Sign in with
Sign up | Sign in
Your question

what is hyperthreading?

Tags:
  • CPUs
Last response: in CPUs
Share
October 16, 2002 9:04:43 AM

hey guys i'd like to know whats hyperthreading but u knoe in a very dumb language. I coudnt understand the THG Guide coloumn about this subject

ummmmmmmmmmmm...............wha?

More about : hyperthreading

October 16, 2002 9:53:18 AM

A "thread" is a stream of instructions from a certain program. The OS bundles instructions in threads so that it can effectively simulate multitasking (i.e. running 2 programs at once). The CPU basically processes one thread, then switches to another thread. By doing this very fast, it can simulate the illusion that it's running 2 programs at the same time. However, due to x86's poor parallel nature, sometimes all the instructions in a thread can't fully max out the processor's resources. Sometimes the processor is left idle due to data dependencies, etc. This is what hyperthreading hopes to alleviate. If the instructions in one thread can't max out the processor's resources, it takes instructions from another thread and fills in the "holes" that aren't being used on the processor, effectively not only doing more than 1 program at a time (better for multitasking) but also more efficiently using the processor's resources.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
October 16, 2002 1:25:47 PM

Quick and simple answer: Hyperthreading is making one CPU pretend to be two so that almost none of the CPU is being wasted when running programs that don't use the CPU to its full potential. :) 

<A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new"><font color=red>Join</font color=red> <font color=blue>the</font color=blue> <font color=green>THGC</font color=green> <font color=orange>LAN</font color=orange> <font color=purple>Party</font color=purple>!</A>
Related resources
October 17, 2002 2:19:34 AM

HT would benefit more K7 that P4 in theory.But major change to the architecture must be made

Now what to do??
October 17, 2002 6:41:13 AM

so does A Xp use hyperthreading? and when intel incorporates it in it's 3.06 GHz Processors will it be able to widen the performance gap b/w itself and A XP


ummmmmmmmmmmm...............wha?
October 17, 2002 11:50:36 AM

Yes and no. Early HT tests revealed performance up by 20% and down by 10% in some apps. It remains to be seen, as those tests were mostly on network benchmarks, IIRC, on Anandtech.com.

--
What made you choose your THG Community username/nickname? <A HREF="http://forumz.tomshardware.com/community/modules.php?na..." target="_new">Tell here!</A>
October 17, 2002 12:52:19 PM

Theoretically, the K7 would benefit tremendously from SMT as it does have a lot more parallel resources, through decoding, scheduling, and execution than the P4 does. You'd have to redesign the core to keep track of which instruction belongs to which thread, but it could be done.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
October 19, 2002 2:49:35 AM

Kinda weird how many Intel features would benefit even more on the dated K7 core. I think they have not yet discovered how to exploit the P7 core to its uniqueness, since what they are adding, the competition could use even more lethally.
Wish AMD would at least try to license those features, they're taking SSE2, why not HT as well?! Provides tons more competition.

--
"Let Go." -Avril Lavigne
October 19, 2002 12:39:53 PM

I don't think these features are of "liscense" material. Concepts such as superthreading and simultaneous multithreading are not exactly something you "patent". It's implementation-specific. I.e. as long as AMD doesn't just take the blueprint for HT from the P4 and mimic it, it's fine for them to implement some type of simultaneous multithreading on their chip. Of course, that may not stop Intel from suing, seeing how the trend today seems to be moving towards patenting everything, including such broad and open concepts.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
October 19, 2002 1:22:12 PM

This is a bit like the controversy over the next generation of Open graphics languages or the HLSLs.

But it is interesting to see if AMD or other CPU makers might actually try to use such idea and perhaps make it their own. Right now, I could clearly see that the best one who could benefit TREMENDOUSLY would be Opteron SledgeHammer. Adding HT on it would not only be already powerful for a K7-K8 execution unit core but as a server chip it would be a huge bang for the buck. Why oh why doesn't AMD even try to think of these things WE simpletons come up with! :eek: 

--
"Let Go." -Avril Lavigne
October 19, 2002 6:37:48 PM

Remeber that the P4 was litterally designed around HT since its been onboard since the Willy. I dont think there is a simple answer to implement something very "Intel" into something very "AMD" or "IBM" the cores are different in soo many ways. I dont see the technologies crossing in the fashion you are wishing would happen.

Like I keep telling FatBurger I personally think the Hammer is late not because of SOI or 0.13 or production times it's SSE2. Cant even imagine the time needed to build that around the Hammers FPU's or 64bit registers.

-Jeremy

<font color=blue>Just some advice from your friendly neighborhood blue man </font color=blue> :smile:
October 19, 2002 7:02:02 PM

Well, SSE/SSE2 does require dedicated 128-bit registers separate from the rest of the x87/MMX registers so it would take some time to build. However, seeing how they've already demonstrated prototypes of the final silicon, it's not really something with the design process. Which basically rules out anything but manufacturing limitations. Seeing how Opteron is still slated for Q1 2003, I'd say it's a problem with mass-producing enough chips to feed the desktop market.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
September 27, 2012 8:54:36 AM

imgod2u said:
A "thread" is a stream of instructions from a certain program. The OS bundles instructions in threads so that it can effectively simulate multitasking (i.e. running 2 programs at once). The CPU basically processes one thread, then switches to another thread. By doing this very fast, it can simulate the illusion that it's running 2 programs at the same time. However, due to x86's poor parallel nature, sometimes all the instructions in a thread can't fully max out the processor's resources. Sometimes the processor is left idle due to data dependencies, etc. This is what hyperthreading hopes to alleviate. If the instructions in one thread can't max out the processor's resources, it takes instructions from another thread and fills in the "holes" that aren't being used on the processor, effectively not only doing more than 1 program at a time (better for multitasking) but also more efficiently using the processor's resources.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.


Superb, appreciate your explanation !!
a b à CPUs
September 27, 2012 9:27:59 AM

This has got to be the oldest thread revive I have ever come across...
!