Imagine the points that would be awarded with that kind of processing power. . . .

http://www.xtremesystems.org/...

Hi Movieman!
I am interested by your experiments. Since I anticipated that the HT topic would surface again in discussions with the start of Nehalem processors I have just run a series of tests with my old P4 HT. My purpose was to have a real life answer to this question:
Is the global throughput of a HT processor much better by feeding it with two simultaneous WUs than by giving it those two WUs one after the other?
In Boinc's terms, is it more productive to allow two processors for a single HT core or is it more efficient to restrict it to one.

I have never considered disabling HT itself because I think that when a processor has to deal with more work than it can it will always be better if it can use features designed specifically for making it easier and more efficient. And in our machines, even with one WU at a time, a processor has always other things to do.

Reading your post I have the feeling that your experiments have been about disabling the HT feature itself. Could you please confirm, and if yes would it be possible for you to redo the same test as mine (4 WUs alone, then the other 4 alone too, vs all eight WUs all at once, but all with HT ON) to see what happens on the global throughput on one hand, and on crediting on the other hand.

On crediting my tests show that there is a considerable difference between claimed and granted credits between these different running modes (and it is normal), so I can wait until you are allowed to run the tests online in real life (if you accept to do it, of course).

Cheers. Jean.

MM,

Mulling this over, a benchmark at full throttle and results crunched at half the speed with HT. Wonder if the credit methodology is going to gag at this. Looking forward to hear about that, and sure we will ;>)

thanks

I have been following Intel Nehalem [i7] Press releases, fan sites and videos for over a year:

Intel Core i7: The Essential Guide