Quote:
|
Originally Posted by cyrnel
"Flop out" meaning to go somewhere else and get work done because that app is taking too long. Sure they would see an improvement, unless the test were designed to prevent it.
|
Yes, I perfectly understood what you meant by "flop out." I can only imagine that you felt it needed clarification because you are just
that surprised by my assertions. Of course, that's why I made them... If I told everyone what they already know, it wouldn't make for as interesting a post. It's always more fun to surprise people with things they
didn't know, although I'm a little surprised that this is one of them...
Quote:
|
Knife, I'm a bit baffled by your contention. The XP thread scheduler is neurotic and wasteful but it will show an increase in work with multiple cores and multiple single-threaded applications. NT has had parallelism built into the kernel since way back when. XP Home will see some less benefit due to its further restrained scheduling support vs. multi-core + HT. But the benefit is there to experience.
|
There's nothing neurotic about the NT scheduler. It's a round robin scheduler and it's why you probably wouldn't see any benefit from the second core. Incidentally, there's almost no difference between Home and Pro and there's absolutely no difference "under the hood."
It seems that I have not sufficiently explained how my claims are true and I'm tempted to give a long and detailed dissertation but I want to make sure there's are no miscommunication first, since I would hate to exert all that effort just to find out that you all almost understood what I was saying, anyways. Wasted effort sucks...
Quote:
|
Try setting affinity (for a few more % benefit) with task mgr or one of many launch utils. Monopolize each core with separate tasks and time vs. disabling one core completely. I don't know how you interpret the increase in work as other than an increase in performance. Not only is it measurable, but it's obvious to users making rigorous use of multiple single-threaded apps. XP's scheduler is less efficient than tuned, multi-threaded apps, and it's possible to sabotage the benefit with I/O or common resource blocks, (including user input) but otherwise independent apps do eat this stuff up.
|
It is very hard for an application to use 100% of a single processor's power and, even when they do, they do so infrequently. For instance, IDE's only use the full power of the processor while compiling. The rest of the time, which happens to be the majority of its use, it's nothing more than an over-glorified word processor. It just sits there, doing nothing but waiting for you to press the next key. When you do, it does a little bit of something while still only using a fraction of the processor's computing power and then goes back to doing nothing but waiting for you. The vast majority of applications are like this. They do nothing but wait for you. So, surely, you can see that even if you have several applications doing this, you're still not utilizing the full power of even that single processor, much less a second core! I'm hoping that you already understand all this and that I'm just reviewing but, if you haven't, then I really must start from the very beginning! Otherwise, my second point should be of interest to you...
Secondly, even if you have
an application that has sufficient computing needs to constantly utilize the processor, such as any video renderer, you still need another program to do the same to utilize the second porocessor. Otherwise, even on a single processor, your other applications will be content with the few quanta it will get from the scheduler since all they do is sit around waiting for user input, anyays.
Now, in the example above, if you did have a second processor/core, the other applications will be scheduled on it and there will be a
literal increase in performance but it will be very slight and certainly not noticeable, which is what I'm referring to when I say "makes no difference." I mean, really, if you can't tell there's a difference, do you care that there was a difference? I can only assume it's this that you're thinking of when you're "a bit baffled" by my claims...
If there was anything you didn't understand, please point them out and I will do my best to clarify!