Sign in with
Sign up | Sign in
Your question

What CPU war?

Last response: in CPUs
Share
June 19, 2004 1:50:52 AM

June 18, 2004


I find it amusing that Intel and AMD CPU fans form sides to an engineering argument. If we all really were engineers, what argument would there really be? Facts are facts and simple application studies can determine your best choice, and the best processor for your tasks. What “choice” is there but the right one, after all? So I say there is no AMD verses Intel war. The question is what choice should you make?

Intel and AMD depart from CPU processing most significantly by the depth of their prediction pipelines, and subsequent memory needs to fill that pipeline. Intel CPU’s, P4’s for sure, like lots of closely coupled cache memory to supply a speedy but deep prediction pipeline. Why is it speedy and why does it need so much memory? Imagine that the CPU knows that I want a glass of milk every time I eat a sandwich. Not only will the CPU/cache it fill a glass FULL of milk (lots of memory to supply ALL of the predicted milk) out of fast local memory (the fridge), but it will also do it as fast as it can at full CPU clock rate. But what happens when I don’t want either that much milk (I can’t drink 32oz!), or switch to coffee right in the middle of my sandwich? The CPU has to start a reverse course flushing process. Get rid of the milk, run back to the fridge (local memory), find coffee in the local fridge or possibly go farther to the grocery store (RAM) and run then run all the way back to the table and start pouring coffee. Not only does it waste time if I change my drinking taste, but the process can’t just stop pouring milk immediately. It has to finish what it started over time much as the last remaining stream of milk must fall through the air to the glass. Since the CPU’s branch prediction is so sure you want milk, the stream is fast and the pitcher held pretty far from the glass. It has no knowledge of your picky taste. Once it gets moving foreword, it wants to stay moving smoothly and quickly forward in a big way. No sense sloshing all that milk around! This is great if you do repetitive tasks. Eat milk with a sandwich ALL the time. CPU’s like repetition. Maybe that’s why artificial intelligence is artificial. The machine can simply fix mistakes faster than humans can recognize it! The machine really doesn’t “know” what you want at all it just predicts it.

An AMD processor is a little different in that the pipeline is small and not so deep. The CPU runs at a two-thirds or so clock rate reduction compared to an Intel CPU but does more work per clock cycle. And, the CPU doesn’t pretend to be all that smart. It meters out a little milk at a time into a smaller 8oz glass. So it doesn’t need a huge fridge to supply a large amount of milk to load from a huge local memory. It can also walks back and forth from the fridge to the dinner table much faster than the Intel part so if you CHANGE your mind, it’s very efficient at swapping milk for coffee. It has less to milk to rid itself of, it gets to the fridge faster, the fridge is smaller so coffee is easier to find, and the trip back to you is faster. With the CPU doing more work per clock cycle than an Intel part, it easily makes changes with much less need for a large expensive local memory. A littler fridge and a fast host are fine. But this is a disadvantage IF you want a LOT of milk, and the smaller fridge runs dry, and the CPU has to go to RAM memory to get more milk.

The issue at hand is do you do predictive tasks or not? Games are highly unpredictive as game speeds ramp up. Just what will you be doing and when? Only the shadow knows. When predictive tasks that need LOTS of the same data streams come into play, Intel rules. But these tasks seem less in number than unpredictive one that we use most often. Or, the differences in predictive tasks are measured in a few seconds or so and don’t really matter. Do I care if a video clip is decided in 20 or 25 seconds? I do care if my computer stutters playing games or on interactive tasks involving any eye – hand – monitor inputs. Watching a CPU crunch a set of numbers is not really a big deal if its 5-10% between the two. For me it is, anyway.

Cost is not an issue with true performance users because we pick the right CPU for the job and cost is secondary. But if you aren’t a performance user, I suggest that the short pipeline AMD structure will benefit you more often on most applications, and for less money, than an Intel part. Now that the 939 boards are out, the cost is even less than it was before with even faster memory speeds. AMD likes faster RAM because it has to use it more than an Intel part, but this is still faster than flushing out bad branch predictions to large local on die memory based on application tests.

Pride and assumed superiority can get in the way of good engineering. Why do we take sides so firmly when there is no side to take? One does this, the other that. But we will take thing “personally” which seems to mean that I should be hurt to hear the truth. Why? If I take things as simply true, or false, I learn something. If I defend a bad aspect of a CPU or any other item, I’m stuck defending a thinning circle of knowledge. I have a Mercedes. I didn’t make this car, it just owes me what it is, no more no less. I bought what it is, not what it isn’t. An expensive car, for sure, but do I tattoo a MB logo all over my body and defend it to the death? No, I don’t. I sent in a five-page letter to Mercedes and pointed out PAGES of things wrong with their car. You know what? They sent a 20-page latter back outlining 2,000 changes that will make the next car better! This is the way it is supposed to be. A good consumer who is realistic with the product, and a manufacturer who is, too. I’d have to say AMD is much better at this than Intel, or it’s customers. I’d say most people need stop cycle anti-virus enabled 64-bit computing with lower CPU costs much more than expensive Intel Extreme Edition P4 processors. I’d never buy a subzero 100 cubic foot fridge, but the P4EE seems to be just that. Like lots of cold milk?

For BOTH manufacturers, I’d say who needs hot and leaky 90nm die CPU’s verses a larger die with less leakage and maybe two parallel CPU cores that really work? The me too trap may consume AMD if it follows Intel’s lead before real low leakage dielectrics can steam the wasted power in 90nm die products. Not to mention my Intel P4 locks up at random intervals with its secondary “virtual” core enabled which speaks of great marketing, and a flaky CPU. It just isn’t a benefit when it doesn’t work ubiquitously.

Those who hang with good people get better themselves. AMD users seem more knowledgeable about their processors, and also feel comfortable admitting their weaknesses. Intel seems more like a Harley, once you paste it all over your body, most people feel kind’a stuck trying to defend what ever it is, or will become. Not only does the customer defend lack of change, the manufacturer sees no point in providing any! With enough money, you can fence it out till you HAVE to change, though. Intel and Microsoft can both use this tactic. Big companies make money, small one’s make products.

There is legitimate Intel users who also understand their CPU. But most don’t really know why they really bought their CPU. I have a P4 3.0Ghz 533MHz bus 845 chipset system that is barely better than an AMD 2800+ Barton system at 1/3 MORE money. I built the Intel system when AMD did have thermal issue to work out. But AMD has worked them out and their current crop of CPU’s are well suited for short branch prediction users. With the 939 socket boards taking either 32 or 64 bit processors with more economical and faster dual memory, it seems AMD has done well for itself.

More about : cpu war

June 19, 2004 2:44:10 AM

Holy crap what a long post. We need a condensed version.
Related resources
June 19, 2004 2:57:23 AM

No, we need him to delete all the cut and paste he did. Once was fine, but he has at least 4 copies floating around this forum.
It would also be nice if he new what the function of prefetch was.
June 19, 2004 3:42:05 AM

ROFLMAO! :lol: 


"I am become death, the destroyer of worlds. Now, let's eat!
June 19, 2004 4:17:49 AM

i really like the passive agressive tone he takes up in the end. its quite an art to be soe lame.

"I see" said the blind man to his deaf dog
June 19, 2004 4:23:04 AM

I can make a cappucino faster with a prescott beat that!

i need to change useur name.
June 19, 2004 4:33:43 AM

LOLLOL :lol:  at least intels are good for something! :wink:

"I see" said the blind man to his deaf dog
June 19, 2004 5:11:08 AM

I found it very interesting. And I really liked the food analogy. As a matter of fact I think I am going to go get a glass of milk now. :p 
June 19, 2004 5:12:12 AM

but you can only have either an 8oz serving or a 32 oz serving, there are NO choices!

"I see" said the blind man to his deaf dog
June 19, 2004 5:42:44 AM

If you liked the food analogy, then let's give credit where credit is due.

It's ripped right out of Scott Mueller's "Upgrading and Repairing PC's"

Cheers,
Steve

P4c 2.6/I875/nVdia 5200
June 19, 2004 11:56:19 AM

I also found it very interesting, and I'll take 2 8oz glasses of milk

thank you
June 19, 2004 4:09:01 PM

Other posters often times don't look for answers in OTHER posts, so I link it into associated post topics.

I suppose you're for big government to tell you what to do, how much to read and what to think. If reading more than a line or two crashes your RAM, move on. Lame individuals lean on one another, it's a form of the "intelligence" I mentioned in my post, ganging together to learn much about nothing. Except that is, nitwits seem to love company. Information is simply true or false. Guess what the majority of yours is.

As for plagiarism, what web site? I didn't grow up in the computer era so blindly learning to copy others work isn't one of my interests. If it parallels someones work, link to that site, please. It's probably good information that will help people decide where their CPU purchase fits in. Good news though, with post like this group has made, no one will plagiarize your work.

And passive aggressive? The truth hurts those not in search of it no matter how it is delivered. There, how did I do?
June 19, 2004 4:30:30 PM

LOL... Whoo you're such a rebel, you're sooo rebellious, You don't do what authority tells you to. OOPDIFUCKINGDOO, i honestly don't give a flying f*ck about your ideas. This is one of the most useless posts out there, milk? cookies? coffee? WTF?!

Anyhow, i have to go to work now, will continue this when i get back.

<A HREF="http://www.ebaumsworld.com/celebgay.html" target="_new">click here now!!</A>

<A HREF="http://www.ebaumsworld.com/epilepsyr.shtml" target="_new">DON'T CLICK HERE!!</A>
June 19, 2004 4:35:33 PM

It would have been better if there were more true in your ``theory``.I also disagree with most here.The philosophy of benchmarking.To me it just a test to have a idea not a final jugement.

i need to change useur name.
June 19, 2004 4:36:47 PM

Dual channel milk hmm...

i need to change useur name.
June 19, 2004 4:56:50 PM

I'm not sure where you guys get your educations, but you should listen a little better in class. No theory was mentioned. The twp CPU types are vastly different in pipelines. The test are not theory, the easily show where deep and shallow pipelines excell. Software can indeed be a mixture of the pipeline requirements. No surprise there. But when eight out of ten test favor shallow pipelines, and the ones that don't are less intrusive in operation, the shallow pipeline method is pretty compelling.

For the record, I have a P4 system, but don't support EITHER CPU just because I own it. I support their strengths or weaknesses, no more.
June 19, 2004 6:04:04 PM

yes the two chips are different, but the endless debate over benchmarking wotn end. There are a slew of benchmarks that favor one cpu or antoher. And this isnt limited to differnt tasks, but within the same tasks. There are some audio encoding apps that favor p4, some favor athlon 64, thats where the confusion comes from, there is no one set of tests that can say one cpu is an out right winner. Thats why its best to base decisions apon your particular use and the app you use most, not general benchmarking, especially any synthetic test.


Also, it would be wise to try and not insult people's intelligence, your credibility looses alot when you do that.
June 19, 2004 6:29:46 PM

"I'm not sure where you guys get your educations"

Not sure where you got yours either. How many "educations" you done got??

You plaguarized a book, a copyrighted book. Specifically, page 65 of the 1575 page book, and the analogy is used extensively throughout. You just replaced "Hot Dog" with "Milk." Congratulations you innovative thinker. I personally don't care, just give credit where it's due.



P4c 2.6/I875/nVdia 5200
June 19, 2004 6:50:27 PM

How old are you...seriously?

AMD's dual-core sex fights due in 2005
June 19, 2004 8:16:28 PM

Not that you seem to be proud of your cave, who are you? It wouldn't be a picnic without the ants, would it? It's a shame that a reasonably sophisticated web site is patrolled by the equivalent of web site terrorists, simple post a silly reply and fade into the?

Me, I have a Purdue University EET degree and was one of the first to received the EET outstanding alumni award for technical industry achievement (35 out of 70,000 were given this award). I have worked in the wire and cable industry for over 20 years and am the second highest patent holder in the company's 100-year history. I'm 46 years old and can be reached for reasonable comment at any time. I don't need to hide or avoid new FACTS to improve the ability to lend a hand to people in this news group.

You conspirator's seem to be bent on your own demise because you have to take fiction as fact in order to function. I have no clue whom, or what the book is you mentioned when I posted my reply. He or she should be flattered the concept was used again. For the less technical, it is a good learning tool. Please (I already asked that this be done) post the TWO different links I'm claimed to have plagiarized so far (more to come I'm sure!).

Your appeals to ignorance (say or believe something that can't ever be proven either true or false) is your own demise. My post and reputation will reinforce who is really the better contributor. Many of you are very knowledgeable and I respect, and recognize, your consistent and accurate achievements. Don't let this discourage you. The truth is a wonderful insulator from common intellectual nuisance.
June 19, 2004 8:27:00 PM

Yes, I concur with your comments with regards to those who wish to be offended (they will be anyway). The rest of us of substance pay no heed to it all. But, it has been said to not get in a kicking fight with a fat man. I think you are 100% correct, and I will get my butt kicked. I apologize for my comment.

Now, you are also 100% correct in your comments on benchmarks. Statistically, with so little audio encoding and the like being done on applications, the AMD CPU is ahead in most real world applications over a WIDE base of uses. The Intel is a 100-pound gorilla on specific apps.

I only point out that it is best to be true to yourself when selecting a CPU and that for most users, an AMD is probably better IF YOU DON'T have a clue what you'll do with your PC. Real time apps are far more aggravating when they grunt and groan than crunch and burn bit crunchers.
June 19, 2004 8:43:51 PM

Rebellious? When is drawing an analogy rebellious?
What is authority telling me to do I'm not?
What is authority telling you?
Don't read the post if you're above it all.
June 19, 2004 9:13:42 PM

Ok i got a question for your super degree.

How SMT affect Branch history

i need to change useur name.
June 19, 2004 9:23:34 PM

Quote:
This is one of the most useless posts out there, milk? cookies? coffee? WTF?!


stfu, his post made sense and i enjoyed it


better than the stupid flame wars that everyone takes so personally


=)

-------
<A HREF="http://www.albinoblacksheep.com/flash/you.html" target="_new">please dont click here! </A>
<A HREF="http://www.subhi.com/keyboard.jpg" target="_new">This is you, interweb junky</A>
June 19, 2004 9:32:31 PM

http://www.cs.wisc.edu/~mulvihil/752_paper.pdf

Here, I found your answer. Remember, a driver must also support SMT (creative seemed to have issues early on).
I'm not so sure why you feel hyperthreading has to be 100% effective to be a valid technology.
Are you supporting the concept of 100% compatability?
delete all the non compatible applications till we can pretend that it is now 100% compatible?

You asked who I was, I told you. I'm flattered you think Im so super. Not admitting a lack of knowledge is the most difficult barrier to learning it. I was curious about your request and found a good paper on it. Glad you asked.
It is amazing how well it works with good drivers and software. Most CPU cycles are wasted by current software.

It seem true physical CPU's will pretty much displace hyperthreading and we won't really get to see mature software for the technology. Lock-ups or no, it still doesn't give the needed performance gains verses single proc systems.

Driver people, are drivers easier to write for multi CPU systems or are there simply more code that is bugged to apply to physical CPU drivers? Have virtual core CPU's been used in earlier systems verses true dual physical processor systems?
June 19, 2004 9:39:55 PM

yeah i totally agree with you, the average user doesnt do any specific computing so anyone in that group would be fine with an athlon system. in my years of building and testing systems, i have grown tired of all the benchmark wars. synthetic benchmarks are really a waste of time and are only for frivilous competitions.

as far as your creditials, there is no need for you or anyone to prove the experience they have. in the years that ive bene in the industry form getting my degree to running my own pc business, another thing i know is that even those with creditials can have the elast knowledge in the tech business, not refering to you of course. things are evolving constantly, and many get left behind or not care to keep up. the point is, just throwing around titles doesnt mean a person know more in this field.
June 19, 2004 9:52:47 PM

i dont think hyperthreading was meant for long term adoption. i think at the time, intel needed something to keep its edge, and it came up with this. while i like the technology, its biggest strength is using multiple apps at once, which owuld be something anyone could see, and not quite as much wiht specialized tasks. also hyperthreading really works best on the current p4 platform becuase of its architecture. it wouldnt be nearly as succesful on an athlon 64 or even p-m dothan becuase of thier architecture. those two designs dont use the same long pipeline structures and therefore arent idle as much as p4 northwoods or prescotts would be if it didnt have HT. not to mention things like better branch prediciton for dothan and ahtlon 64 versus prescott,the p4 just needed something to put those idle clocks to use and keep its pipeline filled, or as close to it as possible.

i think its a nice tech, but i dont see it as having a long term future. i agree with you about dual core, physical multiprocessing is still the best way to go, software just cant reach that level today.
June 19, 2004 9:56:40 PM

My intitial post was an attempt to lighten the mood buddy. I saw the "bandwagon" thing taking off. I think sometimes people just don't realize that although they may be joking or even just a bit huffy in the moment the comments stay up in print. I think sometimes though the reader can take it personally and then it becomes cruel whether it was intended to be or not. Anyway, I wasn't being sarcastic or commenting one way or the other on the validity of what you said. I thought it had some interesting thoughts.

I think I did actually go get a glass of milk after I posted that comment though.

...I like milk...
June 19, 2004 10:15:49 PM

i severyone who goes to purdue boreing and self vindicating like rower30?

its posting, not essaying. and we should all remember the logical falacies of argument, otherwise we start to sound like rower30

"I see" said the blind man to his deaf dog
June 20, 2004 5:12:46 AM

I disagree with you on SMT but due to time constrain i will cut this short.


Are you on realworldtech?

i need to change useur name.
June 20, 2004 12:15:06 PM

>It seem true physical CPU's will pretty much displace
>hyperthreading and we won't really get to see mature
>software for the technology

Hyperthreading and CMP are definately not mutually exclusive. HT is just a way to make more effective use of a (long) pipeline and increasing its efficiency. Simply put: insert instruction into the pipeline where there would otherwise be bubbles. This method is presented towards the OS as a dual thread capable cpu, much like a 2 way SMP or dual core system, but that is about the only resemblence.

Having two or more cores, be it on the same chip or on different cpu's doesn't negate this (potential) advantage at all. The only thing that changes is that if you only have 2 concurrent threads, and you have 2 physical cores (SMP or CMT), HT will not be able to provide a benefit, much like there is no benefit in having a single core cpu with HT running single threaded apps.

If you doubt the viability or usefullness of HT in combination with CMT, I suggest you look take a look at the Power5. If you doubt the usefullness of HT in combination with SMP, I suggest you look at some server benchmarks of 2 or 4 way Xeons with and without HT. It still offers just as tangible benefits, as long as you have enough threads to process in parallel.

So dual core chips will in no way replace HT. HT will still be a technology that allows you to make more effective use of a single core. performance gains might not be quite as big as with dual core, but then it also takes up neglectable die estate compared to multicore.

> Lock-ups or no

HT is (almost²) no different from a dual core or 2 way SMP system from a programming POV, that even includes OS support. In the (extreme rare) cases an app or driver would misbehave on a HT enabled cpu, I bet you a fortune it would behave just as bad on a 2 way SMP or CMP system. Code that crashes on such systems typically have race conditions, or rely on loop based timers (which could just as well cause crashes on faster-than-expected CPU's, remember the K6-300 issue with Win98 ?). Blame incompetent developpers, not the technology.

(²) the only differences I'm aware off is optimization; HT enabled virtual cores share the same L1, L2 and trace cache, unlike CMP/SMP implementations. This can result in cache trashing when one virtual core monopolizes the cache, and with each context switch the content of the cache must be reloaded from ram (or a higher level cache).

Furthermore, the OS ideally must be aware a logical cpu <> physical cpu, otherwise it will always try and assign the same thread to the same cpu whenever possible, which is good with 2 physical cores as it increases cache hit ratio, but can be decremental with 2 logical CPU's where the cache is shared anyhow. This is a minor change to the scheduler, just modify the thread/cpu affinity code, something that has been done for XP, but not Win2K.

>Driver people, are drivers easier to write for multi CPU
>systems or are there simply more code that is bugged to
>apply to physical CPU drivers?

You may want to rephrase that question, but if I understood you correctly: writing code for HT is no different than writing it for SMP/CMT. If you want to maximize performance, you may need some minor additional work to reduce chances of cache trashing, but this would be no different for a dual core CPU with a shared L2 (or L3).

>Have virtual core CPU's been used in earlier systems verses
>true dual physical processor systems?

Again, its not one approach versus the other. That is like saying "AMD64 versus SSE3". But no, AFAIK, the xeon was the first implementation of a SMT ("hyperthreading") cpu. Meanwhile Power5 also has it, and there are rumours Itanium would implement it. Note Power5 is already multicore, and Itanium will be soon as well. Its not "or" its "and".


PS: in case you wondered about my credentials, I'm 13 years old, I failed miserably in primary school since I just couldn't grasp 10 times tables; my biggest personal achievement is my diploma of 25m breastroke. No one else in my family ever achieved that so late in their lives or required more than the 7.5 minutes it took me. I also think its a record in the 50 year existance of our school.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 20, 2004 1:32:30 PM

You don't disagree with me, you disagree with the data in my post.

POST your disagreement with documentation, please.
June 20, 2004 2:22:04 PM

Someone asked a question, I honestly answered it. No more, no less. It was a reasonable question and I'm sorry if an accurate answer was somehow "throwing" anything around.

I disagree that there is no need to prove experience. Your posts, spelling, and thought process all "prove" your experience. Some are aggressive with degreed individuals because they themselves have not spent the time to advance their ability to discuss things in an even and fair manner.

Life doesn't discount ANY experience no mater where it comes from, that I agree with. There are educated idiots and an education is only as fresh as its use.
June 20, 2004 2:33:57 PM

There you go guys and gals, trooper has a great set of points on HT and the long pipeline in Intel CPU's. This makes good factual sense for Intel to figure out how to fill this long pipe.

I do some multi app work but the work done in the background isn't really being expected anytime soon. I lazily get around to it at the desktop level.

A server, though? Everyone is waiting on a server. This is still better with real dual CPU's.
June 20, 2004 2:47:14 PM

Be careful P4man, they might think were buddies. Wow, a good, long post that makes perfect sense. Oh, I didn't ever offer my past till I was asked about it. What's the shame in that? people deserve an truthful answer.

With HT turned on, several systems we use will lock-up if your Outlook Express is turned on. shut off the computer, reboot, and try again and it is fine. Till next time, anyway. The random nature (with wintel software no less) is consistent with HT enabled. Turned off, all is well.

What part of HT is being handled differently at each boot for this to be so consistent? It renders these PC's HT incompatible.
June 20, 2004 3:54:01 PM

>Be careful P4man, they might think were buddies.

Let's set the record straight then, since I don't like pretentious fudders that talk out of their clueless rear.

>What's the shame in that?

No one asked, the comment you responded to was sarcastic, in reaction to <i>your</i> rethorical 'where did you get your eductation' question, but apparently sarcasm is lost on your award winning genius.

>With HT turned on, several systems we use will lock-up if
>your Outlook Express is turned on. shut off the computer,
>reboot, and try again and it is fine. Till next time,
>anyway. The random nature (with wintel software no less) is
>consistent with HT enabled. Turned off, all is well.

Considering intel sells roughly 15 million Pentium 4's per quarter, and over 90% of them ship with Windows (therefore, OE) installed and HT enabled, I seems rather unlikely we wouldn't have head of any related stability issues. There might be something screwed with *your* machine(s), could be flaky driver, virus, BIOS that is not up to date, overheat, bad PSU, anything really, but it is NOT a general problem. So stop spouting FUD.

>What part of HT is being handled differently at each boot
>for this to be so consistent?

"Is it true only car cylinders turn around compression when the carburator and not the injection system is tuned for low emission in the exhaust charger or battery" ?

Man, I sure hope you know more about wires and cables...

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 20, 2004 5:16:23 PM

im just thinking that Ht is at the end of its life, if intel does indeed adopt the p-m architecture for desktops and gives up on netburst. if they do, and sense aperently HT doesnt work well with the p-m, im guessing becuase of the shorter pipelines, then HT will slowly fade away as netburst does.
June 20, 2004 5:24:49 PM

I doubt it. Power5 has 2 way SMT, Sun's upcoming Niagra features 4 way SMT per core (and 8 cores/chip), future Itaniums are more than llikely to feature a form of SMT (corse grain), and if I'm not mistaken, Fujitsu (and therefore now Sun) is also working on SMT Sparc64's chips.

Dothan based desktop/server chips may not feature it, but its a fair guess to assume future x86 chips will include it (again), be it K9 or Meron, or both.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 20, 2004 6:26:12 PM

P4man,

My rear is indeed clueless (I have lots of clueless body parts!), it doesn't even know....
Ok, my rear is taken care of. In place of a real knowledge base I guess I could replace it with aggressive profanity and chest thumbing that isn't seen as, "pretentious fudders" talking about themselves, either?

I can take issue with most all of what I know since it came from those before me. So did almost ALL of your knowledge, too. So what's the issue with "knowing"? If you "know" something before me you're smarter? Even if your knowledge isn't really yours either? I'd say that's fudding (good word). So that's it...we're ALL FUDDING around.

What FUD, I have no fear of many HT PC's locking. They simply do. Sure, if you message ALL the parameters, or buy DELL PC's where all the messaging is done for you, you might have an artificially perfect HT world. PC's simply have fewer configuration issues with HT disabled since it adds to issues already there. A stone stock PC still can have issues and single proc OS'es have been around awhile. What is it about system issues that have got everyone so defensive?

You make assumptions about HT's perfection that is too intertwined to really support. I turn it on and more issues arise. Well, there seem to be more issues doesn't it? You say its FUD / bad. I say it simply is. HT can sometimes do "this" (generally no), most of the time it can do that. Add issues and they statistically multiply the matrix of possible "none compliant" configurations of two raised to that power. You also seem insistent that I'm against HT in general. No, it simply doesn't do as much as people think it should, as supported by Tom's Hardware guide tests. Is it FUD your afraid of (isn't any) or that the HT world has smears on its glass exterior you're somewhat ashamed of (shouldn't be)?

This is the exact type of response I pointed out in my, "got milk" post that started all this. We have strong tendency to make closed-minded stands that get in the way of really developing anything past the idol crowd stage pushed on use by marketing. HT is perfect...right? The tattooed Harley crowd rears it's head.

Off of that for a second, you seem very well versed in the pursuit of HT perfection (not that it is there yet) and Intel architecture. And, based on your posts, you seem at least pretty concise on these issues. You deserve credit for that and your facts seem correct. Even if they are someone else's, just like mine.

Questions ( I know nothing about this, in other words, or my rear, it talks, I asked it!);
1.0 In a dual HT enabled system, is the motherboard seeing "four" CPU's? Two real / physical and two virtual?
2.0 If a pipeline prediction is incorrect and flushed, is the data in BOTH CPU's pipeline common to that processor flushed? If so, would this not have a double negative if too many wrong predictions were made?
3.0 Would this would be a negative if you use short duration type desktop tasks verses longer consistent one where you have convinced me that HT really will benefit this user?
4.0 Dual core type HT enabled tasks are more server centric than desktop, possibly?

Sometimes we get what we ask for, only to find it's not what we need. But it sure feels great seeing two CPU's when we boot. That is until the evaluations say it comes up short of expectations. Where are those application (most are not synthetic) benchmarks that show an Intel CPU beats an AMD in desktop use with HT enabled, and loses with it turned off? If we don't use scripted "real" applications what do we use to "prove" our point? More profanity?
June 20, 2004 7:24:44 PM

i only ment as far as intel p4 chips in the desktop area. i dont think this extends to the server arena, im sure SMT will go on for some time there, im just saying it probably wont in desktops. Anew form of SMT may be developed and used in desktops, but the current HT intel uses wont be used in p-m based mobile and desktop chips, so its likely HT will drop out of favor in the desktop area, while other forms of SMT will still be useful in server areas for sure.

I dont really see why you would think K9 or Meron would impliment a form of SMT, there arent really any hints at such a thing, especially on the amd side. in fact it looks like they are more concerned with physical multi threading and letting the software developers work on adapting to it.
June 20, 2004 7:58:29 PM

>I can take issue with most all of what I know since it came
>from those before me.

What a great point. The issue is not what you know or dont know, its what you proclaim to know when you don't. I'm allergic to that. You keep stating over an over hyperthreading results in instability, even when it has been pointed out to you over and over what a nonsense claim that is. If you do not know, don't spread fud, and if you think you know because you have an unstable setup, that perhaps is less unstable with HT disabled and therefore you conclude HT is definately the cause, and this phenomena has got to be universal, then that shows something about you as well.

>we're ALL FUDDING around.

Maybe you do not realize what the word FUD means.

>What FUD, I have no fear of many HT PC's locking.

By claiming there is a causal relation between HT and instability, you inflict fear (uncertainty, doubt), which is not warranted. Hence, you spread fud.

>PC's simply have fewer configuration issues with HT
>disabled since it adds to issues already there.

I've asked you before, and will ask you again: please show us half a shred of credible evidence hyperthreading results in instability, and it is therefore warranted someone owning or buying a P4 should be warry of enabling it.

>What is it about system issues that have got everyone so
>defensive?

its your false conclusions you keep posting as facts that get me offensive.

>You make assumptions about HT's perfection that is too
>intertwined to really support. I turn it on and more issues
>arise. Well, there seem to be more issues doesn't it?

Guess what; I've got an old Athlon system from a friend sitting here, with a Tbird 1400/133 FSB. When I set FSB to 133/266, it crashes every 5 minutes, sometimes it won't even boot. When I set it to 200x10.5 (1050 MHz) it works just fine. So, can I conclude there are issues with Tbird 1400 ? Using your logic, I could. People should think twice before setting it to the FSB is was designed for.. or risk instability. Makes sense no ? Unless, maybe, could it be the box has crappy cooling, inadequate powersupply or the memory doesnt support 266 DDR ? Maybe I didnt flash the bios, maybe I've got a KT133 that doesnt run at 133 Mhz ? Maybe even the chip is a 200 Mhz part and I should set a 14x multiplier ? There are plenty of reasons that cause this behaviour, and since AMD sold millions of those chips, and I've not read anywhere they do not work as advertised, common sense says something else is the problem.

>No, it simply doesn't do as much as people think it should,
>as supported by Tom's Hardware guide tests

Performance of HT has been known for a long time to varry somewhere from -5 to +15% on average. If you expected more (or less) blame your expectations. Besides, you really think this board needs your expert opinion on HT performance ? You reallly think posters here have not read two gazillion P4 reviews ?

> Is it FUD your afraid of (isn't any) or that the HT world
>has smears on its glass exterior you're somewhat ashamed of
>(shouldn't be)?

LMAO ! You have clearly not been around here for very long. I can't wait to be labeled an intel fanboy for the first time in my life; please make it happen.

> HT is perfect...right?

Not anymore than AMD64 or SSE is "perfect", its just another technology that helps performance under certain circumstances. Its cheap in die size, and has little downsides, there is nothing wrong with it.

>1.0 In a dual HT enabled system, is the motherboard seeing
>"four" CPU's? Two real / physical and two virtual?

Yes.

>2.0 If a pipeline prediction is incorrect and flushed, is
>the data in BOTH CPU's pipeline common to that processor
>flushed? If so, would this not have a double negative if
>too many wrong predictions were made?

No. Each logical processor has its own context, if the pipeline is flushed, this only affects one context. The other logical cpu is not affected.

>3.0 Would this would be a negative if you use short
>duration type desktop tasks verses longer consistent one
>where you have convinced me that HT really will benefit
>this user?

I have no idea what you're asking. What do you mean "longer task" ? I hope you do realize just opening a print dialog in word would translate in maybe tens of millions of clockcycles and the cpu might be interrupted and switch threads and contexts perhaps several thousand times between your click and the appearance of the dialog ? So please define "a long task".

>4.0 Dual core type HT enabled tasks are more server centric
>than desktop, possibly?

Tasks arent "dual core or HT enabled". Software can be multithreaded or not. Or you can run several single threaded cpu intensive apps at once. Some apps lend themselves well to multithreading, and for instance any half decent rendering program can be configured to spawn as many threads as you like. For other programs, like games, creating multithreaded code will be a lot harder. But again, there is nearly no difference between creating software that takes advantages of several physical cores or logical cores. Obviously, server software tends to be much more multithreaded friendly since it is designed to support several users simultanously, and designed to take advantage of SMP systems.

> Where are those application (most are not synthetic)
>benchmarks that show an Intel CPU beats an AMD in desktop
>use with HT enabled, and loses with it turned off?

What sort of nonsense criterium is that to judge a technology ?

> If we don't use scripted "real" applications what do we
> use to "prove" our point?

My point was never to claim hyperthreading enabled enormous speedups. My point is that it does NOT inflict instability like you claimed, unless with software that is so crappy that it will not run on SMP or CMP systems either. Creative Labs drivers are indeed a prime example or worthless coding, and are known to give problems on all sorts of configurations, which may include HT for all I know. If there is a lesson to be learned, it is to avoid Creative products at all costs, something most people here have known for ages.

Damn.. I don't think I ever wrote such a long post to defend an intel technology..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 20, 2004 8:08:39 PM

>I dont really see why you would think K9 or Meron would
>impliment a form of SMT, there arent really any hints at
>such a thing, especially on the amd side

Simply because it makes sense. Dothan doesnt have it, since its just a respin of the Pentium Pro which clearly didn't sport it yet. K8 doesnt have it because AMD already faced enough challenges as it was (HTT, ODMC, AMD64, SOI,...) and it would probably have had an even smaller impact as on P4, but I'm fairly certain as time goes by, and also AMD keeps on making longer pipelined designs that SMT will make sense on AMD cpu's just as well. The benefit may not be enormous, but neither are the costs. Pretty much all the low hanging fruit is gone nowadays, rest assured AMD is looking into CMT as yet another trick in the book.

Besides, the Risc world is usually also a good indication of things that will be on our desktop only a few years later. AFAIK, there is no upcoming risc cpu that doesn't sport some form or SMT.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
June 20, 2004 10:12:20 PM

P4man,

No, you create the conspiracy. I never said, "and this phenomena has got to be universal" you did. You seem afraid that any indiscretion that does indeed occur on admittedly rare occasions will bring the house down. You try to slam you opinion that it's all perfected out there and to dismiss those few that DO have issues. They are simply posting them to solve a configuration problem that DOES have to be addressed BECAUSE of a HT enabled situation.

If a configuration has to be changed to make anything work, it is indeed a fault of that component's interaction if all else worked fine before. Yes, it may indeed be software, but not UNLESS HT is enabled is it a problem. Users just want to know why! Me, I need to know why Outlook Express crashes the system. I don't think BIOS needs to be OE5 aware, or XP isn't. What is it?

I DO know HT causes instability when turned on, its gone when HT is turned off. What more need be said. HT crashes the system with Outlook Express. How on earth can I not know that?

I will not post what I don't know. I haven't posted what I don't know. HT crashes my system. I know that. You don't know why, fine. Someone else may. There are kernel fixes all over out there "fixing" HT problems. Why?

Installing an HT enabled CPU into an HT "compliant" PC under NORMAL set-ups (no silly outside of spec overclocking) doesn't always (it is rare - I seem to keep pointing this out to you) result in a stable system.

Instead of trying to bury the problem, start considering solutions. What would you tell a client with a PC YOU built with this problem? You're nuts? Go away? Don't tell anyone? Disable it and don't bother me? You don't get what you pay for?

Lets consider SOLUTIONS and not act like it's an Intel slam dance issue to undermine the beloved empire with FUD.
Take a deep breath and repeat, HT is not a universal problem, a very few group of users need solutions to infrequent HT enabled stability problems and would like to define solutions.

Going to your very useful side, and again appreciating your knowledge....
What I mean by short / long tasks is this,
We run a program that calculates SPC numbers over several product attributes and hundreds of data points over several frequency ranges. You load the parameters click "start calculations" and it crunches for about and HOUR. It is in the background with other tasks running up front. This is a LONG task with (I can't say it because I would plagiarize Carl Sagen) CPU cycles. Not a simple work document. Encoding audio seems to be a LONG time task running in many minutes to an hour. 10-15% can make a difference here, no?

I still don't understand (getting closer) how a single HT enabled CPU can empty out a bad pipeline prediction if the data from multiple tasks is "threaded" into the pipeline and multiplexed at the output into two serial and independent program outputs. It seems all the data in the pipeline would have to be trashed and start over with BOTH processes. It seems like two cars going down the same road when the prediction says, whoops, we guessed the wrong lead car. Well, both cars have to back out so the lead car can be replaced with the right car and the second car again goes down the road again behind the "correctly" predicted lead car. Or, is the lead car simply discarded somehow at the output and the second car accepted while a replacement is then "sent" for the wrong initial lead car?
June 20, 2004 11:34:49 PM

well wether that makes sense or not i up for debate. for one, its obvious amd didnt need HT , it still have the power to outperform a p4 in alot of areas, likewise for dothan in some situations.

i just think right now, neither of us can say that it will go etiher way, but i will say this, i have big doubts amd will look to implimenting HT anytime soon, since the id bet amd is looking into other architectures that dont focus on extending the pipeline, and there fore wont take advantage of HT. plus the other thing is that there would need to be a large compelling software need for HT that would push amd to do that. they are heading for dual core, that is thier answer to HT. sure they could add HT on top of that, but there is really no need to at this point.

i dont think that your theory doesnt make sense, its just that based on trends now, id say its improbable that will happen as soon as you would think. now if for some reason intel pushes Ht on dothan, or changes its mind aobut brining dothan to desktops, then things would be different, but right now that doesnt seem to be the case at all.
June 21, 2004 6:50:09 AM

>f a configuration has to be changed to make anything work,
>it is indeed a fault of that component's interaction if all
>else worked fine before.

And you are supposed to be a scientist ? Apparently my thunderbird example was too difficult for you to understand, so lets try a simpler example to show your logical phallacy: my systems works fine, great, no problems, but when I run Prime95 it crashes. when I don't run prime95, it doesnt crash. Surely, prime95 has got to be the reason, no ? It doesn't matter that prime runs fine to tens of millions of computers, I don't need to know that. Not running prime=ok, running prime=crash, that's all I need to know. People, don't run prime cause it can crash your computer. Never mind the fact my cpu is overclocked by 50% and only exihibits this instability under extremely heavy load. Never mind the fact that other, similary taxing software would likely crash my computer just as well, never mind thinking or using logic.

>I DO know HT causes instability when turned on, its gone
>when HT is turned off. What more need be said. HT crashes
>the system with Outlook Express. How on earth can I not
>know that?

Because you attribute the cause to merely a symptom. You blame HT as if you'd have tried eliminating all other possible factors, yet you haven't even tried. Does the problem still ocur with the latest bios ? does it occur with a different PSU ? Does the problem still exist when you removed all virusses and spyware from your machine ? Does you your computer still crash when you disable certain periphirals or drivers ? Do those periphirals and drivers work on a 2 way SMP machine ?

Here is an educated guess: remove the soundblaster and drivers from your machines, and watch how it magically cures all problems. Now, would you say the problem is HT and not the SB drivers ? Even when any other soundcard will work just fine, with or without HT and the SB with the same drivers will cause the same lockups on otherwise completely stable 2 way workstations, servers, etc ?

> will not post what I don't know. I haven't posted what I
>don't know. HT crashes my system. I know that

No you <b>don't know</b> that. You wrongly <b>assume</b> that. enabling HT reveals a problem, but you have no way of knowing it is the cause. I tell you it is NOT the cause, and I challenge you again to provide any credible source that shows it is or could be. I can give you two million links of problems with soundblaster drivers, crappy psu's, spyware, badcap motherboards, BIOS issues, etc. Its fair to say it is known these things can cause problems, I have not ever read HT can, nor have I read dual cpu systems are known to be unstable.

>Instead of trying to bury the problem, start considering
>solutions.

Actually, you're the one that should have asked "what could cause this" instead of simply stating "HT can cause instability" as if it where a known fact. You're the one making the wrong assumptions on which I called you. I gave you plenty of possibilities allready.

>What would you tell a client with a PC YOU built with this
>problem?

Probably something like: "I didn't ship you that system with that soundblaster POS, why did you replace the soundcard ?" or "Why the hell did you flash your BIOS to a three year old one ?".

> You load the parameters click "start calculations" and it
>crunches for about and HOUR. <snip> 10-15% can make a
>difference here, no?

So what is your question or point ? Yes, 10% of an hour is 6 minutes.. ?

>I still don't understand (getting closer) how a single HT
>enabled CPU can empty out a bad pipeline prediction if the
>data from multiple tasks is "threaded" into the pipeline
>and multiplexed at the output into two serial and
>independent program outputs. It seems all the data in the
>pipeline would have to be trashed and start over with BOTH
>processes.

Why ? at any given stage in the pipeline, there are only instruction from one thread. why would you flush those when another thread hits a branch misprediction ? Using your (flawed) analogy the cars on just on different roads, and one doesn't follow the other one at all.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
!