I am resuscitating an old PII 400 and figured on running BOINC/Rosetta when the kids aren't doing homework. I'm sure a cure for cancer will be found before it finishes crunching its first WU. Will adding additional RAM improve the crunching speed or is BOINC efficiency mostly processor based?

I'd say anything less than 256 of ram and that will be the bottleneck. . if you have that or more, then it's probably cpu bound. I notice a large difference between our machines with 128 and 256 of memory.

---

Indeed, crunching speed is mostly CPU-dependant. However, if you're serious about the crunch time (I've seen the projected time on a seti WU at a few thousand hours) then it could be trying to use the swap space as primary memory.

Like H&FS said, if you're under 256, an upgrade will be cheap and easy, and WUs should take no more than a day or two (or three? I haven't crunched with sub-GHz CPUs since my 333MHz died).

---

I did a bit of calculator work. . here are how 3 systems compare based on hardware. Note, these are our employee's workstations, so there may be some variation due to them being used.

All 3 machines have Intel CPUs
500mhz with 128MB RAM = ~27,600 seconds
500mhz with 256MB RAM = ~20,200 seconds
1Ghz with 256MB RAM = ~11,900 seconds

---

I did a bit of calculator work. . here are how 3 systems compare based on hardware. Note, these are our employee's workstations, so there may be some variation due to them being used.

All 3 machines have Intel CPUs
500mhz with 128MB RAM = ~27,600 seconds
500mhz with 256MB RAM = ~20,200 seconds
1Ghz with 256MB RAM = ~11,900 seconds

In practice, 256 is the minimum sweet spot for Rosetta only, IME. Four of my systems all run 98 SE in 512. When I view them with system monitor, as often as not there's no swap in use, and still over 256 physical ram free.

I'd say that with 256 you could probably run Rosetta and one other app (just not CPDN), with "leave in memory set true" and get away with it.

---

Thanks for the advice!

---