Sometime during Thursday, April 17th (UTC) we are going to install the new FMA3 (aka AVX2) version of LLR on the PPS project, in preparation for the Shakespeare's Birthday Challenge that starts on Sunday.

Especially on Intel Haswell CPUs, but to some extent also on Ivy Bridge and Sandy Bridge CPUs, this new app is significantly faster. It also consumes more power, and generates more heat. Please make sure your computer is up to running the new app.
...

RHMash
Orange, I would assume any performance gains by reducing the threads used on these would be very minimal given the short run times.  But I could be wrong.  I never heard/read about reducing cores and threads for such a big increase until that last challenge....  If I remember right, the improvements were due to the amount of memory on the cpu each task required.  PPS are smaller primes and should require less memory than the PSP primes, so I wouldn't predict much of an increase, especially on newer cpu's, other than maybe improvements from less heat....

Khalifrio
Sounds like my i5-3570k should be ideal for this since it does not have any HT to start with. I ran one WU when I signed up for Prime Grid to get ready for this challenge but its been long enough ago that I don't remember how long it took to run. I will finish off the tasks I have and load up a set of Prime Grid WU's and see how my i5 does for a comparison to you folks i7's.

Khalifrio
I came across this while browsing other forums. While it does not directly relate to Prime Grid, I'm sure the applications work differently, it does show the difference of HT on or off while running Rosetta. Just a little fyi for those interested.
 
http://boinc.bakerlab.org/rosetta/forum_thread.php?id=6419#76624

<app_config>
    <app>
        <name>llrPSP</name>
        <max_concurrent>4</max_concurrent>
    </app>
</app_config>

Punchy
 

planetclown
Another option is to keep HT on and limit the application to one task per physical core (not thread).  For example, I used an app_config.xml file with the following contents during the last PG contest to limit myself to 4 tasks even though I had 8 threads available.  

Does that have a different effect than just limiting CPU usage to half the "processors" in BOINC manager?  Are you sure that BOINC manager is smart enough to use one thread per core in that configuration?  I would expect that it simply limits the number of tasks and relies on the OS to schedule them; the OS may or may not assign them to separate physical cores.  Not trusting software myself, I go to the BIOS and turn off HT.