• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

HyperThreading's Value

Status
Not open for further replies.

Archaea

[H]F Junkie
2FA
Joined
Oct 19, 2004
Messages
11,934
I am a PC tech, I got in a discussion with one of my other PC tech friends about the value of hyperthreading.


I found a video at Tom's hardware that pretty much sums up HT's value...IF my bud would watch this - the argument would be dead.

http://www4.tomshardware.com/images/thg_video_5_p4_ht.zip

I say it's 100% worthwhile - makes a computer feel snappier, he says its a joke. He's only ever had AMD, I've been in the Intel camp since the P4 1.6 @ 2.4 days.

My upgrade path has been this
Pentium 1 (133Mhz)
AMD 200
AMD K6-2 450
AMD Duron 600 @750ish
AMD Duron 800 @900ish
AMD T-Bird 1.1? @ 1.2ish
AMD T-bird 1.3? no overclock - limited by motherboard
P4 1.6a @ 2.4
P4 2.4b @ 2.8
P4 2.4c @ 3.0
P4 3.2c @ 3.7

My last AMD was a t-bird 1.33. I currently have a 3.2 Northwood overclocked to 3.7 at stock voltage.




For those interested in the long conversation...


Nelson says:
Just buy an AMD system.
Archaea:
no
Archaea:
i like my intel's stability
Archaea:
and hyperthreading RULES

Nelson says:
LoL
Nelson says:
hyperthreading is a joke.
Archaea:
it's not a joke
Archaea:
read - google it
Archaea:
makes a big difference in everyday usability
Nelson says:
I have.
Archaea:
http://www.gamepc.com/labs/print_content.asp?id=pentium4c&cookie_test=1
Archaea:
“As we've shown in previous tests, Hyper-Threading can add anywhere from 5% to 30% performance gain in lots of applications, and is incredibly useful for heavy multi-taskers. While we were skeptical at first, Hyper-Threading really does make a heavy-usage system feel much smoother. “

Archaea:
doesn't say anywhere in any review I've found that it's "a joke"
Archaea:
don't be arrogant next time if you don't know something
Nelson says:
"feel" isn't real world
Archaea:
it totally is
Archaea:
i've read that all over the net
Nelson says:
Never said I knew everything.
Archaea:
your system FEELS more reponsive
Nelson says:
this is the same as "my car FEELS" faster but isn't.
Archaea:
that's probably the most commonly noted opinions on hyperthreaded systems
Nelson says:
HT isn't worth getting tremendously excited about, anyway. It does provide a small speed boost, generally, provided your operating system and motherboard and are able to turn it on, which is not a problem for Win2000/XP and pretty much any current Socket 478 board. But a P4 with HT is not the same as a real dual-processor system. Both HT pseudo-processors are competing for the same resources from
Archaea:
i understand the concept
Archaea:
generally 15%
Archaea:
overall performance
Archaea:
BUT
the concept is that multitasking
Archaea:
which EVERYONE DOES in general everyday use
Archaea:
"feels" faster
Archaea:
because your computer acts as if it has two processors
Archaea:
and can manage the multitasking better
Archaea:
http://www.extremetech.com/article2/0,1697,1152222,00.asp
Archaea:
believe me i've read up on this a lot, and having p4 for the last several cpu upgrades
Nelson says:
I don't need HT to act like there is two processors.
Archaea:
i've researched each time
Nelson says:
15% is a joke
Archaea:
granted only the last two have had hyperthreading
Archaea:
it's not
Archaea:
15% average
Archaea:
some have none
Archaea:
some have an actual dedregation with hyperthreading
Archaea:
some have a lot of benifit
Archaea:
but overall 15-30 % is what most review sites state
Archaea:
15% being a low ball park figure
Archaea:
and it's not just the 15%
Archaea:
that is not much
Archaea:
it's the fact that you can alt tab smoother, run winamp while doing other stuff
Archaea:
etc
Archaea:
because your processor splits up the threads its doing
Archaea:
playing a game goes to the core cpu
Archaea:
background tasks take the 2nd hyperthreaded piplines
Nelson says:
HT is a joke, and intel should of just went with DC.
Archaea:
i'm done talking to you
Archaea:
there are 100% unlimited google links that say hyperthreading is a very worthwhile alternative to an expensive (at the time it was introduced) dual processor solution
Archaea:
yet you keep saying it's a joke?
Archaea:
WHY?
Nelson says:
Because software solutions are a joke til the hardware makers can MAKE a new chip to do what they are intending. Unless the software is supposed to regulate how stuff is processed.
Nelson says:
HT is a software solution
Archaea:
it's not only software
Archaea:
there is a hardware component
Archaea:
managed through software yes
Archaea:
but so is EVERY pc component
Archaea:
and since the OS supports it
Archaea:
all applications thereby support it
Archaea:
it's a speed boost
Archaea:
and if you have the rather of speed boost or no
Archaea:
why not?
Nelson says:
no, not all applications support it.
Archaea:
yes it's not dual core
Archaea:
but it WAS a step in the dual core direction for those without dual core
Archaea:
and you'll be hard pressed- if not find it completely impossible - to find a person on the web who has had a hyperthreaded system and says they thought it was a joke....
Archaea:
it can be disabled in the bios
Archaea:
nobody does
Archaea:
not all singular applications support it --- no
Archaea:
but since the OS supports it
Archaea:
most things well
Archaea:
if in the fact that nothing else - alt tabbing to those programs works faster
Nelson says:
woohoo alt tab
Archaea:
the program coding ITSELF may not take advantage of the hyperthreading capability but the OS does
Archaea:
which means that multiple process threads going on at the same time will run better
Archaea:
so if you are encoding video and listening to an mp3....neither of those applications may directly support hyperthreading - but since the OS does...both processes will be smoother
Archaea:
encoding will be done quicker - etc.
Archaea:
http://www.tomshardware.com/2002/11/14/single_cpu_in_dual_operation/page25.html
Archaea:
quoted from tom's hardware
Archaea:
“In real world situations the feel of Hyper-Threading is much better that our benchmarks actually show; this is why we also produced the video, which demonstrates this. Hyper-Threading is particularly beneficial when at least two applications are making heavy demands on the processor simultaneously. “
Nelson says:
2002 is old.
Archaea:
that's when the technology came out
Archaea:
watch the video
Nelson says:
You stick with your HT and I will stick with real solutions, being hardware.
Archaea:
you don't have a "real" solution
Archaea:
and neither could you or i have afforded it in "2002"
Archaea:
yet we could afford Hyperthreading
Archaea:
and saying it's a joke is just being stubborn
Archaea:
dual core at the time HT came out was 3x's more than single
Archaea:
motherboards cost 2 to 3 times more than single processor motherboards
Archaea:
There was no such thing a dual core - just dual processor
Nelson says:
You pay for what you get.
Archaea:
bah
Archaea:
just realize that if you ever applied at a job i was interviewing for and said something and blatantly wrong as HT is a joke - you'd never get a job from me....
Nelson says:
I wouldn't want to work for you.
Nelson says:
=P
Nelson says:
specially if it was @ help desk
Archaea:
one last point...tom's hardware has a video that shows a p4 3.0 with HT enabled beating a p4 3.6 with HT disabled.
Archaea:
By a LONG time
Archaea:
you have to have divx to see the video
Archaea:
but here's the link
Archaea:
http://www4.tomshardware.com/images/thg_video_5_p4_ht.zip
 
AMD's offerings have changed since the 1.3GHz Thunderbird days. Although HT does do quite a good job with multitasking. If you are gaming however, definately go AMD.
 
Got bored reading all the convo, but where you say its works with 2000, it works, but isn't optimized for it, i had some problems running HT under Win2k Advanced server, especially in games, Windows XP is optimized for it and works great, i've played with non HT 2.8ghz intels and can tell a good difference and wouldn't to back to single core again due to it :)

Now since you didn't pay anything for the HT integration it does make it worthwhile, but its only a stepping stone to Dual Core, Dual CPU setups

Edit: the New dual core EE's actually have you pay for the HT, making it not worth the extra money

And like the guy above, a A64 can still beat one in gaming, SMP enabled games (quake4, cod2) do get decent boosts though.
 
I should edit to say that he has an AMD 1800 overclocked to 2200....

I would agree that AMD's hold the upper hand now, but not a couple/several years ago at the time of the Hyperthreading introduction...and yes while dual core is better, hyperthreading is still valuable -

I really don't have much need to upgrade my 3.2 @ 3.7 northwood and Abit IS7 socket 478 stystem for at this generation of games. Doom3, HL2, and Fear all play fine - meaning all their derivative games will work. Considering that my MB was available in may of 2002...that's a great "single system motherboard" lifespan....

http://www.abit-usa.com/products/mb/reviews.php?categories=1&model=79
 
Even I, an AMD enthusiast, admit that HT ROCKS< it pushed in multithreading, it was revolutionary, it ushered in DC. I wish I had HT, but alas, I am going to get a 4200 :D


Can you make a video of your HT system? Maybe that would prove finally it's worth.
 
HT is good, but apparently Intel didn't really implement it very well. There was an article at Anandtech that said that HT was more of a hack for the NetBurst processors which were not designed to use HT. That's why you get the occasional decrease in performance.

That's not to say that Intel is dropping HT. Intel has said that we will eventually see cores that can process 8 threads at a time.
 
The "value" of hyperthreading is now non-existent. You now pay for it in the form of the EE chips. Get a DC chip and call it a day. Any flavor will do. 80+ % beats 15 any day.
 
mwarps said:
The "value" of hyperthreading is now non-existent. You now pay for it in the form of the EE chips. Get a DC chip and call it a day. Any flavor will do. 80+ % beats 15 any day.

But if software was optimized for it, I think it's a great idea to keep HT. For example; imagine quad core CPU with HT on each one so you have 8 simultaneous threads (in theory).
 
HT was valuable on the P4's, because it provided releif from the overly long pipeline, without it perfromance and "snappyness" would have suffered. I wouldn't call it valuable really, it was more of a band-aid for larger issues.

Notice how HT is absent from Intels new offerings, moving to short pipeline, high IPC setups does not benefit from HT like the old P4's did.
 
Well HT is great, but only when compared to single core processors. With dual core CPU's becomming more and more common, HT has less and less of an impact.
 
I usually try to avoid arguing with devlopmentally delayed peoples. It really isn't worth it. Ive had my little is7 system for around 2.5 years now. Its nice. I only upgraded the 2.4b to a 2.4c (damn intel released em without much notice if anyone remembers). Hyperthreading mostly made anything before the barton core look pretty below average. Now is a totally different story, but he only has hopefully a throuoghbred. I say he needs at least a barton to be competetive and a 939 a64 to actually put down hyperthreading.
 
It had a slight amount of value for very few programs. Some programs actually ran SLOWER or performed worse with HT enabled. BUT - as far as I know Intel is about to drop it all together with the onset of Dual Core processors. On servers at work - certain applications require that it be disabled for performance reasons.
 
They should just include HT on all cores, as its already builtin.

dual-core with two threads per core would be nice and not that hard to enable.
 
JAW said:
They should just include HT on all cores, as its already builtin.

dual-core with two threads per core would be nice and not that hard to enable.

They do this for the EE editions.
 
HT was useful for speeding up desktop maneuvering. However, any modern processor should move plenty fast in that environment if you have a reasonably clean system. In actual application performance it was only useful for providing minor boosts when running 2 apps that were processor hogs. IE, it's more useful for saying 'look what I can do' while you rip a dvd and are almost able to play a game at the same time, where you couldn't do it at all with a regular processor. Now it's been made virtually obsolete by dual cores, and even when it was new was only really a novelty as it allowed IE to update a bit quicker while you did something processor intensive in the background. Most apps and programs barely make use of 2 threads, let alone 4 or 8. Even 4 and 8 way systems don't show great performance increases over dual processor (or dual core) systems, so I can't see how having HT on multiple core systems will do anything at all for quite some time (by which time we'll have 4 and 8 core chips anyway).

So I'd agree with your buddy, it wasn't necessarily a joke but it didn't really do too much, certainly nothing to get too excited about.
 
I think its better than nothing, but not by much, and highly overrated by the Intel fanbo1s.
I had a 2.8P4 w/o HT @ work, and replaced it with a 3.2 with HT. Its still slow as hell. I noticed no smoothness gains at all. Now my AMD X2, or my dual Xeons @ home, hell yeah, VERY smooth
 
your friend is an idiot, and you were wasting your time trying to prove something to him. You were trying to have a decent argument, and all he could say was "HT is stupid" over and over again.

Find a new friend, that one is obviously not worth keeping
rolleyes.gif


EDIT:
as far as HT being slower in some things than a HT processor (as people inthis thread are discussing), it is simply not true. Yes, with Windows2000 and very early implementation of HT in XP, performance in a few rare things would suffer, due to improper handling. now, on a single core, HT chip, performance increase goes from 0% up, it does not go down... assuming you are using an updated (ie, sp2) windows XP installation. Now then, i know someone is going to jump on me and say "But the EE dual core HT suffers sometimes!!" Yeah, i know. its becuase the OS doesnt know how to handle it correctly - it will put a full load on a virtual HT core instead of the physical processor, cuasing performance drops. Here, again, we see a software problem, not hardware. With an update to XP's SMP kernal with this in mind, performance should only go up, not down.

HT uses the pipeline in the P4 to work... and was always included in the core (although im not sure about willamette), but it was disabled. It was definatly not a "hack" added to the p4 in the latter days of its life, it was always (afaik) present on the Xeons, it just came to the desktop segment later. it simply uses unused areas of the long pipeline in the P4 (mostly towards the end of it) to handle a second thread, hense why we do not see this kind of implementation on a more parallel, short pipeline architechture like the p6 or "p-m architechture" (doesnt officially have a name)

HT has very little merit left, with true SMP being present in a single processor these days. However, when it was released, it was a very cool technology, and nothing to scoff at. anyone who says it was a 'hack" or not worthwile at all is simply ignorant.

BTW, this is from a guy who has never owned a p4 and refuses to buy anything netburst based becuase i simply thought it was a horrible architechture and a bad move by Intel.

edit2: oh, and HT is no more a software solution than true SMP is
rolleyes.gif
in both cases you have hardware capable of handling two threads at once, one (SMP) is just faster than the other (HT). Install a non-SMP aware OS (like win 9x) on a duallie and see how far that gets you...

edit3: (sorry)
Nelson says:
HT is a joke, and intel should of just went with DC.
this is the most ignorant thing i have ever fucking heard. the die size for a dual core northwood would be ENORMOUS, not to mention the R&D was nowhere near finalizing any sort of multi-processor core in 2002.
 
seanmcd said:
It had a slight amount of value for very few programs. Some programs actually ran SLOWER or performed worse with HT enabled. BUT - as far as I know Intel is about to drop it all together with the onset of Dual Core processors. On servers at work - certain applications require that it be disabled for performance reasons.

Even when a program ran worse with HT enabled, it was usually not more than 1-3% worse. (Usually)
 
I miss HT. It's not as good as dual core, but it's better than single core.

I ordered an Athlon 64 X2 yesterday because even fast single core performance makes multitasking painful. I have to rely on 2 computers now because my desktop becomes unusable when performing CPU intensive tasks, something that didn't happen when I had a P4 HT system. I'm not going to task manager to change process priority or make new shortcuts for everything to work around that problem.

And keeping with the theme above, HT is something that is mindlessly trashed by AMD fanboys. :p
 
pxc said:
I miss HT. It's not as good as dual core, but it's better than single core.

I ordered an Athlon 64 X2 yesterday because even fast single core performance makes multitasking painful. I have to rely on 2 computers now because my desktop becomes unusable when performing CPU intensive tasks, something that didn't happen when I had a P4 HT system. I'm not going to task manager to change process priority or make new shortcuts for everything to work around that problem.

And keeping with the theme above, HT is something that is mindlessly trashed by AMD fanboys. :p

hardly "mindlessly trashed". I posted my opinion based upon personal experiences. My experience was I couldn't tell it was there, period.
 
When I upgraded to a P4 with HT, from my non Hyperthreaded Northwood, I saw an immediate and dramatic difference in XP's responsiveness while running multiple applications. Since upgrading, I haven't encountered a situation where everything came to a screeching halt because I was doing too much at once. It felt like the burden placed on my computer by background processes just disappeared.

Not to flame Steel Chicken, because I have seen this argument played out a million times, but I just can't see how anyone could fail to notice the improvement. It was like night and day. Maybe he is spoiled by all that dual core, dual processor goodness he mentioned. Those systems are certainly faster than a single core, HT P4 - but I still think HT is the next best thing.
 
http://www4.tomshardware.com/images/thg_video_5_p4_ht.zip

Have you guys watched this video?

There is no question to hyperthreading's worthwhile after watching this...

Argument dead.

This video shows a P4 3.6 chip with HT disabled vs. a P4 3.06 chip with HT enabled.

The 3.06 chip KILLS the 3.6 chip and the sole reason is HT.

If you can't view it - install the Divx player and you'll be able to.

www.divx.com
 
Archaea said:
http://www4.tomshardware.com/images...deo_5_p4_ht.zip

Have you guys watched this video?

There is no question to hyperthreading's worthwhile after watching this...

Argument dead.

If you can't view it - install the Divx player and you'll be able to.

www.divx.com

that link is busted...but I found the video in your OP.

Hyperthreading is a bandaid to a broken architecture IMHO. It's very helpful, but comparing a P4 with and without HT isn't the same as comparing a P4 w/HT to a shorter pipeline AMD chip. The argument from the AMD side is that they don't need HT because their pipeline is shorter. I have a P4 3.0C w/HT and a dual core setup and I've used Barton/Winchester/Venice chips also. HT is "snappier" I suppose than the single core chips, but it doesn't change the fact that Netburst sucks ;).
 
kirbyrj said:
that link is busted...but I found the video in your OP.

Hyperthreading is a bandaid to a broken architecture IMHO. It's very helpful, but comparing a P4 with and without HT isn't the same as comparing a P4 w/HT to a shorter pipeline AMD chip. The argument from the AMD side is that they don't need HT because their pipeline is shorter. I have a P4 3.0C w/HT and a dual core setup and I've used Barton/Winchester/Venice chips also. HT is "snappier" I suppose than the single core chips, but it doesn't change the fact that Netburst sucks ;).

During the time period AMD had no shorter pipeline advantage - or if they did in fact have a short pipeline - there was no advantage. The AMD chips of this timeperiod were a good deal slower than a HT P4. Remember we are talking about 2002 time period introduction.
 
hmm? the xp series was ok, but I still wish my A64 had something like HT, cause I multitask so much more than I game.
 
Archaea said:
During the time period AMD had no shorter pipeline advantage - or if they did in fact have a short pipeline - there was no advantage. The AMD chips of this timeperiod were a good deal slower than a HT P4. Remember we are talking about 2002 time period introduction.

I don't think it's really in question whether or not the P4C's were better than the Athlon XP's ;).
 
Enjoicube said:
hmm? the xp series was ok, but I still wish my A64 had something like HT, cause I multitask so much more than I game.

dual core is where it's at.
 
yeah, but I mean offering that extra bang for the buck, HT is pretty much free, DC is NOT AT ALL free, alas I am soon going dual core.
 
if you are going for a single Core CPU I wouldn't go AMD(I've had both) HT's usefullness is HUGE if u do any multi tasking, say for example ur playing game x and u alt tab, u can a) wait for ur browswer to open or b) have it open right away, guess which one does what =p

Single core = p4 6xx series hands down
dual core = amd hands down

thats my opinion anyways ^_^
 
Archaea said:
During the time period AMD had no shorter pipeline advantage - or if they did in fact have a short pipeline - there was no advantage. The AMD chips of this timeperiod were a good deal slower than a HT P4. Remember we are talking about 2002 time period introduction.
ermm... AMD has always used a short pipeline, its similar to the P3 with a much stronger FPU.

the XPs were not a good deal slower than a P4C line in single threaded apps, but the P4C's did have the lead.
 
yeah,the 6XX series was good, I would recommend the 640/650. The 50X series is terrible, no HT, no 64 bit 1 MB cache and about as expensive as the 6XX. I might soon be getting a 3800 or 4200 or 165, I really want DC.
 
the 500 series has HT... with hte exclusion of the shitty budget ones with the 533 FSB.
 
kirbyrj said:
that link is busted...but I found the video in your OP.

Hyperthreading is a bandaid to a broken architecture IMHO. It's very helpful, but comparing a P4 with and without HT isn't the same as comparing a P4 w/HT to a shorter pipeline AMD chip. The argument from the AMD side is that they don't need HT because their pipeline is shorter. I have a P4 3.0C w/HT and a dual core setup and I've used Barton/Winchester/Venice chips also. HT is "snappier" I suppose than the single core chips, but it doesn't change the fact that Netburst sucks ;).

That's BS.

SMT ( or HT for Intel ) isn't a bandaid.It is simply a way to better use a cpu's resources.

And it isn't related in any way to long pipelines or Netburst.There are quite a few cpus out there that have it , some like Niagara have a 6 stage pipeline which kinda shows how good your argument is.

The reason why AMD doesn't have it now ?
Incapable of doing the extra work needed.It needs quitte a bit of knowledge and willingness to spend time and money which AMD didn't have at the time.But they will probably implement it too in the future and then you'll show up with " xx (insert name for AMD SMT) rocks!!! " :eek:

It added 5% more die area to the P4 but it counted for 95% of the validation and testing time.

Btw, the concept is 40 years old.Quitte a time before the "broken architecture"...

Read , it helps in getting wiser :

http://en.wikipedia.org/wiki/Simultaneous_multithreading
 
Enjoicube said:
notice that i said 50X not 5XX :rolleyes:
in that case, the 50X processors are no where near hte price of the 6XX series. i assumed you meant the higher ends ones since you compared them in price.
 
savantu said:
That's BS.

SMT ( or HT for Intel ) isn't a bandaid.It is simply a way to better use a cpu's resources.

And it isn't related in any way to long pipelines or Netburst.There are quite a few cpus out there that have it , some like Niagara have a 6 stage pipeline which kinda shows how good your argument is.

The reason why AMD doesn't have it now ?
Incapable of doing the extra work needed.It needs quitte a bit of knowledge and willingness to spend time and money which AMD didn't have at the time.But they will probably implement it too in the future and then you'll show up with " xx (insert name for AMD SMT) rocks!!! " :eek:

It added 5% more die area to the P4 but it counted for 95% of the validation and testing time.

Btw, the concept is 40 years old.Quitte a time before the "broken architecture"...

Read , it helps in getting wiser :

http://en.wikipedia.org/wiki/Simultaneous_multithreading

I didn't say that no other processor used something similar to HT, but Intel's use of SMT is directly tied to their long pipeline and inefficiency otherwise, thus the bandaid comment. No need to be so condescending like I'm a little school boy (or AMD F4nboy).

And note that all I said was the argument from the AMD side is... I didn't say it was MY argument (and specifically phrased it that way).
 
kirbyrj said:
I didn't say that no other processor used something similar to HT, but Intel's use of SMT is directly tied to their long pipeline and inefficiency otherwise, thus the bandaid comment. No need to be so condescending like I'm a little school boy (or AMD F4nboy).

And note that all I said was the argument from the AMD side is... I didn't say it was MY argument (and specifically phrased it that way if you'll read...and remember it helps in getting wiser ;) ).

Don't feed the trolls, man. It burns. You should know better :p
 
kirbyrj said:
I didn't say that no other processor used something similar to HT, but Intel's use of SMT is directly tied to their long pipeline and inefficiency otherwise, thus the bandaid comment. No need to be so condescending like I'm a little school boy (or AMD F4nboy).

And note that all I said was the argument from the AMD side is... I didn't say it was MY argument (and specifically phrased it that way).
i was under the impression that the width was important just as much, if not more than the length ;)
 
Status
Not open for further replies.
Back
Top