Computers used in the SETI project

[toolie]

I was pounding out 2 blocks every 4 hours on one of our machines at work, and 2 blocks every 8 hours with another. Those were the only two that I could use, the others were all in use.

The specs are (or were - the machines are in use full time now and I can't run SETI@home http://www.linuxnewbie.org/ubb/frown.gif):

Both Dual Processor SGI Octanes
Both have 512 Megs Ram
The fast one has 2 R12k 300Mhz CPUs
The slow (heh) one has 2 R10k 250Mhz CPUs

They added up a lot of blocks in my name during the few weeks I had the use of them http://www.linuxnewbie.org/ubb/smile.gif

[ghorton]

PII 400. Is the Unix client faster or slower than the Windows one (with graphics turned off)?

BTW, for those running in Windows--turn the graphics off (it's in the screen saver setting). My average time shows about 14 hours, but it's been closer to 10.5-11 since I did that.

When I figure out how to get files onto my P166 laptop, I may try it there too.

[Beowulf_Ghost]

I'm currently working on a Beowulf cluster and was wondering if anyone knew of a PVM or MPI version of SETI@home that I could run on it. Or would it make more sence to make a script that would run a seperate block on each node.

------------------
Silence is Foo!

[conman]

I changed from the windows setiathome version on win98 to the gnulibc21 version on the same machine and found virtually no difference in speed.

[toolie]

Beo: The best thing you can do in that case is to run a script that gets a separate block to each box. The computations aren't very parallel processing friendly. Even on the same box with a greater number of CPUs than one, you have to run separate instances for each CPU.

[Squid_boy]

Is there any way to get seti to work through a proxy with a user name and password? Our network recently changed to this system to log traffic and I can no longer use Seti in Linux or windoze.

[toolie]

SETI does work through a proxy. Use the same proxy as your web browser. I don't remember the syntax, but its in the documentation (I think the command line options are what I looked at to figure it out).

[bleg26]

I have 3 'puters crunching SETI data.
My Linux/SAMBA machine is a PII233 averageing 16hrs/block.
My Main box with RH6.1/'DOZE98 is a PII350 averageing 9hrs/block.
And a AMD K6-2-450 running 'DOZE98 averageing 14hrs/block.

I had a pentium 66 crunching data, but was averaging 3-4days/block. I took SETI off of it and only use it for IPMASQ now.

When I upgraded to version 2.03 from version 2.0 in 'DOZE98, my PII350 and AMD sped-up from 22hrs/block(PII) and 30hrs/block(AMD).


MY OPINION:
AMD processors are pretty decent, but their math co-processors suck..HEHE http://www.linuxnewbie.org/ubb/redface.gif

[This message has been edited by bleg26 (edited 31 March 2000).]

[tnordloh]

Noticed the same thing with my AMD. The pentium II 350's at work are cranking them out (in windows 95) at around 10 hours per block, while my linux 500 mHz k6-2 is doing them in about 11-12 hours. Doh.

[MGP]

Making a blanket statement like...

AMD processors are pretty decent, but their math co-processors suck..

is not really accurate. I know it's your opinion and as far as the K6-2 and to some extent the K6-III family of AMD CPU's is concerned -- it's true. The K6-2 is particularly bad, since it has no on-chip L2 cache. The K6-III makes up for some of it's FPU's poor performance by including 256K of full-speed on chip L2 cache. The K6-2 and K6-III do have superior integer performance though. Clock for clock they are faster than PIII's and even in some cases the Athlons. So running regular Linux applications or even normal Windows apps for that matter, they are good performers. I built a lot of K6-III machines for friends to use as general web surfing and office app systems and they were all impressed by the speed of them. Linux with a GUI runs great on a K6-2 or even better on a K6-III system running at 450 MHz and 128MB of RAM.

The K6-2 and K6-III family were designed by AMD to compete against the Celeron in low end systems and they did that quite well. But as everyone knows (or should know by now) the new king of the low end CPU's is the Duron. Very inexpensive and excellent performance -- basically a full blown Athlon CPU with a little less on-chip L2 cache. Anyone who has used one knows how fast these inexpensive CPU's are.

The Athlons are another story though -- they have superior FPU performance vs. the PII and PIII's. And clock for clock they are considerably faster. AMD now has high end CPU's that compete at the same level (and somewhat higher, depending on the application) as the PIII's for a considerably lower cost.

I have all AMD systems (except one older PII 450) running SETI@home now. My Linux server is a K6-III+ running at 550MHz and cranks out SETI work units in about 8 hours. The Athlon systems are mostly running Win98 and since SETI@home required everyone to switch to the newer client that does additional science calulations the work unit time have increased to about 10 hours. Before that the Athlon systems would crank out a work unit in 4-5 hours typically. Not bad FPU performance at all... http://www.linuxnewbie.org/ubb/biggrin.gif