Forums | MacLife

Amount of RAM doesn't play any part in SETI performance. Only raw FSB (in which the Mac lacks) and processor cycles matter. The G4/5 are not optimised to take full advantage of Altivec or 64-bit processing; and there's nothing we can do about it.

The PC, on the other hand, boasts at least 533 MHz Quad-pumped Front Side Bus and over 2 GHz of processing cycles a second. With those sort of numbers, a typical 2GHz P4 can churn out a unit in 3.5-4 hours. As for megahertz efficiency, a 800MHz G3 will certainly outperform a 800MHz P3 by quite a lot, but its not enough to overthrow the mighty P4.

May I suggest running 2 clients simultaneously? I don't have OS X to try out, but I believe guides are available at www.macaddict.com/seti and www.macaddict.com/setiex . Running 2 clients maximizes your CPU usage, because Seti@Home program spends most of its time writing to the hard disk, and that CPU time can be used efficiently by a 2nd client.

Should I still run a second client if I only have 1 processor?

Amount of RAM doesn't play any part in SETI performance. Only raw FSB (in which the Mac lacks) and processor cycles matter. The G4/5 are not optimised to take full advantage of Altivec or 64-bit processing; and there's nothing we can do about it.

This is true, as long as you have at least 16 meg available.  The large L3 cache on some Macs does help (depending on what else you're also running).  For example, my dual 1 GHz Quicksilver has 2x 2 meg L3, so the slow FSB doesn't hurt as much as it otherwise would.  Depending on what else I'm doing on the computer, this Mac turns in around six WUs per day, or around eight hours per WU (per CPU).

May I suggest running 2 clients simultaneously? I don't have OS X to try out, but I believe guides are available at www.macaddict.com/seti and www.macaddict.com/setiex . Running 2 clients maximizes your CPU usage, because Seti@Home program spends most of its time writing to the hard disk, and that CPU time can be used efficiently by a 2nd client.

Where did you get your information?!  sah (the command-line version) spends almost all of it's time doing serious number crunching (FFTs and the like).  It only "checkpoints" to the hard disk around once a minute, and even then only writes a few kilobytes at most.  In other words, sah is not I/O-bound.  The application / screensaver version eats CPU drawing the "pretty graphics", but it doesn't access the HD any more than the CLI.

I'll put it another way:  I'm running two copies of sah on my dual-CPU machine (nice'd, so they only use "spare" cycles), and my CPU usage is 100% on both CPUs.  If it were I/O-bound like you say, then there would be a significant idle percentage showing up, and there isn't (it's zero, as expected).

In other words, don't run more copies of the sah client than you have CPUs to feed it.  Yes, you can run more than this, but you won't get any gain, as each one will get (on average) less than 50% of a CPU; there's only so much CPU to go around

The only way I can imagine getting sah to rattle the hard disk significantly would be if you have less than around 16 meg available physical memory available.  In this case, the HD would then be thrashing from the Virtual Memory system constantly having to swap pages in and out of memory.  Of course, in this case, you still wouldn't get a gain from running multiple clients per CPU, since you would be forcing even more thrashing in that case.

Well this new 3.08 client is really good. I stopped for a while (I think I was using 3.03), now my Dual Gig MDD is moving fast enough that I can sit here and watch it build the stacks in the progress bar.

I should be on track for about 1 every 4.5 hours with the Screensaver version.

I might try instaling 2 cli versions if I get time.

i use a few PCs to do it, PCs have to be used for something