that's true. Normally the software will only run on one core. There may be some special treatment with multiple operating systems or multiple users on an X2 system, but it are not for the timid.

I bought an AMD64 4000+ single core machine to tear up the workunits. It's 2.4 GHz and the time is comparable to the Pentium 4 3.2 GHz. I'm doing about 4-5 workunits/day, somewhat less than the number of Rosettas that I did before.

So, if I were to buy a amd 939 X2 3800
put windows xp sp2 on it, and run the fah project, How long exactly would each work unit take?
Remember, on my p4 3.2ghz machine, it takes roughly 5hr 37 min

it would take around 1oo minutes, amd rules:)

Hello whocrazy,
I think that mariduru was being facetious. I was just looking over a comparison benchmark test comparing Intel's 64 bit code against that of the Athlon 64 earlier today. AMD looks pretty good, but I could not think of any rule of thumb that would have let me predict all of the differences.

So with that understood, I'll try to come up with a ballpark figure. AMD rates their chip at 3800 compared against an Intel 3.2 Ghz chip (3200). So the difference is 3800 - 3200 = 600. Call it a 1/6 difference. If Intel runs a work unit in 5 hours and 37 minutes (337 minutes) then a core of the AMD chip should run the job in [337/6 is about 56 so. . .] 337 - 56 = 281 minutes. So with 2 cores, each running a BOINC Windows work unit, I will guess at 140 minutes so overall throughput is one work unit per 2 hours and 20 minutes rather than 5 hours and 37 minutes.

This probably overstates the AMD advantage a little. But it is reasonable to assume that you can cut 3 hours off the job (remember that I am averaging throughput for 2 cores here). Just remember to double up on memory. (1 GB minimum)

mycrofth

Good choice, whocrazy.

Right now we are running UD.exe on Windows, a single-threaded distributed computer agent which is well tested and reliable. We are implementing BOINC for Linux. It is a new distributed computer program that is rapidly changing (and has numerous bugs). But in time we will be running it on Windows as well for all our projects. Right now we have some people running the Human Proteome Folding Project on Windows on BOINC. BOINC can take advantage of multi-processor or multi-core systems.

My guess is that in a few months everyone with multiple processors will switch over to BOINC to take full advantage of their computer power. But right now, running BOINC can be a bit of an adventure. Like my father driving Model A Fords down red clay roads before the highways were paved.

Ummm. . . No, that's a poor analogy. BOINC is already much more stable than that. It is just that I have been spoiled, so I notice even minor glitches.

Even after years, these are still the early days of distributed computing.
mycrofth