Sanity check needed

[sdbardwick]

HELP!

I'm having one of those moments where I can't tell if I am just missing something obvious, or if my own arrogance is blindng me to the truth.

Please take a look at this thread at mersenneforum.org (my username is the same), and please tell me:

Am I (a) just plain wrong, (b) misunderstanding what the OP is saying, (c) not being clear in my theory, or (d) beginning to engage with a crank?

The folks over there are generally very intelligent and thoughtful, hence my hesitation.

Now if Intel had odd yield problems and shipped true quad cores but marketed them as hyperthreaded dual cores, that might explain things...

(Background info: Prime95's main computational code is highly optimized hand-tuned assembly language. It has few "bubbles" where hyperthreading can pop in and utilize otherwise idle resources.)

 

[ddrueding]

It looks like their mental block is thinking that hyperthreading actually gets them more performance.

From what I understand, the "hyperthreaded" core only gets the leftovers, and that wouldn't be very much in this case.

So if the two cores being used were the two actual cores (as it should be if the scheduler is worth a damn), performance will be very close to the same as using all "4" cores. If you can assign it to use one entire actual core (the "real" one and the "hyperthreaded" one), then performance should be closer to using 50% of the CPU.

But really, if he is so set on dicking with the CPU scheduler, he should disable HT completely and take control of it manually.

In short, you may be crazy, but in this case I think you are right

 

[time]

I read the thread. Your posts are remarkably clear and easy to follow.

As Ddrueding says, the guy is being stupid about hyperthreading. Unfortunately, I can't think of any simple way you can convince him that he's deluded - he sounds like he'll believe what he wants to believe.

Here's a reasonable precis of hyperthreading that highlights the performance issues.

 

[sdbardwick]

Thanks for looking; I do appreciate your wasting time on such a trivial matter.

I'm probably just going to walk away from that discussion, (ok, maybe I'll leave a snide remark) because (a) the OP cannot furnish any supporting evidence that I'd accept as genuine, (b) he has the capability of testing my theory but decided not to, which leads to my invoking the adage that(c) one should never argue with an idiot, as they will drag you down to their level and beat you with experience.

 

[MaxBurn]

My advice is move on, yes your right but arguing on the internet isn't going to do anything. This link is for you.

http://xkcd.com/386/