Well, as of tomorrow (06/19/2006), it's Goodbye SETI@Home
& Hello Folding@Home!
* It'll be cool joining this team's forums for it, because most folks here have one heck of a powerful system - & that ought to translate out to many results units generated (for a potentially great cause).
Once more - The reason(s) for my delay in joining was simple:
A.) I wanted to see how HIGH of a SETI@Home "RAC" (recent average credit) granted per unit I could get outta this AMD rig, vs. the older Intel P4 3.2ghz 2nd machine I keep here (was my primary unit before this one & is now relegated to server-work only)
&
B.) To see how far I can get into my teams "top computers" rankings (made it as high as 12 w/ my P4 before, & this time I can't seem to crack the top 25 w/ an undoubtedly superior CPU in my AMD now).
Also, what I am finding is, on these types of projects (because I also view them as a long-term benchmark as well as being decent pursuits):
1.) Intel CPU's do seem to do better, & these types of projects are cpu-cycles dependent (& it seemed that even though the P4 only has 512kb of L2 cache on it, its 800mhz superior clock outweighs L2 cache amounts, because my AMD has 4 times that (2mb shared between the dual cores present), but runs @ 2.4ghz)...
&
2.) That the compilers out there for software DO tend to favor Intel CPU's, over AMD ones (this is widely known @ SETI@Home) & optimize custom clients for their work better...
Even though AMD cpu's blow AWAY Intel ones in the Fpop (floating point) ALU fpop superiority, & also have more L2 cache by far than a Pentium 4 does, which SHOULD matter here (the built-in seti benchmarks prove this every time)?
Intel cpu's still win out, & it's due to optimized clients being better for them as well as more mhz, because the compilers used apparently are geared more to Intel CPU's. It's GOT to be this, what else could it be?
(E.G.-> The AMD cpu I have here knocks the you-know-what outta my older Intel P4, hands-down on every test there is, including SETI@Home's built-in benchmark, but the results show quite the opposite oddly enough!)
It probably will be the case for FAH as well, but no biggie (because I heard its units aren't as "uniform" in their finish times either as are the SETI ones, which also aren't always "the same" each time they finish either, but tend to not vary as widely as finish times are concerned)...
APK
P.S.=>
I do wish, however, that somebody did an "optimized client" for FAH!
The optimized custom client build types that optimize for various instruction sets (floating point processing stuff usually, via MMX, 3dNow, SSE/SSE2/SSE3) DO run massively faster & are JUST AS ACCURATE, @ least in
SETI@home + far faster & just more efficient...apk
.
