Hyper Threading + Seti and other projects

[Soulkeeper]

anyone experimenting with hyper threading on the P4's or Xeons
wonder what speeds would come out if you were crunchin two seti packs at once with hyper threading enabled

 

[Engineer]

The only thread I can remember with hyper threading....

Hope this helps.....

 

[Poof]

Thanks to TA member Barbary's seminal work on hyperthreading back in January , there were additional discussions on this topic at Ars including:

This initial one
And another
Generic P4 discussion

One of the interesting issues that came up is how to bench SETI on a hyperthreaded machine in a way that would allow an "apples to apples" comparison to the current TLC database results, ie., when hyperthreaded, you see a production increase out of the box when compared to non-hyperthreaded, however apparently (if I am interpreting this correctly), the actual reported WU times are the same between hyperthreaded and non-hyperthreaded. So basically, the calculated time to process the WU drops significantly when hyperthreaded (leading to more WUs/day), but that isn't reflected in the CPU time reported in the result.sah file, if I'm not mistaken, or something like that... :Q

Roelof Engelbrect who authored Setispy and who helps manage the Ars TLC bench database, weighed in on the situation regarding the best way to have hyperthreaded benches reported...

 

[Soulkeeper]

cool thanks