SetiBOINC on P4-HT, 1 or 2 processes?

[PatrickV2]

Hi there,

One of the machine I'm running BOINC on, with the optimized binaries I download from lunatics, is a P4 2.8GHz, HT-capable CPU.

It's running the following kernel:

Linux <name> 2.6.26-1-686 #1 SMP Sat Jan 10 18:29:31 UTC 2009 i686 GNU/Linux.

BOINC recognized the 2 virtual CPUs and runs two instances of SetiMB/AP at the same time.

When benchmarking, BOINC reports:

13-Mar-2009 22:21:57 [---] Running CPU benchmarks                               
13-Mar-2009 22:21:58 [---] Suspending computation - running CPU benchmarks     
13-Mar-2009 22:22:30 [---] Benchmark results:                                   
13-Mar-2009 22:22:30 [---]    Number of CPUs: 2
13-Mar-2009 22:22:30 [---]    683 floating point MIPS (Whetstone) per CPU
13-Mar-2009 22:22:30 [---]    1140 integer MIPS (Dhrystone) per CPU
13-Mar-2009 22:22:31 [---] Resuming computation

Anyway, my question is: since a P4-HT only provides basic HyperThreading capability, what would be faster: running 1 SetiMB/AP process at the same time (and not incurring the overhead of the HT-ing), or do I indeed get the highest RAC when running 2 at the same time?

Regards, Patrick.

[Jason G]

...
Anyway, my question is: since a P4-HT only provides basic HyperThreading capability, what would be faster: running 1 SetiMB/AP process at the same time (and not incurring the overhead of the HT-ing), or do I indeed get the highest RAC when running 2 at the same time?
...

I tested this on my p4 3.2GHz ast year.  I get a fairly whopping 30% better overall throughput running Hyperthreaded, and the machine is more usable for other things too.  As they say though, "Your mileage may vary"

[PatrickV2]

...
Anyway, my question is: since a P4-HT only provides basic HyperThreading capability, what would be faster: running 1 SetiMB/AP process at the same time (and not incurring the overhead of the HT-ing), or do I indeed get the highest RAC when running 2 at the same time?
...

I tested this on my p4 3.2GHz ast year.  I get a fairly whopping 30% better overall throughput running Hyperthreaded, and the machine is more usable for other things too.  As they say though, "Your mileage may vary"

OK, thanks for the feedback; I'll leave things as they are.

Regards, Patrick.