Intel responds to Barcelona benchmarks

[Deleted member 3]

That is a new one to me, about AMD building chips for Intel (while back) and also about a few folks from AMD coming to Israel to work on a second chip for Intel.

The story reported, in MaximumPC and CPU magazine and at least a dozen others, is that Intel, shortly after AMD released the 64 processor, had two independent teams working on the next P4 successor chip (or whatever PX chip it was). One team came with Netburst, and of course, we all know how that ended. The other team started over from scratch. Something unheard of. They built off the centrino technology and pentium M technology and thus C2D was born. The only thing Intel lacks to completely compete with AMD is integrated memory architecture. Apparently, this is coming in the Penryn. AMD better take a few queues from Intel and try some of their new architecture out.

Centrino isn't a technology, it is a label for a system that has a certain set of hardware. Pentium M wasn't exactly from scratch either, it's based on the Tualatin. As for AMD making Intel chips, new to me as well. They did make Intel clones until Intel launched the Pentium. They kept using it's socket for a long time though.

 

[suraswami]

I think they made their 386 & 486 processors. Some had encryption of 'AMD' on it. Google for history of AMD and Intel and you can find answers to this.

 

[GJSNeptune]

That is a new one to me, about AMD building chips for Intel (while back) and also about a few folks from AMD coming to Israel to work on a second chip for Intel.

Check out this thread, or this reply to that thread.

 

[suraswami]

Ha Ha our friends already wrote about this few days ago!!!!

 

[v-zero]

That is a new one to me, about AMD building chips for Intel (while back) and also about a few folks from AMD coming to Israel to work on a second chip for Intel.

The story reported, in MaximumPC and CPU magazine and at least a dozen others, is that Intel, shortly after AMD released the 64 processor, had two independent teams working on the next P4 successor chip (or whatever PX chip it was). One team came with Netburst, and of course, we all know how that ended. The other team started over from scratch. Something unheard of. They built off the centrino technology and pentium M technology and thus C2D was born. The only thing Intel lacks to completely compete with AMD is integrated memory architecture. Apparently, this is coming in the Penryn. AMD better take a few queues from Intel and try some of their new architecture out.

Close. Intel set two teams running in parallel to design the next architechture to come after the Pentium-3 set. The Israeli team started from a base on the P6 architechture, the other team started almost completely from scratch in designing the Netburst architechture - there was 'some' previously designed basis for Netburst though.
Intel went with Netburst, as preliminarily it portrayed a product whose scalability far outreached the P6 based design. However, they soon found a use for the Israel based designs in Laptop computers.

We all know the story from here. K7 fell to Netburst (just) but AMD's tweaked K7 (K8) had a much better IPC than Intel could cope with, especially since Netburst fell short of their original expectations after it was revealed that the ultra-long pipeline shift and over-complex design lead to massive electron leakage, and reduced scalability.
And so, seeing this Intel turned to the Israeli team and said: "We want Pentium-M, but with two cores, in a desktop - and we want a slightly longer pipeline to give us much better clocks" - and that's what Core2 became (well, massive oversimplification, but that's the gist).

 

[Fox34]

So the core2 is a modified M with long pipe lines..but the core 2's perform extremely well at a much lower clock than a higher clocked AMD, so adding the longer pipe line to give a higher clock doesn't make to much sense at seeing the Core2 default clock speeds, but after seeing the over clocks that people have done with the core 2's, that makes it sensible.

 

[a111087]

what about the power consumption, 65W isn't 150W, AMD's 16 cores consume way less that intel's 8 cores

 

[Fox34]

Yea I was gonna say something about that, those Intels with freakin cook

 

[bigboi86]

K7 kicked Netbursts ass. AXP's were way better, except for multitasking, which intel had hyperthreading.

 

[WarEagleAU]

so I wasnt too off base with my memory recollection. Still AMD could learn a few tricks from Intel with its C2D stuff. I think we will see great improvements from both camps.

 

[Fox34]

Think AMD will start making their architecture somewhat like the core 2?

 

[EastCoasthandle]

I thought C2D used shorter pipeline then P4??

 

[Fox34]

hmmm, we should verify if it does or not

 

[kwchang007]

I thought C2D used shorter pipeline then P4??

yes it does c2d uses a 14 stage pipeline while the p4 uses a 30 stage pipeline

 

[trt740]

I highly doubt RAM size makes that much of a difference. Plus, you should also consider that Clvoertown has been available for half a year already and Barcelona isn't as of yet. Shortly after Barcelona Intel will also introduce the new Xeon MP's, which should be Barcelonas target. Tigerton will have the same amount of cores and features CSI. At that time both should be compared, since those 2 platforms are in the same market.

Yes the ram would make enought difference would most likely make them run neck and neck. We are talking 900 pixels after all.

 

[Fox34]

So the core 2 has a smaller pipe line, why would they do that? And that sounds like the reason they have the lower overall clock.

 

[[I.R.A]_FBi]

for what u get for ur money, it surely kicks ass

 

[Fox34]

Bah, I'm an AMD fan boy though

 

[kwchang007]

So the core 2 has a smaller pipe line, why would they do that? And that sounds like the reason they have the lower overall clock.

core 2 duo has a "small, fat pipe" which is much better. a longer pipe can run a higher clock, yes but a short pipe has many more advantages. a short pipe takes less time to "fill" b/c it is shorter (im kind of peicing what i have heard together, if i am wrong, please tell me) and it takes less clocks to load the pipe up. also, when the branch predicter predicts the wrong next insturction, the whole pipe has to be flushed, so a long pipe her is worse. also, core 2 can process 4 instructions at once compared to pentium m and p4's 3 instructions at once, plus the whole macro ops thing.

 

[Fox34]

Yea the shorter pipeline with a higher "bandwhidth" makes sense of the core 2, lower clock but still high performence