 to Seti.

I do just let seti work in the background and give my unused cycle time to the project. I also use the stats to judge my systems performance and make sure nothing is out of balance. You know memory leaks, bad extensions, fragmented memory or drive, overall system performance.


But, yes it does make me wonder why my 8600/200 604e , later upgraded to a G3/400 and now my G4/400 all out perform PII's and PIII's, without any tweaks to seti, in processing WU's. Before putting a new slave drive for OS9 and 10 in my system and downloading the new seti client I got 8 hour units on up to 18 max. This lends alot to the argument that processors cannot be truely compared across different architectures using clockspeed as a yrad stick.

Geuss I need to read the docs concerning updates more, the new processes help explain the slow down, plus I found my scheduledHD indexing hogs CPU cycles.

Sorry for the long read

Come on, people!

It plainly says on SETI's site that the new version "does more science" to each WU than the older versions did. That is why it is "slower."

If you are unhappy with your machine's performance, either get a faster machine or drop out of the project. Obviously, anyone considering doing either just wants their name on a list and scientific progress be damned.

The whole point of the Seti@home project is to use the total computing power of the Internet to analyze radio signals for a coherent signal from an alien race, if one even exists.

It is not a competition for who analyzes the most WUs, but who analyzes THE work unit that confirms the existence of an intelligent race outside the Solar System.

No one will win a prize for the most WUs processed but didn't find anything. That is not the point of the project, yet many people have desired to make that the focus of their efforts.

It is worth noting, that the major speed up of Mac versus Wintel is not the CPU itself, but the L2 cache. If your Mac uses software that allows you to vary the use of the L2 cache (XLR8 can choose none, 256k 512k or 1 meg) you can see that (especially with 60 ns RAM) L2 cache size is the big factor. set your L2 cache to pentium sized and you get pentium speed. But the entire program/data fits in a 1 meg cache so never runs at motherboard speeds.

[Kozmik]

L2 cache size can only be set if your cache already is one of the larger sizes. You can't up the size of your cache to 1MB if it's only a 512KB cache. You can, however, adjust the speed. Hmm... I wonder when L3 cache modifiers are going to be available for the new towers with G4e chips...

[mrfcobf]

I've got no problem with the longer processing times, but do have a question:

What happens to all of the work units we've been doing for all this time? Do we "re"crunch them with the deeper search??

Was all this time for nothing now, and the units turned in all have to be reprocessed?

I know it happens, but you'd think maybe there would be a client version then that would process the older units using *just* the missing parts of the process rather than reprocess them again both ways...

Just a thought. I'll keep crunching in my background regardless...

[reader50]

It takes about six months for SETI to get one complete sweep of the sky, and additional sweeps to return to any interesting locations to see if a possible signal is still present. We have a fairly long period of results with the earlier clients, and almost none with the 3.x clients. For now any post analysis has to be done almost entirely with the 2.x results and earlier.

In time 3.x results will build up. If we keep crunching faster than new units come in, I assume that Berkeley will reissue older units. 1.x units would likely benefit the most from reanalysis.

On my G4@400 with MacOS 9.1, 192MB ram and Seti 3.0.3 a unit takes 15 hours to finish. With Seti 3.0 it needed from 10 to 7 hours depending on how many interesting data it found.

Indeed the pupose of SetiatHome is to do number crunching for an underfunded science project. Since no one is paid for their processor time, the only satisfaction many users get is indeed seeing the work units numbers go up, and if this is removed (as by 3.03), some users may drop out. SetiatHome could make some attempt to balance things out weighing the increased data crunching against decreasing users and seeing if they are ahead of the game. Personally, my units have gone to 15+ hours each, and as a result I now run seti only when the G4 is on and actually idle, as opposed to leaving it on at all times.

[Scotttheking]

Hey, to the person who posted that popularpower thing, thanks.
I'm not taking SETI off any of my machines, but I'm gonna run it on my linux machine along with SETI.
That way I can pay for the added electricity costs that I need for SETI.

[druber]

well, poppow isn't paying anything yet, and i wouldn't literally hold your breath waiting. it probably is coming, it just isn't in sight yet. i'm keepin my eye on it, too. i'm sure we'll hear some great noise when those pennies start coming in.

[spleck]

I hope that the Seti@Home project sees how people feel about this.

I have no problem doing more science, but I do wish there was more reward. Not monetary reward, but scientific progress reward. It used to be enough to see the work units become more recent. I also used a program that would show the location of the WU on a star chart. These are things that should be part of the program. If people really knew what the program was doing, then it wouldn't be so much a question of how fast it was doing and whether they were working faster than others.

On the other hand, Seti@Home should have foreseen this and in the sake of keeping the WU processing time the same,they should have reduced the size of each WU by 50%. More but smaller transactions. This would still reduce their bandwidth, keep people happy, and also reduce the number of WUs that expire and have to be sent out again.

spleck

The 3.x SETI client does more science than the old client. All you moronic h4x0rz who want this thing to show higher numbers to make you feel like better h4x0rz, hey, what's that over there? Quake 6? Go check it out! Forget about your SETI numbers! Go get Quake 6! Fetch!

I've been running SETI@Home since nearly day one. I've got it running on
my G3/400 and my four webservers. To those of you who think that you're
going to catch up to me (3836 work units processed last time I checked)
just by running 3.0, let me just say this: TThhhhhtttt!!!!

Seriously, the slowdown is due to the fact that we're doing more science.
Period. I have to keep my computers on anyway so I'm glad that they're
doing *something* "useful" when I'm not using them. And I'm glad that the
good folks at Berkeley have found some ways to lower their bandwidth load.
Those of you trying to "beat" your friends in your clubs will have to find
some other way to see who can piss the farthest.

[Scotttheking]

Druber,
do you have any idea when they will start paying?
Their website says that people who use it now will get higher priority for higher paying stuff later.
So, to that end, I gave their program 9% of the cpu time on my linux machine.
Hopefully they will pay soon.
I just discovered I've got less cash then I though.

[druber]

probably better to shut your computer down and save electricity. i've been watching poppow for several months now, and it doesn't seem that they've gotten any closer to paying for units done. their software is still in beta, as well, though i'm not sure how much that really accounts for the no-pay status.

GJK, you think trying to quickly get through Seti units makes us h4x0r-wannabes? Oo, Quake 6? Good one, there. Go back to hotline, where ya came from. Or, if you'd actually like to contribute, register in the forums and give us some info about what Seti's accomplished so far.

[pcurtner]

Due to power constraints and general bitchiness, I think all SETI@Home members located in California should idle their machines and save some electricity.

(ducking)

I don't mind the slow down if it is being used to process more info. What I do mind is that it take about 9 hours of real time to get about 17 hours of CPU time on my G4 500 MP machine. It seems that work done on a dual processor is counted about 1.8 times. I am using OS X and S@H v 3.0. Oh well somuch for tring to twice the work with one machine.

[Todd Madson]

Under OS 9.04

Under 2.04 Seti, my G4/400/256 was averaging 5-6 hours a block.
Under 3.00 Seti, my G4/400/256 was averaging about 8 hours a block.

3.03 is going to take longer due to the added science.
Two reasons: bandwidth as well as more science.

So there.

[Todd Madson]

Oh, under OS X...WHATEVER YOU DO: DO NOT* use the graphical client.

Use the command line client or it will take FOREVER for your blocks
to complete.

That graphical display is overly processor consumptive anyway. Sheesh.

-t

[proux]

Got to get a hand on one of these dual 733Mhz when they come out,
with the command line client you can run a SETI application on each
of the processors with minimal overhead, and still not slowing the machine
when you need to use it for email or web browsing, sweeet.

[Frogger]

Anybody know how to run the OSX Darwin client at startup? Right now, I have to type "~/Documents/setiathome/setiathome" in Terminal to get the thing to run. I guess I could write a simple script and call it SAH or something, but running at startup would be great.

TIA,
Frogger

[Scotttheking]

If anyone knows how to write a script to run at startup for Red Hat Linux 6.2, I'd appreciate it.