HT getting old?

I just thought that I'd see, am I the only person who fails to see any huge performance increase with Hyperthreading? I can see a POTENTIAL use, but as of now, HT seems rather pointless. I don't really see that the improved benchmark scores come from much more than higher clock speed. As far as gaming is concerned, benchmarks with Intel HT CPUs fall short of AMD equipment, in spite of a speed difference of over 700 mhz. If I am missing something, please let me know. I for one would enjoy net seeing any more HT articles until the technology develops.

I agree. I don't currently see much in the lines of performance for HT. However, I would like to stress that if your gonna put HT in a chip, ya just as well do it earlier. This way, you learn your lessons, and when you WILL need it, you will having working at peak efficiency.

One example....the current P4 architectures. When the P4 Willy came out, the performance sucked. People hated it, and rightfully so. Nobody in their right mind is laughing at the P4 architecture, now, because many technologies within it are maturing.

One extra hurdle that HT has to overcome though, is programming that utilizes HT. I don't know how fast people will start programming for it. That, I think, will be the ultimate test of HT....convincing the programmers, especially the game programmers.



Okay, I got it apart. Now how do I put it back together again???

i'm hoping it will make my computer usable when i'm doing processor intensive things (video conversion). right now i just have to let it sit, but with this hopefully i can get other stuff done too

--------------

<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq¬found=1&code=1" target="_new">mubla otohp eht ni ecaf ruoy teg</A>

HT is fundamentally useless for any processor-intensive applications. It's a handly little multi-process accelerator, and not much more.

In theory, in a program designed <i>just right</i> it can also be a multi-thread accelerator. However, in practice, software engineers can profile the exact CPU usage of their software and arrange and tweak the algorithms to keep the CPU more fully utilized than HT will allow for. And that's <b>MUCH MUCH MUCH</b> easier than writing (and consequently <i>debugging</i> ) a multi-threaded program.

So in other words, HT is more or less a joke to software engineers. The vast majority of those who care will tweak their code <i>long</i> before they add the complexity of writing multi-threaded code. The vast majority of those who don't care won't even tweak or multi-thread their code in the first place, because they simply don't care.

And sadly, in this day and age, there are <b>way</b> too many software developers (and entire companies) who simply don't care. They just want a working product at most. (And usually they just want a <i>mostly</i> working product that they can get to market and patch up later.) They don't care if their code is horribly inefficient. Most just figure that with the continual march of technology, it isn't worth the time to optimize code.

And by the way, I'm not even talking about optimization to a specific processor. I'm talking about simple things like when doing a simple loop over values in an array, to use addition to increase a pointer to the next element in the array instead of using random access with an iterator. (Because iterators for an array use multiplication to determine a pointer, where as using addition to increment the pointer yourself saves the CPU a considerable workload.)

If all software engineers were to just write good code in the first place, HT would be completely useless in games. It's only because there are so many half-skilled monkeys writing software that HT could possibly ever improve the performance of a processor-intensive application in a single-CPU system.

So because writing a well-written single-threaded application will outperform any single-CPU multi-threaded application (even with HT), all that HT really does is make multiple processes run better.

<font color=red>--The turkey may burn and the family may fight,
but video games are forever, so to all a good byte.--</font color=red>
<font color=green>Winter Solstice: <i>The</i> reason for the season.</font color=green>

I would very much have liked to see how the best from AMD compares to the best from Intel in this area. Funny that THG left this very relevant comparision out of the picture, I mean to the average user, the ability to deal efficiently with several programs at the same time, seems at least as relevant as the other types of benchmarks they run.

<i>/Copenhagen - Clockspeed will make the difference... in the end</i>

The whole point of HT is to improve performance of multi-tasking. It is not <i>supposed</i> to improve performance when running a single application. Admittedly, the current variation of HT is not that great, but HTII (on Prescott) and the upcoming HTIII should considerably improve HT performance. Of course, not much is known about the future iterations of HT. In certain specific cases, HT provides a nice performance boost (like running several intensive applications at the same time).

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?</font color=green>

So basically, it is just another addition to the Intel processor, like MMX, SSE, SSE2.....

It does have its value....but then again, that value is minimized in the inefficiency of human programmers.

Then again, some programmers still aren't programming for MMX utilization....



This post is guaranteed fresh for up to 7 days after the date printed on the box.

I understand what you're saying, that HT is not some miracle performance booster. Intel, though, claims that HT can increase performance even more using their new compilers. HTII is rumoured to be SIMD instructions for HT, and who knows what that could do. Intel may even make an "instruction pool" specifically to work with HT, on top of the regular instruction pool that P4's already have. And of course, Intel claims that Nehalem will "greatly expand" on HT, and that statement can be interpreted in many different ways. If you think about it, how can you "<b>greatly</b> expand" on HT? IMHO, the only way to do that with HT is to increase performance if you have not just a multi-threaded application, but simply several <i>processes</i> running on your machine. Intel may pull a miracle out of it's hat and integrate micro-ops fusion with HT, which definitely <i>would</i> increase performance of ALL applications. I know I'm just speculating, but think about it yourself, how do you "greatly expand" on HT?

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?</font color=green>

Maybe they won't bee SIMD specifically, they could be some different type of instructions. Or maybe they will be SIMD, but Intel may add something (like execution units) to make it all work. Prescott is supposed to have <b>some</b> sort of instructions. Originally, it was though to be SSE3, but now it's rumoured to be HTII. Intel has actually stated on a PDF that Prescott <b>will</b> have HTII. Specifically, though, it's not known exactly <i>what</i>HTII is. Here, check out the facts for yourself:

<A HREF="http://www.extremetech.com/article2/0,3973,801644,00.asp" target="_new">http://www.extremetech.com/article2/0,3973,801644,00.asp</A>

<A HREF="http://www.extremetech.com/article2/0,3973,587833,00.asp" target="_new">http://www.extremetech.com/article2/0,3973,587833,00.asp</A>
This is a good article which explains hyperthreading.

Finally, here's the PDF where Intel states Prescott will have HTII: (warning, this PDF is 22MB, so it takes a while to download - HTII and Prescott is on Slide 26)

<A HREF="http://images.visualwebcaster.com/8064/pdf/highres.pdf" target="_new">http://images.visualwebcaster.com/8064/pdf/highres.pdf</A>

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?</font color=green>

I understand what you're saying. What I said, though, was that the HTII instructions are a "rumour", I never said it was fact.

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?

"... In the semiconductor industry, it's good to be paranoid ..." - [Andy Grove]</font color=green>