AMD vs Intel

[jlswier1988]

Ok so i hear AMD does there processing models as whati t runs at like a Intel 3.2ghz so basically what i want to know is how a AMD 3200+ is stock at about 2.0ghz can compare and run just as well as a Intel 3.2GHZ prescott processor.. can someone please explain this to me

 

[InlineFive]

The AMD processor has more pipelines which are shorter. In simple terms:

AMD processors = 9 cylinders
Intel processors = 6 cylinders

There are more factors but that's a chunk of it.

 

[jlswier1988]

So does that also factor in prices?

 

[shoRunner]

performace = ipc * frequency

amd has a higher ipc vs intel

 

[Suspicious-Teach8788]

If you're talking 3200+ vs P4 3.2, that's what... $190 vs. $220 respectively?

To make it short, I pick AMD any day in this case.

I'll be honest. I was an Intel fan up until early this year. I refused to believe that AMDs were that good, but seriously. The 3200+ kicks the 3.2 anyday around. Furthermore, you can overclock the 3200 (2.0) up till 2.8 GHz on STOCK COOLING. A 2.8 GHz is what.. 4200+? Hahahaha. Add in some whacky cooling I'm sure you can get higher.

How high does a 3.2 go? Not very. Stupid Prescott cores.

Athlon 64 is just far more efficient and so it demolishes the P4 in terms of IPC. The pipeline of a K8 is 12 stages long. A Prescott has 31 stages I think? Wow. It also has a crazy branch predictor that makes it so friggn hot.

 

[Duvie]

IPC = Instructions per cycle (clock).....

Therefore amd is handling at 9ipc versus 6ipc of INtel about 1.5x more work per each clock cycle...hence why a 2.0ghz model can equal around a 3ghz intel...again 1.5x....Intel can make up for this with things like SSE codes and what not, but since AMD has a cross licensing agreement with INtel it usually implements the same features...Itel usually implements early on when there is little software coded for it to give it an y significant advantage. AMD usually adds it a few core revisions later to its cores and thus usually comes into market when apps actually start using them.....

There are other factors that can also hurt evenmore...the fact is the tremendous length of INtel's pipeline has inherent issues. The pipeline is so long it adds latency. It also means that if the cpu has a branch misprediction the penalty grows even greater, but all this while the pipeline has been idle waiting for more data to be processed IE nothing is getting done...In comes Intel's "crutch"...HT Hyper threading.....without this AMD would be kicking the ass-of Intel in EVERY SINGLE benchmark...the only benches intel currently wins can be attributed to HT....Some can give increases in performance ranging from 1-20%+....Basically what it does is allow the cpu to not wait for those idle times and keep the pipeline full by trying to implement more then one thread simultaneoulsy...The data is ready to go to fill in if there is a branch misprediction and idle time....

 

[Zebo]

Duvie A+ explained it and the Calcs are real simple:

Skt 754 512KB L2 Athlon A64 = speed in Mhz x 1.5 = Prescott speed
Skt 939 512KB L2 Athlon A64 = speed in Mhz x 1.55 = Prescott speed
Skt 939 1024KB L2 Athlon A64 = speed in Mhz x 1.57 = Prescott speed

This is a generalization across all testing and is accurate. AMD will be a little stronger than these multipliers represent in some things, Intel in others. But in overall testing it's correct.

EX: Approx how does a 2600 Mhz winchester compare to P4? Using above numbers: 2600 x 1.55 = 4030mhz P4

 

[Duvie]

Got an idea on a dual X2 setup Zeb??? I mean how do you calc in the inherent bandwidth issues of the shared memory bus for the presscotts??? Or can we say an 4800+ X2 which is 2.4ghz 1mb each core would be like 2 3.76ghz PDs??? Seems like that may still be up for the AMDs....

Is the Venice core have any advantage over the Skt 939 512kb L2 Athlon A64?

 

[Sylvanas]

I remember someone on these forums gave a good analogy of how AMD owns Intel a while back and it went a little like this: You need to transport a load from point A to point B. Mr AMD gets in his 18 wheeled semi trailer and carts the whole load in one go, sure it may be slower but it gets the job done. Meanwhile Mr Intel gets in an Indy car and transports one package at a time, therefore taking longer. Pretty lame but I understood it when I was new to all this.

 

[trungthiendo]

intel or amd...doesn't matter which one u want, get the one that u appeals to you,

3.2E prescott run hot but they do hit 4ghz fairly easy if you have an E0 stepping and with good ram

get whatever u can cheapest

 

[Zebo]

Originally posted by: Duvie
Got an idea on a dual X2 setup Zeb??? I mean how do you calc in the inherent bandwidth issues of the shared memory bus for the presscotts??? Or can we say an 4800+ X2 which is 2.4ghz 1mb each core would be like 2 3.76ghz PDs??? Seems like that may still be up for the AMDs....

Is the Venice core have any advantage over the Skt 939 512kb L2 Athlon A64?

I could do it Duvie with available data at Anandtech and TechReport in excel but since the chip is'nt really released yet I'm not to enthusiastic. Lots of time for not the real deal. But i definitly will and make file available when X2 comes and it's tested proper. Real ram on both platforms, real mobos.

I don't see any advantage to venice from a clock for clock standpoint over winchester, do you? The winny does a couple percentage points better in some things Venice in other things..But lets get real here, well within margin for error. The SSE3 just aint all *that*, look at the DivX and Mainconcept MPEG Encoder, hardy anything off it.
http://www.xbitlabs.com/articles/cpu/print/athlon64-venice.html

 

[ahock]

This is a real big tradeoff from Intel but they had log pipeline thinking that they can scale the frequency not knowing that they will hit the power waal eventually. But say without this power wall I think Intel already reach 7 GHz to 10 GHz (not sure) by this time and could have kick AMD ass all throughout. AMD have seen this power wall problem that is why stick to their design. Another thing is basically Intel didn't abandoned the design they left from P3 which is simialr to AMD's design right now. They made some improvements on their P3 which we now know as PM

 

[Lithan]

Ok, here is exactly as simple as it can be made.

A64's are better processors that run at lower MHZ.
Prescotts are worse processors that run at higher MHZ to make up for it.

Right now, A64's don't only beat Prescotts in performance, but also in design, since their architecture is so much newer and more revolutionary (and prescotts biggest architecture change over northwood was that they do less /mhz... not exactly a good change.)

 

[ahock]

Cant blame them if this doesn't screwed up due to power problem.... I think they are making more money due to provious concept of more clock = powerful processor. Imagine if they released the Tejas which as far as I know can reached up to 10MHz plus the Centrino based system then by all means I think they are a 40 Billion dollar company by now.

 

[Quentin]

Hey, AMD is missing out on a great advertizing campaign. They could have the Jolly Green Giant walking alongside one of those stupid little blue men from the Intel ads. "Greenie" would be ambling along at a leisurely pace while the blue guy is running flat out and sweating like crazy to keep up. Then they pick up a bag of groceries at the store and "AMD" keeps going while the blue guy finally passes out.

It might help explain how Intel needs high frequency to do the same work of a more powerful CPU that alos has lots more reserve built in.

 

[clarkey01]

Jesus, another thread on this !

 

[Duvie]

Originally posted by: ahock
This is a real big tradeoff from Intel but they had log pipeline thinking that they can scale the frequency not knowing that they will hit the power waal eventually. But say without this power wall I think Intel already reach 7 GHz to 10 GHz (not sure) by this time and could have kick AMD ass all throughout. AMD have seen this power wall problem that is why stick to their design. Another thing is basically Intel didn't abandoned the design they left from P3 which is simialr to AMD's design right now. They made some improvements on their P3 which we now know as PM

The power wall was called physics!!! It is called more specifically "transistor leakage" and they both are or will suffer it at around the same range. The physics of it is the porocess and materials being used can not yet stop the amount of leakage that is occuring. This leakage inturn results in heat as well as the cpu maker having to up the current to get it stable at those speeds leading to even more heat. fact of the matter is that Intel when they shrunk to 90nm should have realized a reduction in power yet even with the voltage drop they raised the actual power consumption...Perfect signs of transistor leakage as well as the higher heat output versus the older 130nm process northwoods....

Now AMD may eventually suffer this since they both are using the similar SOI and SS processes....However expect AMD to probably suffer it near the same 3.4-3.8ghz range. So AMD has more headroom for its single core design and thus will only pull away evenmore in this single task apps..intel has no choice but to get those software developers to quickly adopt SMT or it is going to get real ugly in certain areas...That is it. Intel is done there will be no 4ghz chip so when the FX57 arrives it only jst builds on existing leads. The FX59 near years end and we are pulling away further.

AMD didn't need dual core yet anyways but is forced to ramp up sooner by a desparately seeking INtel to get some assemblance back as a leader.

 

[imported_whatever]

Originally posted by: Duvie
Now AMD may eventually suffer this since they both are using the similar SOI and SS processes....However expect AMD to probably suffer it near the same 3.4-3.8ghz range.

Actually, one of the big reasons that AMD is not having leakage seems to be SOI - Intel is just using Strained Silicon, AMD is using Strained Silicon On Insulator, which reduces the leakage (the On Insulator part does)

 

[Duvie]

Originally posted by: whatever

Originally posted by: Duvie
Now AMD may eventually suffer this since they both are using the similar SOI and SS processes....However expect AMD to probably suffer it near the same 3.4-3.8ghz range.

Actually, one of the big reasons that AMD is not having leakage seems to be SOI - Intel is just using Strained Silicon, AMD is using Strained Silicon On Insulator, which reduces the leakage (the On Insulator part does)

OK...What do you think was done to give the 6xx series a slight better advanatge over the 5xx series in heat issues...Is it all just tricks or did they do something (refine process) to reduce the leakage which in return reduced the power consumption and thermal output....and that was with adding more l2 cache which I have always heard adds more heat with the added transistors...

 

[clarkey01]

Originally posted by: Duvie

Originally posted by: whatever

Originally posted by: Duvie
Now AMD may eventually suffer this since they both are using the similar SOI and SS processes....However expect AMD to probably suffer it near the same 3.4-3.8ghz range.

Actually, one of the big reasons that AMD is not having leakage seems to be SOI - Intel is just using Strained Silicon, AMD is using Strained Silicon On Insulator, which reduces the leakage (the On Insulator part does)

OK...What do you think was done to give the 6xx series a slight better advanatge over the 5xx series in heat issues...Is it all just tricks or did they do something (refine process) to reduce the leakage which in return reduced the power consumption and thermal output....and that was with adding more l2 cache which I have always heard adds more heat with the added transistors...

I think 3.2Ghz will be the limit for the K8, thermals will be in prescott country by then anyhow. And with the dual core and 65 nm process, 3 Ghz tops will be the limit for dual core. Im only guessing but what the hell.

 

[Duvie]

Yeah it is all speculation on our part on that anyways since AMDs roadmap doesn't specifically outline target speeds...

3.0ghz would be phenomenal enough and insure nothing INtel could offer off of the prescott line in forms of smithfield will ever gain the leads back...

 

[clarkey01]

Originally posted by: Duvie
Yeah it is all speculation on our part on that anyways since AMDs roadmap doesn't specifically outline target speeds...

3.0ghz would be phenomenal enough and insure nothing INtel could offer off of the prescott line in forms of smithfield will ever gain the leads back...

Makes you think though, if prescott and tejas and even nehalem have made it (Prescott getting to 5 ghz) then AMD would have been in trouble and would have had to ramp instead of taking thier time and stretching out the PR's on chips.

Hell hector would of been at FAB 36 with a whip " I need 65 nm process online by tonight or smack bottom time". I would like to see the K8's and K9's headroom with the 65 nm process, Im guessing they a good idea since there stuff is now installed.

 

[clarkey01]

Fairly sure Intel projected Nehalem 10.25 Ghz for 2005......And we'r at 3.8 Ghz...

If they had nehalem out AMD would of been in trouble.

Hector must be saying in thier prayers " Thank God for prescott"

 

[ahock]

I agree clarkey01 if nehalem Tejas had just live to its expectation I think AMD is now in big trouble. Problem is they knew that Intel with their current process will eventually hit the power wall.... I'm not sure but I think this is the reason why they design their chip to dual core already from the beginning. They knew all along that their current (AMD) process will not scale its frequency. Remember IBM is first to come with the dual core and AMD and IBM have the same process tech.

As far as AMD getting ready their 65nm and just behind Intel with this?.... I guess in my opinion, this is not the case.... They haven't even fully utilized their 90nm..... and as of now they are still using the 130nm..... yes they have some but question is, are they capable of ramping it up for volume demand? AMD is so dependend with IBM in terms of process tech.... Intel already have the 45nm in their roadmap......

 

[Ready]

How does a 200MX Cyrix compare to a P4 3200 or an AMD 3200+
I think the Cyrix will kick the crap out of both of them end of story!

And the 200Winchip will kick the AMD and Intel butt also!