SETI Benchmarks

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
Hi there people.

Do any of you run the SETI program? If so, I am very interested to find out how long a unit takes you to process + on what machine that was.

Cheers
 

748s

macrumors 6502a
Dec 14, 2001
686
24
Tiger Bay
between 14 and 19 hours on a g4 867 640mb ram. average times seem to be 17 to 19 hours. a g3 400 (firewire) powerbook i had was consistent on 29 hours. seti osx on the 867. seti os9 was on the powerbook.

[Edited by 748s on 12-14-2001 at 07:33 AM]
 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
Are you sure? I know somebody that processes unit on a G4 433 I think, it only takes about 11 hours!!
 

menoinjun

macrumors 6502a
Jul 7, 2001
567
0
My average is 16 hours on a dual 533/512/10.1. I find that slow considering how many people can do it in 10hours or less, but whatever. They might be set up differently.

-pete
 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
There seems to be a huge difference in times! Do you set set it up to process data with a blank screen + only when the screensaver is on? This should not make a huge difference, especially not nearly 6 hours extra! I'm lost.
 

748s

macrumors 6502a
Dec 14, 2001
686
24
Tiger Bay
i usually have it run when i'm not working but doing stuff like surfing, downloading etc. so that may explain the difference. next unit i get i will run overnight to see what sort of time it takes.
 

dantec

macrumors 6502a
Nov 6, 2001
605
0
California
I used to use SETI, but I found it horrifcly slow. It has been working for 9hrs & 19 minutes and has only done 48 percent. I don't know how some guy managed to push our 24 packets in a weekend!?! I also have a G4 867 but I only have 384 Ram.
 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
SETI will not increase in speed once the RAM increases over a certain point (probably 256MB). I am really supprised that it takes everybody so long to process a unit.
 

748s

macrumors 6502a
Dec 14, 2001
686
24
Tiger Bay
it use to take around 11 hours in phase 1. earlier this year they started sending out phase 2 units that took the powerbook around 28 hours instead of 11 hours. an SGI onyx would do 24 units on a weekend. think they do a phase 2 unit in around an hour.
 

Ensign Paris

macrumors 68000
Nov 4, 2001
1,781
0
Europe
Dual 800

I have a dual-800/ which processes about 11 packets in a weekend. Running it in the Terminal Mode is much faster that running it as an carbon app.

Guy

[Edited by Ensign Paris on 12-30-2001 at 07:02 PM]
 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
Ok, question. SETI give us a CPU time, but where does it tell us the actual time it takes us to process a unit?
 

Ctenactis

macrumors newbie
Dec 9, 2001
4
0
CPU time

In my limited experience cPU time has little effect on 'real' time. If you are running other apps the CPU time remains average although this average can vary by as much as 50%.

I run on mainly 233 G3 machines and average an 24hours CPU time.
 

pimentoLoaf

macrumors 68000
Dec 30, 2001
1,976
1
The SimCity Deli
iBook 500 w/256mb RAM takes 20-hours on the average to process running in the background.

Unlike the PC version, macSETI seems to not like anything else running for it to work. In the past hour, it only crunched about 1% more -- and all I was doing was reading online news and posting comments here.

Run overnight and when I'm not around, it works out to one work unit per day.
 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
As much as I like the ibook I am not very impressed at how long it takes to process a unit! Is this to do with the mac version of seti, I cant believe that the mac is slower than pc's!! Afterall, a v.good time for a desktop should be around 10 hours!
 

pimentoLoaf

macrumors 68000
Dec 30, 2001
1,976
1
The SimCity Deli
My pc (a Toshiba 4080xCDT laptop @ 366mhz & 192mb RAM) whips through a unit in about 11 hours, so I think it must be a programming thing. Perhaps all those built-in math routines in Pentiums has something to do with it, while the RISC might be difficult for some people.

Then again, isn't SETI programmed in C++, and the same code is on all machines? Might then be a question of compiler optimization.

 

Falleron

macrumors 68000
Original poster
Nov 22, 2001
1,609
0
UK
I am at the moment running a AMD650Mhz, 128Mb RAM + it takes me 15 hours to process a unit! I'll be whipping through those units as soon as I get a mac!
 

pimentoLoaf

macrumors 68000
Dec 30, 2001
1,976
1
The SimCity Deli
Yup -- same version.

My Toshiba's p2 is a "mobile variant" which I am led to believe is faster than a normal pII. Waaaaaaay too many p2's and p3's are Celerons, which have limited math routines... I think.

Any pc hardware folks out there know what I'm talking about?

 

pimentoLoaf

macrumors 68000
Dec 30, 2001
1,976
1
The SimCity Deli
Of course, it also depends on the WU (work unit).

Those without gaussians (bell-curve signals) are obviously going to go through processing superfast, and those with loud noise are going to be processed even faster, often within a few minutes before getting another WU.

Remember: it's doing all these calculations over a range of doppler shift values, and all that math takes time. Those running the program as a strict screen-saver will wait an eternity for just one WU to process, even though the total processor time for one will amount to a few hours. Running it overnight as a standalone program is the only way to go.

Which, BTW, is how I'm up to 777 WU's processed -- since the program started back in '99.
 

greatm31

macrumors member
Oct 19, 2001
69
27
well i've got a pc with a p3-500. it chunks down units in around 16 hours, with seti v. 3.03.
 

pimentoLoaf

macrumors 68000
Dec 30, 2001
1,976
1
The SimCity Deli
You know...

The way PDA's are evolving, that 9600 will be less advanced in a couple of years to a handheld, and to see SETI crunching away on something that small...

:eek: