I have a few dual Xeons kicking around (ah the joys of a boss who buys hardware when I ask for it). I'd never really considered hyperthreading until yesterday - I enabled it on one of my machines, just to see 4 framebuffer penguins with my shiny new 2.6 kernel (yes a late transitioner). Now I'm starting to wonder how HT and 4 virtual processors instead of 2 real ones will affect my crunching. Anyone have any benchmarks?

Dan

> I have a few dual Xeons kicking around (ah the joys of a boss who buys
> hardware when I ask for it). I'd never really considered hyperthreading until
> yesterday - I enabled it on one of my machines, just to see 4 framebuffer
> penguins with my shiny new 2.6 kernel (yes a late transitioner). Now I'm
> starting to wonder how HT and 4 virtual processors instead of 2 real ones will
> affect my crunching. Anyone have any benchmarks?
>
> Dan
>
>

Hyperthreading is a technology which tries to do two things at once with one cpu. The theory is that the CPU is big, and an instruction only uses part of the cpu, so they try to cram a couple of instructions at once, and run them at the same time.
So.. if you have two processes, and one is doing integer math, and the other process is doing floating point stuff, then the cpu is able to kinda run both at the same time efficiently. Having two seti processes will have two things competing for the same part of the cpu doing the same math, so you won't see the same boost as if you were running a web server while watching DVD's for example.

I've experienced about a 33% increase in productivity for SETI with HT enabled.

With HT enabled, you'll lose performance per unit for an overall increase in productivity.

Time to crunch a WU on my main rig

HT disabled - 2:10 - ~12WUs/day
HT enabled - 3:05 - ~16WUs/day