it’s not exactly eating though the units, but the cpu heatsink is only a touch over ambient. So, another eight should be being brought for our testing cluster tomorrow, once approved - I’ve got to sort out halving the hog rack and then I’ll see about spending the £600 required to populate the other half myself, possibly with the slightly higher rated LE chip, maybe a couple of duals.
The crunchyhog seti account is back 
DT.
Good going DT! 
I might just take that back about going through units, especially when you take a look at the other computer that returned this unit 
http://setiathome.berkeley.edu/workunit.php?wuid=238130437
DT.
Nice work dt, good to see a proper farm coming together again, these quad core based setups are ok but you can’t beat the fun of a bunch of motherboards on a rack
Why not a bunch of quads on a rack? I do wonder where the best crunching_performance/cost is these days. If I was still into farming, I’d probably get a low end quad and mildly clock it.
There’s a preview of Nehalem on one of the other sites. Includes the return of hyper-threading so 8 threads total on a quad core single socket. Be interesting to see how they perform in the real world.
[QUOTE=mackerel;415894]Why not a bunch of quads on a rack? I do wonder where the best crunching_performance/cost is these days. If I was still into farming, I’d probably get a low end quad and mildly clock it.
There’s a preview of Nehalem on one of the other sites. Includes the return of hyper-threading so 8 threads total on a quad core single socket. Be interesting to see how they perform in the real world.[/QUOTE]
the testing they will be used for in the workplace needs IO, not cpu power 
It’s more a test of our messaging system than raw power. I agree though, I do wonder what the best bang for buck really is :chin:
DT.
Trouble is, you get into the various needs of each project. My personal main project is still rieselsieve, with 2 primes to my name so far. For that, you ideally need two things. Core 2 generation CPU, and clock. Core 2 has up to 2x the IPC of other processors at prime finding, and scaled well with clock. So for me to farm that, a C2Q is the obvious choice.
I have no idea how other projects compare. For example, I’m currently dabbling in a little bit of folding for another team. But I don’t know what the relative strengths/weaknesses of each processor type are for that. Does cache have much impact? Is it memory bandwidth or latency sensitive? Is SMP or 2x single instances better for unit count (not score). Should I run win or lin? So it goes on… (only put it on my testing box, 2x single core cli, linux).