Intel abandoning HyperThreading??

[purrkur]

In this article, The Inquirer asks why HT isn't a part of Intels future roadmap. I for one have always regarded HT as a marketing driven technology (similar to the GHz race that intel pushed with the P4 marketing) so I am not surprised to see this happening.

I know that many stand by HT as being excellent technology and something that does increase performance and I do realise that it some cases it does, but I have also seen it decrease performance for certain types of apps just as much as it would increase for others. In my experience, the mainstream user doing all sorts of computer work will only see a few percent increase in overall performance with HT. I have been debated on this issue before but I think that Intel shutting down this technology finally proves me right. It simply isn't worth implementing since the average joe is lost on this technology level anyways so marketing doesn't gain much.

Am I stepping on any toes here??

Fwiw: Intel seems to be dumping a lot of technology. They have pretty much said that the Pentium 4 technology has reached its end and it won't be used in future products (see my rants elsewhere on why I hate the P4). It also seems as if they are in trouble with the Itanium line of products. Interestingly enough, the world and their dog knows that AMD has hit home run with integrating the memory controller into the CPU. Intel is rumored to be moving engineers from the Itanium line to Xeon in order to do the same but they won't have anything out on the market until 2007.

The only thing Intel seems to be doing right these days is their development of low-power based products like the Excellent Pentium M....

[davd_bob]

I am not surprised to see this happening.

I agree that it was more "blow then go," but since Intel was able to turn the HYPE into bucks harvested from ignorant buyers ...I am kinda supprised they would drop the HT cash cow.

I know that many stand by HT as being excellent technology and something that does increase performance and I do realise that it some cases it does, but I have also seen it decrease performance for certain types of apps just as much as it would increase for others. In my experience, the mainstream user doing all sorts of computer work will only see a few percent increase in overall performance with HT. It simply isn't worth implementing since the average joe is lost on this technology level anyways so marketing doesn't gain much.

I agree, but the mainstream user usually has no clear concept of preformance anyway , so they believe whatever Intel tells them.

I have been debated on this issue before but I think that Intel shutting down this technology finally proves me right.

Am I stepping on any toes here??

C'mon purk...
Theres no I in team.
It would be also nice if the folks with big toes you mentioned would be more intrested in just spreading the truth and less in taking credit for knowing what the truth was.

Fwiw: Intel seems to be dumping a lot of technology. They have pretty much said that the Pentium 4 technology has reached its end

You mean the Pentium Pro is finally at the end of its life cycle...

The only thing Intel seems to be doing right these days is their development of low-power based products like the Excellent Pentium M....

Intel DOES run cooler then American Micro Devices.
But thats about the only thing they do better then AMD.

How about this for a statement...
AMD Intel

[Billl]

In this article, The Inquirer asks why HT isn't a part of Intels future roadmap. I for one have always regarded HT as a marketing driven technology (similar to the GHz race that intel pushed with the P4 marketing) so I am not surprised to see this happening.

Acually you couldn't be more wrong. If you were actually doing something math intensive like S.E.T.I. you would know that HT is far from some marketing ploy. My P4 Northwood blows anything AMD ever had in it's class right out of the friggen water! It literally does half again as much work as the AMD.

I know that many stand by HT as being excellent technology and something that does increase performance and I do realise that it some cases it does, but I have also seen it decrease performance for certain types of apps just as much as it would increase for others. In my experience, the mainstream user doing all sorts of computer work will only see a few percent increase in overall performance with HT. I have been debated on this issue before but I think that Intel shutting down this technology finally proves me right. It simply isn't worth implementing since the average joe is lost on this technology level anyways so marketing doesn't gain much.

Did it ever occur to you that maybe it's just not needed since you now have two real cores? Lots of technology gets left out as you move on to the next thing. It's called progress. One thing you fail to consider here, with Intel you at least had the ability to use it if you wanted to or turn it off. With the AMD's your stuck with what you have. Can't turn on what isn't there! I wouldn't be suprised to see HT return in the future. Don't count it out as dead just yet.

Am I stepping on any toes here??

Not actually stepping on any toes. I have nothing against either manufacturer. One thing I've found over the years, Intel's chips tend to be much better over clockers. Intel is much more conservative in binning their chips. If I'm running a mission critical server whose chip do you think I want in my machine? Probably why Intel has the lion's share of the servers out there. Personally I want a chip that over clocks well. But that very fact makes Intel in most cases the hands down winner for over clocking. But if my business depended on my servers up time you can bet your ass an Intel chip is going to be in it!

Fwiw: Intel seems to be dumping a lot of technology. They have pretty much said that the Pentium 4 technology has reached its end and it won't be used in future products (see my rants elsewhere on why I hate the P4). It also seems as if they are in trouble with the Itanium line of products. Interestingly enough, the world and their dog knows that AMD has hit home run with integrating the memory controller into the CPU. Intel is rumored to be moving engineers from the Itanium line to Xeon in order to do the same but they won't have anything out on the market until 2007.

I must of missed your P4 rant. But I can't understand why you hate them. Unless it had something to do with the way they were kicking AMD's ass until the 64's showed up? It seems perfectly reasonable that Intel would be dumping technology. It's called progress, out with the old in with the new. In my opinion the P4 Northwood was one of the most successful chips Intel ever produced. That said I'm very impressed with AMD's 64 bit chips. Intel really has it's work cut out to regain it's leadership role. If it ever can.

The only thing Intel seems to be doing right these days is their development of low-power based products like the Excellent Pentium M....

Well at least we can agree on one thing. My guess is you will see those cores in the next multi core chips. It should get interesting. We all benefit from this battle. Personally I'm glad AMD is keeping Intel honest. Just imagine what chip prices would be if Intel had the market all to themselves. One thing that baffles me though is a lot of people think AMD can do no wrong. What makes you think AMD wouldn't be putting it to us just like Intel would if it could? Yet a lot of people here act like AMD is the savior. Kind of cracks me up. AMD is a company just like Intel and they want to make as much money as they can. Wake up.

Billl

[davd_bob]

NNNOOOOOOOOO,

Guys,
Be nice.
Otherwise the OOH flamewar, I mean, discussion is down the hall.

BTW,
I think you both know more then me on most PC topics.

[Billl]

NNNOOOOOOOOO,

Guys,
Be nice.
Otherwise the OOH flamewar, I mean, discussion is down the hall.

I don't see this as a flame war at all. Were just expressing our opinions. We just don't agree on a lot of things


Billl

[Wolfram]

Wouldn't a system with a combination of dual cores and Hyperthreading be even more responsive?

[purrkur]

Sorry for the extremely late reply guys! I have been so busy that I haven't had time for the longer rants

I agree that it was more "blow then go," but since Intel was able to turn the HYPE into bucks harvested from ignorant buyers ...I am kinda supprised they would drop the HT cash cow.

Well, the truth comes out sooner or later. In this case I think HT is widely overrated for general performance increases. Sure, there are apps out there that will use it well, but others won't. The balance means you are not getting much out of it.

I agree, but the mainstream user usually has no clear concept of preformance anyway , so they believe whatever Intel tells them.

That is unfortunately true in many cases. There is a lot of ignorance out there as well. I have spoken to clueless admins that buy Intel just because nobody has ever been fired for buying Intel.

Intel DOES run cooler then American Micro Devices.

I think I have seen this from you before but you never specify exactly what CPU'syou are talking about. It is like saying something like "BMW is faster than Mercedes". Without specific models the argument isn't really stable. If you want to talk the latest and greatest from both manufacturers then you are wrong. You can see that here where Toms Hardware says:

As far as power goes, Intel is an inefficient energy-guzzler with up to 30 percent more power consumption than the AMD system.

This when comparing the two top dual core systems from both manufacturers. You can also see more of the same here if you look at the facts and figures on that page.

How about this for a statement...
AMD Intel

Hehehe! Don't get me wrong here. I hate Intel just as much as Billl hates AMD, i.e. not at all. But I am seriously tired of how their marketing always gets to have a say in their technological development, and how they rush products to market and how they will brainwash you with their marketing only to change their minds a bit later on down the road. I would understand this if they were a small time underdog but they are actually the market leader in x86 hardware. They should try to act like it too.

Acually you couldn't be more wrong. If you were actually doing something math intensive like S.E.T.I. you would know that HT is far from some marketing ploy. My P4 Northwood blows anything AMD ever had in it's class right out of the friggen water! It literally does half again as much work as the AMD.

I wrote:

In my experience, the mainstream user doing all sorts of computer work will only see a few percent increase in overall performance with HT.

I should have been a bit clearer. I know that certain applications do see a sizeable increase in performance. I also know that certain applications will suffer. If you are doing more with your computer than running a single app then the pluses and minuses will balance each other out. I do believe that you do get an increase in performance, but not nearly by as much as Intel wants users to believe. Anandtech has given their take on what I am talking about here. Believe me when I say I am not entirely dissing the technology. For certain apps and simple repetitive tasks it can give sizeable increases (as seen here at Toms Hardware). However, I don't have a single computer here at home or in the office that simply does a single such repetitive thing and nothing else. My point is that I think that the cost and complexity of this is not worth it, or better spent elsewhere. But you never know, like you said, it could resurface in the future once Intel gets it right.

Did it ever occur to you that maybe it's just not needed since you now have two real cores?

Did it ever occur to Intel? Why are they marketing an "EE" CPU with dual cores and HT on both cores?

Lots of technology gets left out as you move on to the next thing. It's called progress. One thing you fail to consider here, with Intel you at least had the ability to use it if you wanted to or turn it off. With the AMD's your stuck with what you have.

Yep, progress is what AMD is pulling out of its hat and Intel is trying their best to follow in their footsteps. AMD developed the 64 bit extensions. Intel copied, called it something else the funny thing is that when you run 64 bit code on their CPU's, it will run slower than when you run the same apps as 32 bit code. The opposite is true for AMD, as you can see here. This leads me to the conclusion that Intel rushed its 64 bit implementation out to market and making a poor job of it too. As for dual core, Intel has already admitted that they rushed that out to market as well. And we all know how Prescott was going to bring Intel into the +4GHz area. That never happened. Rushed to market???

That said I'm very impressed with AMD's 64 bit chips. Intel really has it's work cut out to regain it's leadership role. If it ever can.

.

Oh, I am sure they can get a leadership role again. They are a huge company with huge resources. I have been watching their roadmaps and what I see is both exciting and interesting. All they need to do is to get marketing out of the development area and get back to making good products. The technology based around the Pentium M's is definitely a step in the right direction and that seems to be where they are headed.

One thing that baffles me though is a lot of people think AMD can do no wrong. What makes you think AMD wouldn't be putting it to us just like Intel would if it could? Yet a lot of people here act like AMD is the savior. Kind of cracks me up. AMD is a company just like Intel and they want to make as much money as they can. Wake up.

Hehehe! I'm awake! I realise that you think I am an AMD fanboy but I am not. I swear by my Pentium M/Centrino powered IBM T42 notebook

I guess what pisses me off is that as soon as AMD is having problems, (for example in the early Athlon days when VIA was making shitty chipsets for them), everybody was quick to point out that AMD should be ignored and left for dead. As I have pointed out above, Intel has been making a lot of mistakes lately but I don't hear anybody saying that Intel shouldn't be used. I am not saying that anybody should, but I would like to see a fairer discussion of these manufacturers. I also don't like the way Intel sometimes acts a la Microsoft. They do push their weight around too much at times when they have messed up, instead of admitting that they have messed up. Right now they are even having stability problems with third party chipsets, similar to what we saw in the early Athlon days!

As for AMD, yep they are the underdog and in the high end CPU's they do better (while Intel owns the low power segment) but you are right, if AMD was the big bully and Intel was the underdog then I am absolutely sure they would act the same. Pretty much like Apple would if they were the size of Microsoft

As a final note, I would like to say that I don't like to mention work all to much but we have been doing tests on Intel and AMD hardware (latest and greatest) with both Windows and Linux. I can definitely say that in a 3-tier setup under extremely heavy load, the top stuff from AMD will seriously outperform anything from Intel. This to a degree that we thought we were doing something wrong! However, both HP and IBM technicians have verified our tests and also said that they have seen plenty of the same elsewhere. The integrated memory controller in the AMD chips is simply a technological delight and Intel doesn't have anything that can remotely compete, until 2007 when they say they will bring out similar technology.I can also mention that both AMD and Intel hardware ran flawlessly during our tests. Neither platform showed any signs of unstability ever and we flogged them long and hard.

I don't see this as a flame war at all. Were just expressing our opinions. We just don't agree on a lot of things

Agreed I think we do agree on more things than our posts show actually

Wouldn't a system with a combination of dual cores and Hyperthreading be even more responsive?

In some cases it would but it could also mess things up if your threads end up in a hardware resource conflict.

[purrkur]

I finally found a couple of articles that talk about what I have known all along. This article called "Hyperthreading hurts server performance, say developers" goes on explaining how Citrix servers as well as SQL Server 2005 see degraded performance when used on high end hardware using Hyperthreading. The article points out this weblog from a Microsoft developer who tested this.

Well it is about time that somebody else supports me on what I have known all along! I have just ordered two servers with dual EM64T 3.0GHz Intel CPU's supporting hyperthreading, 4 gigs of memory etc. I will be testing SQL Server 2005 in 32-bit and 64-bit versions on these two machines. I have decided that I am going to do official tests on this for our applications and I am pretty sure what the outcome will be (since I have seen this happen on our servers using SQL Server 2000 as well as our maths intensive clients).

Before you start flaming, note that I have said that HyperThreading gives a performance increase when running simple multithreading stuff where the threads do not end up fighting for resources (such as ALU or internal cache). However, that isn't really representative for majority of computers/servers that are doing multiple operations at a given time.

I firmly stand by what I have said about Intel Hyperthreading being a marketing driven technology and not a engineering driven technology. I also stand by what I have said about other implementations of Hyperthreading from IBM and SUN working the way they should. It is just Intel's implementation that sucks. It has a feel of being rushed out the door before it was fully implemented which Intel seems to be doing a lot these days.

[hyperspace]

HT => HyperMarketing

When you create a virtual processor from "within" the physical processor, you are not really getting a true perfomrance gain. Multi-Core with dynamically shared cache and memory for each core, is the way to go.

Haven't heard a whole lot about IBM or Sun.

[purrkur]

Haven't heard a whole lot about IBM or Sun.

You can find information on Sun here and here.

You can find information on IBM here and here.

Note that on other CPU's, the technology is called SMT or Simultaneous MultiThreading and sometimes even Hyperthreading.

Ars Technica also had a good description of how IBM implemented SMT over here.

[purrkur]

Oh, and I found mention of this on Slashdot (over here). I loved reading the comments. Now everybody is saying that hyperthreading is crap all of a sudden

[KliK]

Well the HT seems like a nice technology...and the SMT seems to have advanced it (in the previous articles by 40%)...so it is the way to speed up the SingleCore CPU...but, my opinion is that Intel hasn't really developed it properly and release it before the complete testing ('cause of the AMD competition & FLIPs-MFLOPs in P4 systems!)...

now they are abandoning it, 'cause:
1. if development requests millions or billions of dollars
2. they didn't get the HT to speed up the more tasking programs (SQL Server or sthg similar)
3. they use the MultiCore CPU...


problem here is, there isn't so much programs that can gain from using the MultiCore CPU (they are not written to execute MulitiCore CPU or Multi CPU tasking!!!)...how much of us is using the BP6 for computing or servers?! and on the load (I use the UD agent which computes LIGANDFIT or ROSETTA or SMALLPOX research!) i don't see that my CPU consumtion goes to 100%...it is around 60-70% and variates...
how is it with the SETI@home?! maybve some one can fill me up?!

[Billl]

Oh, and I found mention of this on Slashdot (over here). I loved reading the comments. Now everybody is saying that hyperthreading is crap all of a sudden

I guess it all just depends on the use. For me HT has been a blessing. Crunching SETI it sure makes a big difference.


Billl

[Billl]

problem here is, there isn't so much programs that can gain from using the MultiCore CPU (they are not written to execute MulitiCore CPU or Multi CPU tasking!!!)...how much of us is using the BP6 for computing or servers?! and on the load (I use the UD agent which computes LIGANDFIT or ROSETTA or SMALLPOX research!) i don't see that my CPU consumtion goes to 100%...it is around 60-70% and variates...
how is it with the SETI@home?! maybve some one can fill me up?!

I run 2 processes on my P4's and each takes 50% CPU usage. I get about 70 - 80% improvement over my none HT AMD machines running the same clock speed. Not sure if that answered your question?


Billl

[KliK]

problem here is, there isn't so much programs that can gain from using the MultiCore CPU (they are not written to execute MulitiCore CPU or Multi CPU tasking!!!)...how much of us is using the BP6 for computing or servers?! and on the load (I use the UD agent which computes LIGANDFIT or ROSETTA or SMALLPOX research!) i don't see that my CPU consumtion goes to 100%...it is around 60-70% and variates...
how is it with the SETI@home?! maybve some one can fill me up?!

I run 2 processes on my P4's and each takes 50% CPU usage. I get about 70 - 80% improvement over my none HT AMD machines running the same clock speed. Not sure if that answered your question?


Billl

actually i was asking about the BP6 system...just for comparison with mine!

[kuun]

this was such a long and informative thread... so long and so many pages to go off to.. i forgot.. i THINK someone said in this thread that Intel overclocks better than AMD, and that AMD run hotter than INTEL.

i don't remmeber who or where..

but... From my experience in overclocking...

overall AMD seems to be the better overclocker.. as you can reach higher FSB speeds with some AMD chips by unlocking the multipliers (i've done a few chip mods myself) and with higher FSB, you gain a performance boost with the faster data transfer from the ram to the CPU.. BUT.. certain AMD CPU's i have noticed do not overclock very well.. and these seem to be the mass produced chips that jsut have the multipliers changed and locked from one chip speed to the next.. kinda like the celerons did abck in the day.. My Xp2200+ does NOT overclock very good at all.. but then again i haven't tried to OC with the new ddr400 ram i put in it either.. It seems that the Duron line are some of the best overclockers (the Applebreds are upgraded durons) while it also seems the XP line were the worst overclockers..

again.. this is from my personal experience..

Intel seem to be general good overclockers.. but don't have as much gain (im talking percentage wise) on processors as some AMD's do. Intels also seem to be more stable when under continuous heavy load (SETI, or intense gaming such as Quake or Unreal Tournament for example) This is why i would personally use Intel for server applications that use the CPU's constantly.. this is also why i use the celerons to power my Storage server.. But for general gaming and internet browsing, playing videos etc.. AMD is my choice..

I use both, for different reasons..

As far as AMD being hotter.. my XP2200+ does run hotter than my ex-fiance's p4 1.5g and my p3 1G and the old celeron2 900 as well as my grandfathers p4 2.4g pre HT cpu (i don't know whats in it cause it's not mine) BUT.. the temperaturs difference is only by 5C at best. hell the XP2200+ runs hotter than my duron 1.2g... and my duron actually runs coller than msot other CPU's i've used.

again.. this is all from personal experience.. i have no websites or benchmarks to prove anything.. this is a matter of opinion..

Intel make good server CPU's, while AMD are better general all rounders..

Both are good at what they do.. and suck at what they don't

i have read reports on the pentium M's.... i must say.. the results are indeed impressive