Intel doesn't have a REAL dual core processor.

[dbentley1267]

I have read through most of the posts here

some are rather ridiculous some were good etc...

but i have been in the hardware and software testing for the last 7 years, 2 at big blue that is ibm for those who do not know and 5 here at xpoint

as far as many comments on intel vs amd processors i would not even know where to begin

i am not a "processor guru know it all" like some here but i have a few observations if you do not mind

first i think dual core anything is rather cool from both camps, and this is why

i sort of got spoiled when i was at ibm, i have owned so many dual cpu systems it is not funny (honestly)

when all the fuss over pentium 3 500, amd k6 450 - 500 etc, my friends in the gaming community were just crazy and everyone was bragging who was the best

that is until i walked in with a dual pentium 3 xeon 550 with 2mb of cache per chip and 512 mb ecc and a 64 mb quadro 2

now who is talking etc...

well i have owned

1 dual pentium pro 200
4 pentium 2/3 400/500/677/800 (ibm m pro, tyan, intel)
3 pentium 3 xeon 550 2 mb l2 cache (IBM z pro)
2 pentium 3 xeon 866 (ibm z pro and sgi)

now where i fell off of the band wagon was in the advent of the new xeon's they were much too expensive so i got stuck running a single core pentium 4

this is where i got real ticked off at intel and amd amd stopped their short run of what was ahtlon mp?? and intel did not enable smp on their p4

so true computing was left for those with big pockets

it does not matter how fast your single processor is at some point it will come to a grinding halt, single core single thread and you know what it happens to both amd and intel (PERIOD)

i was so spoiled at having smp systems that having the true ability to run many tasks at the same time and still have a system that is usable

now some one was talking about motherboards, who is best etc

i have run oh i dont know 10 or 12 intel motherboards at least abit 2 (they were ok) and tyan 3 (all were bad), aopen (4 we wont go there I was really disappointed) gigabyte too many

the most stable motherboards were always intel i ran at least 6 dual boards and 4 single boards and you know what they were rock solid

sure i could not over clock and maybe i missed some little frill or two but go back and read many of the reviews out there

you will find that almost all of the reviews say the same thing intel boards are rock solid stable

now as far as amd, well i was one of those early birds who got a real bad taste in my mouth, i remember having to have special order power supplies, there were a ton of driver issues motherboard stability problems etc...

now i know these have been worked out and things are way better than before but even today who is best

does it really matter that i can run a bench mark 20 seconds faster

or that i can run 10 fps more

what about power? oh that is right i guess we got consumption meters on out computers so we know how much of our light bill is from our pc

my my my, where have we gone? i read through the posts and some of you guys are fighting like little kids (my dad is bigger than your dad)

WOW

and some of you just get down right vicious

so who is the best?

well as i read it amd wins the war on benchmarks

they used to win on price

but again what are we buying into, what are we really getting?

it boils down to this

10 seconds that is a whole lot!!!

20 fps more, gee i thought i remember reading somewhere that 40 fps was max we could see a difference (correct me if i am wrong)

20 watts of power

etc etc etc etc.....

hey guys let us have fun

that is what life is about aintit

 

[ssvegeta1010]

Originally posted by: dbentley1267
20 fps more, gee i thought i remember reading somewhere that 40 fps was max we could see a difference (correct me if i am wrong)

We can see much more than 40 fps. This issue has been beaten to death in Video.

 

[TuxDave]

Originally posted by: clarkey01

Originally posted by: dmens
yeah well, didnt have much to work with. stupid netburst. we all know smithfield is not a clean solution. but its still dual core. get that into your thick skulls, fanboys.

Funny you rip your own core but it'll sell millions anyhow....If only CPU's sold by thier performance hay.

Hey, just because you work for a company doesn't mean you have to fall in love with every aspect of it. Otherwise that would be an 'Intel' fanboy right??? And I'm sure how much you all hate Intel fanboys. You guys should appreciate that he will give you the technical aspects of Intel while at the same time realizing its shortcomings.

 

[TuxDave]

Originally posted by: harpy82

Originally posted by: dmens
sorry to break it to ya, but there's arbitration logic on the smithfield die. its stupidly primitive since it was a rush job, but its there. and stop redefining the term "dual core", or using "true dual core". get that revisionist crap out of my face.

as for X2's inability to be chopped into single cores, thats just too damn bad. good luck on yields.

wat's so bad about not chopping down dual cores ??? dual cores are meant NOT to be chopped down..... if u can ( like intel as they said)... it is not dual core...

stop trying to make urself sound right.... the more u do.. the more nOOb u sound like....

Why aren't dual cores supposed to be able to get chopped? That's like saying processors are not supposed to be speed binned. It's really a big waste of money if you design a dualcore system and if one core has a defect that you must throw the entire thing away. If you can design it and have some fuses to disable the faulty core and make the processor act independantly, that's a good business plan and good design.

 

[wasserkool]

Intel's dual core suffers severe bandwidth bottleneck when communicating between the two cores. a AMD X2 will kill it any day any time.

For those intel fanboys here bashing X2, they are just jealous that their sorry little DUAL CORE CraPU from intel gets murdered by the X2, I don't feel sorry for you, please DIAF

 

[Viditor]

I think the terms "Twin Core" and "Dual Core" have some merit...you could become more wordy and call them "Integrated Dual Core" and "Non-Integrated Dual Core", but it amounts to the same thing.
Obviously the integrated variety is going to always outperform a non-integrated one all other things being equal. Things being equeal is the key...there are apps that even the single core Intels accel at, and they still do whether they are single or twin (whatever).

The term "real" is strictly a marketing one, and it has no bearing in this forum...I remember when someone said that AMD64 wasn't a "true" 64bit chip...more hogwash. Let the marketing guys fight it out...in the meantime, we know the difference.

 

[TuxDave]

Originally posted by: Viditor
I think the terms "Twin Core" and "Dual Core" have some merit...you could become more wordy and call them "Integrated Dual Core" and "Non-Integrated Dual Core", but it amounts to the same thing.
Obviously the integrated variety is going to always outperform a non-integrated one all other things being equal. Things being equeal is the key...there are apps that even the single core Intels accel at, and they still do whether they are single or twin (whatever).

The term "real" is strictly a marketing one, and it has no bearing in this forum...I remember when someone said that AMD64 wasn't a "true" 64bit chip...more hogwash. Let the marketing guys fight it out...in the meantime, we know the difference.

If by 'twin core' you mean two different dies in one package, then there should be a distinction. But as far as I know that's a pretty weird way of doing dual core design and I'm pretty sure both companies have them integrated on the same die. (can we pop off the heatspreader to check anyways?) If you mean 'twin core' meaning there's a lot of copying and pasting... well I'm not really sure about the details of AMD's design process but I can make a couple educated guesses. When you have a dual core design, there's bound to be lots of copying and pasting. They won't make hire two teams to build their own FPU for each core when they can just build one and copy it over to the other. It's simply a big waste of time, people and money in my opinion but hey, what do I know. Look at the X2 diagram on Anand's review. How much in that big box labeled 'CPU 0' is the same as in 'CPU 1'?

HOWEVER, the companies will have different ways to manage the two cores and I think that's what made the difference between AMD X2 and Pentium D. So you're absolutely right on the use of marketing terms here. Marketing terms are for non-technical people. Performance numbers don't lie (usually) and whoever is faster is simply faster. Bottom line.

 

[Viditor]

Originally posted by: TuxDave

If by 'twin core' you mean two different dies in one package, then there should be a distinction. But as far as I know that's a pretty weird way of doing dual core design and I'm pretty sure both companies have them integrated on the same die. (can we pop off the heatspreader to check anyways?) If you mean 'twin core' meaning there's a lot of copying and pasting... well I'm not really sure about the details of AMD's design process but I can make a couple educated guesses. When you have a dual core design, there's bound to be lots of copying and pasting. They won't make hire two teams to build their own FPU for each core when they can just build one and copy it over to the other. It's simply a big waste of time, people and money in my opinion but hey, what do I know. Look at the X2 diagram on Anand's review. How much in that big box labeled 'CPU 0' is the same as in 'CPU 1'?

HOWEVER, the companies will have different ways to manage the two cores and I think that's what made the difference between AMD X2 and Pentium D. So you're absolutely right on the use of marketing terms here. Marketing terms are for non-technical people. Performance numbers don't lie (usually) and whoever is faster is simply faster. Bottom line.

It all comes down to cache coherency and MOESI protocol...
Apps on multi-core systems (or multi-cpus) will often need access to threads available in the cache of another core/processor. With an integrated DP, the access takes one clock because it travels directly to the second cache through the crossbar...with a "twin core", the access takes 6-7 clocks because it must be sent back to the FSB and back again. There is also the question of the FSB acting as a bottleneck (because all other traffic goes through it as well).
So, my use of the term "twin core" refers to 2 cores (whether they're on the same piece of silicon or not is less relevant) not directly connected. A "dual core" is where they ARE directly connected (either through a crossbar as in the AMD solution, or via a shared cache as in the future Yonah solution).

 

[dbentley1267]

EDIT: I have also read that DFI has a BIOS release for their i865PE LGA775 board that includes the following changes:

Major Reasons of Change:
1. Added C1E BIOS control.
2. Support EM64T CPU.
3. Support Dual Core CPU.
4. Enhance the support of SATA/SATAII HDDs.[/quote]


where di this come from? when i read this i went to dfi and all is correct except the dual core???

HELLO

 

[TuxDave]

Originally posted by: Viditor

Originally posted by: TuxDave

If by 'twin core' you mean two different dies in one package, then there should be a distinction. But as far as I know that's a pretty weird way of doing dual core design and I'm pretty sure both companies have them integrated on the same die. (can we pop off the heatspreader to check anyways?) If you mean 'twin core' meaning there's a lot of copying and pasting... well I'm not really sure about the details of AMD's design process but I can make a couple educated guesses. When you have a dual core design, there's bound to be lots of copying and pasting. They won't make hire two teams to build their own FPU for each core when they can just build one and copy it over to the other. It's simply a big waste of time, people and money in my opinion but hey, what do I know. Look at the X2 diagram on Anand's review. How much in that big box labeled 'CPU 0' is the same as in 'CPU 1'?

HOWEVER, the companies will have different ways to manage the two cores and I think that's what made the difference between AMD X2 and Pentium D. So you're absolutely right on the use of marketing terms here. Marketing terms are for non-technical people. Performance numbers don't lie (usually) and whoever is faster is simply faster. Bottom line.

It all comes down to cache coherency and MOESI protocol...
Apps on multi-core systems (or multi-cpus) will often need access to threads available in the cache of another core/processor. With an integrated DP, the access takes one clock because it travels directly to the second cache through the crossbar...with a "twin core", the access takes 6-7 clocks because it must be sent back to the FSB and back again. There is also the question of the FSB acting as a bottleneck (because all other traffic goes through it as well).
So, my use of the term "twin core" refers to 2 cores (whether they're on the same piece of silicon or not is less relevant) not directly connected. A "dual core" is where they ARE directly connected (either through a crossbar as in the AMD solution, or via a shared cache as in the future Yonah solution).

Ah, so the only distinction is the front-end handling. I still find the 'twin core' vs 'dual core' distinction not really important. In circuit design, the 'dual core' theory was to build two fully functional processors (or whatever) where each individually is slower than what you could build with a single core. However for the same performance (given software optimizations) you should be able to achieve a lower power target.

(for simplicity... #1 exclude leakage and #2 exclude multi-processor overhead.)

Power ~ Voltage^2
Maximum Clock Frequency ~ Voltage

So you reduce voltage 70%, get 50% power, build two cores to get back to the original power envelope but now you have two cores running at 70% speed and given the right application, that may beat a single core at 100% speed.

I agree though, you should have a good front-end handling of the cores like what AMD has done in the X2 and be able to salvage the die if only one core was faulty. That's good design IMO.